Çompytational DesignThinking

Edited by Achim Menges and Sean Ahlquist

The current transition from Computer-Aíded Design (CAD) to Computatíonal Design m
archítecture represents a profound shift in design thínking and methods. Representatíon
is being replaced by simulation, and the craftíng of objects is moving towards the
generatíon of íntegrated systems through designer-authored computational processes.
Wliile there is a particular history of such an approach in architecture, its relatíve
newness requires the continued progression of novel modes of d.esign thínkíncr for
the archítect of the 21 st century. This AD Reader establíshes a foundatíon for such
thínking. It íncludes multifaceted reflections and speculatíons on the profound. ínfluence
of computational paradigms on archítecture. It presents relevant principies from the
domains of mathematícs and computer science, developmental and evolurii
bíology, system science and philosophy, establishing a discourse for computational
desígn thinking in architecture. Rather than a mere technícal approach, the book
discusses essentíal intellectual concepts that are fundamental for both a discourse on
computatíonal design and its practice.

This anthology provídes a unique collection of seminal texts by authors, who have
íntroduced tíie seeds and key developments for a computatíonal approach to design,
thus playing a considerable role in shapíng the field. An ímportant aspect of this book is
the manner in whích adjacent fíelds and historical texts are connected. The sources of
both original ínspiration and scientific thought are presenteei alongsíde contemporary
writíngs on tíie continually evolving computatíonal desígn discourse. Emerging from tïie
field of science, principally the subjects of morphogenesis, evolution and mathematics,
selected texts províde a historical basis for a reconfigured mindset of processes that
generate, arrange and describe form. Juxtaposed against more contemporary statements
regarding the influence of computation on design Üünking, the book offers avenues for
establishing novel thought processes for architecture, theoretícally and practically.

• The first reader to províde an effective framework for computational thinking in design.
• Includes classic texts by Christopher Alexander, Ludwig von Bertalanffy, Johann Wolfgang
von Goethe, Ernst Mayr, Nicholas Negroponte, D'Arcy Wentworth Thompson, Gordon Pask
and William Mitchell.
• Encompasses significant prevíously published texts by Peter J Bentley and David W Corne,
Jòhn Prazer, John H Holland, Sanford. Kwinter, Achim Menges, Kostis Terzidis and
Michael Weinstock.

Professor Acliim Menges is Director of the Institute for Computatíonal Design at

Stuttgart University. Currently he is also Visiting Professor in Architecture at Harvard
University's Graduate School of Design and Vísitíng Professor for the Emergent
Technologíes and Design Programme at the Arctiitectural Assocíation in London.
Edíted
Sean Ahlquist is a research associate at the Institute for Computational Design
at Stuttgart Uníversity. He hás taught at the University of Califórnia Berkeley and
Califórnia College ofthe Arts and founded the design firm, Proces2. He holds a Master
Acliim
of Architecture degree from the Emergent Technologies and Design Programme at the
Architectural Association in London-
and Sean

^
I S B N 978-D-470-66565-7

FSC
MixedSources
Product groupfrom tí&ll-mana^&d
forestsandothei

Cnt na, n-COÍ.C[U785

9 11"78"0"47Õ-"66-5657
wiley.com
6UD|UIL|1 u6lS9Q
lèüoiíeindLuo^
READER

Çomputatipnal
Design Thinking
Edited by
Achim Menges
and Sean Ahlquist

®WILEY
A Jbhn Wiley aiid Sons Ltd, PubUcaúon
To our families, Eva and Elena, Elaine and Ara, for their unending patience and support Acknowledgements

Thls book is inspired bythe stimulating discoijrse ofthe camputational design cammunitywe were fortunate
This edition fírst pubhshed 201 1 enough to be part of over the last few years. As there are too many individuais that have genuinely influenced
© 2011 John Wiley & Sons Ltd and encouraged ourwork on the bookto be named here, we would !ike to thankthem collectively. We would
also like to offer aur síncere gratitude to Helen Castle for making this book possible and for her unwaveríng
support of this project. In addition, we wou!d !ike to extend our deep gratitude to the authofs, their heirs, estates
Registered office and transiators for their kínd permissions to let us re-/publísh some of the most decísive texts on the topic at
John Wiley & Sons Ltd, TheAtrium, Southern Gate, Chíchester, West Sussex, P01985Q, United Kingdom hand. We sincerely regret that acquiríng the rights for reproductíon of a number of key illustrations of some of
the reprinted essays was prohibitively expensive. We hope that our readers will be willing to refer to the original
publications in such cases, or utilise other reference materiais, printed and electranic.
Fordetails ofour global editorial offíces, for customerservices and for information about howto applyfor
permission to reuse the capyright material In this book please see our website at www.wiley.com. The author and the publisher gratefully acknowledge the people who gave their permission to reproduce material
in this book. While every effort hás been made ta cantact copyright holders for their permission to reprint
This is a collective work. The copyríght ownership of individual extracts is indicated in a copyright notice at the
material, the publishers wouid be grateful to hear from any copyright holder who is not acknowledged here and
end ofeach text. will undertake to rectify any errors or omissions in future editions.

Ali rights reserved. Ho part of thís publícation may be reproduced, stored in a retríeval system, or transmitted, in
any form or by any means, eiectronic, mechanical, photocopying, recording or otherwise, except as permitted by
the U K Copyright, Designs and PatentsAct 1988, withoutthe prior permission ofthe publisher Editorial Note on Presentation and Editing ofTexts

Wiley also publishes its books in a variety of electronic formats and by print-an-demand. Some content that
This anthology aims to be inclusive, therefore some texts must appear in abbreviated form in the course of
appears in standard print versions of this book may not be avaílable in other formats. For more information about
presenting others in their entirety. Some texts have been edited to shorten them, to exclude references that
WIIey products, visit us at www.wiley.com. require more space for a fui! explanation and to preserve the fíow of argumenL Information for the sources of the
edited and complete texts is given in the copyright line after each text, The term 'From' preceding a copyright
Designations used by companies to distínguish their products are often clainned as trademarks. Ali brand names
line signifíes that a specifíc extract or extracts hás been taken from a longer text. Otherwise texts are given in
and product names used in this book are trade names, service marks, trademarks or registered trademarks of
their entirety or edited ta indicate the argument af the whole.
their respective owners. The publisher is not assodated with any product or vendar mentioned in this book. This
publication is designed to provlde accurate and authoritative informatian in regard to the subject matter covered.
Texts have been clearly marked to show where they have been edited. The fallowlng conventions have been
It is sold on the understanding that the publisher is not engaged in rendering professlona! sen/íces. If professional
used throughout. Suspended points '...' are used to denote amissions ofwords or phraseswithin a sentence.
advice or other expert assistance is required, the services of a competent professional should be sought.
Suspended pointswíthin a square bracket [...] are used to denote omissions extending from a complete sentence
to a paragraph. Three asterisks * * * mark omissions of more than one paragraph, and may denote exdusion af
Executive Commissioning Editor: Helen Castle
a complete subdívision ofthe original text, A paragraph may end [...] ifthe last sentences ofthat paragraph are
Project Editor: Miriam Swift
omitted or ifthe following paragraph is omitted. A paragraph may also start with [...] if one or more sentences at
Assistant Editar: Calver Lezama
the beginning afthat paragraph have been omitted or if a previous paragraph hás been onnitted.Typographical
errors and errors of transcription have been corrected when discovered in the anthologised texts. Otherwise
ISBN 978-0-470-66570-1 (hb)
idiosyncrasies of speHing or punctuation remaín unchanged.
978-0-470-66565-7 (pb)

Notes and references have been included onlywherejudged to be necessaryto thetext as printed. In some cases
Designed byArtmedia, London
text notes have been renumbered for this edition, in which case a note to this effect Is given after the copyright.
Printed and bound in Great Britain byTJ International, Ltd, Padstow, Comwall, an 15014001 Envíronmental
Management System Accredited Company, using vegetable-based ink and FSC paper

Frant cover image ® Achim Menges and Sean Ahlquist

/)" Mixed Sources

jroup Ïïom^slJ-mdnctged
á oth&rcflnfroHstísouïïËS

FSC org Certnu. ses.coc.i-u!

120 An Introduction to Creative Evolutionary Systems
Achim Menges and Sean Ahlquist Peter J Bentley and David W Corne

io INTI 131 Constrained Generating Procedures

Computational DesignThinking John H Holland
Sean Ahlquist and Achim Menges

142 RealVirtuality
30 Formation andTransformation Manuel DeLanda
Johann Wolfgang von Goethe

149 A Natural Model forArchitecture

32 On theTheory ofTransformations, orthe Comparison of Related Forms John Frazer
D/Arcy Wentworth Thompson

158 Morphogenesis and the Mathematics of Emergence

42 Variational Evolution Michael Weinstock
Ernst Mayr

168 Philosophy of Mathematics for Computational Design:

50 The Meaning of General System Theory Spatial IntuitionVersus Logic
Ludwig von Bertalanffy Jane Burry

sá Systems Generating Systems 179 Associative Design: From Type to Population

Christopher Alexander PeterTrummer

68 The Architectural Relevance of Cybernetics 198 Integral Formation and Materialisation:

Gordon Pask Computational Form and Material Gestait
Achim Menges

78 Towardsa HumanismThrough Machines

IMicholas Negroponte 2ll The Computational Fallacy
Sanford Kwinter

86 A New Agenda for Computer-Aided Design

WilliamJ Mitcheil 216

218 U
94 Algorithmic Form
KostasTerzidis

102 Architecture and Practical Desígn Computation

Mark Burry
Preface
Computational design comprises both an immense chance and a considerable challenge in this spedfíc context. This became even more apparent to us when we were in the
to architecture and related design disciplines. Most often, the challenge is understood to fortunate situation of acquiring and viewing hundreds of related titles for the library of
be merely of a technical nature, as computational design requires acquiring scripting and our newly founded Institute for Computational Design at Stuttgart University, whi.le at
programming skills that traditionally do not form part of the disciplinei repertoire and the same time building up the educational cumculum. Based on the realisation that the
education.There are a vast number of books addressing this aspect, both by equipping the seminal texts, from which a conceptual model for criticai computational design thinking
interested designers with the related coding skills as well as exposing them to reference in architecture may originate, are scattered across a vast body of literature, we attempted
projects from academia and practice. Yet, over many years of teaching computationgl to collect/ filter out and selectively sample the most infíuential pieces by authors that
design, we have realised that the main challenge does not lie in mastering computational have either provided a foundation to which a computational approach to design could
design techniques, but rather in acculturating g mode of computational design thinking. be pursued or have played a considerable role in shaping the fieid. We believe that this
A syndrome of this intellectual struggle may be the fact that in contemporary practice, requires connecting to adjacent fíelds and includlng historical texts that present original
withveryfewexceptions, design computation still merely serves the purposeofextending concepts on thinking computationallyjuxtaposed against more contemporary statements
or accelerating weil established design processes. This, arguably, can be further traced thatareinfíuentíaltotheformation oftoday'sdiscourseon computational designthinking
into the traditional paradigms of design education, based upon a model of investing in architecture. The resuit is the book in your hands.
skills with an overlay of theoretical and practical application. We fínd the capadties We do not consider the book as an afl-encompassing survey, but rather as point of
of computational design are only fully unfolded at the convergence of exerdsing both departurefora continued progression of novel modesofdesígnthinking forthe architect
computatíonal thinking and practice. ofthe21stcentury. Duetothecomplex nature ofthe concepts presented oneshould not
Computational design, when aiming to tap into Íts profound architectural potential, expect a linear progression of arguments, but rather a fíeld of multifaceted reflections and
requires gnd enables us to rethink some central cfisdplinary concerns. The intellectual speculations on the design possibilities arising from computation. We hope the collection
foundation for the nature of computational design rests at a particular confíuence of wil! contribute to establishing a foundation for such thought processes in architecture.
domaíns in fields ranging from mathematics, computer science and systems science to
bioíogyand philosophy.Wefound, however, that no bookcoveredthisspectrum oftopics Achim Menges ana Sean Ahíquist, 2011

8 ComputationaL Design Thinking 9 Preface

'^'-'-.-^ï

Introductíon by which the computer hás been integrated Ínto architectural design. Fundamentally,

Computationat Design Thinking the distinction rests in the approach towards design, rather than in particular skil!
sets or knowledge. A computer-aided approach assumes an object-based strategy for
Sean AhLquist and Achim Menges encapsulating information into symbolic representations - methods of organising
information. in contrast/ a computational approach enables specifíc data to be realised out
The manifest form - that whích appears - is the resuit of a computationai of initial abstraction - in the form of codes whlch encapsulate values and actions.4These
interactíon between intemal rules and externai (morphogenetíc) pressures that, distinctions are refíected back in the philosophical perspectives by which architecture,
themselves, origínate in other adjacent forms (ecology). The (pre-concrete) bom of computer-based processes, is perceived.
internai mies compr/se, in theír activity, an embedded form, what is today clearly One criticai aspect of working in a computational manner is the processing of
understood and descríbed by the term aigoríthm. (Sanford Kwinter)1 infomnation algorithmically. The idea of the algorithm can be traced back to the work of
the Norwegian mathematician Axel Thue, who invented a rewriting program operating
Computation hás a profound Ímpact on both the perception and realisation of architectural on strings of sígns in 1914.5 While it was later demonstrated that Thue's problem was
fonn, space and structure. It shifts the way one perceives form, the way in which form actually recursively unsolvable,6 his dedsion-making procedures encapsulated ali
is purposed, and the way in which form is produced. The fundamental concepts which basic aspects of algorithmic computing. Most generally, one can extract from this the
underlie computational theory and techniques expose form as a subsidiary component definition of an algorithm as a set of procedures consisting of a finite number of rules,
of environment, and environment as a complex web of energies in dynamic exchange of which define a succession of operations for the solution of a given problem. In this way,
both regeneration and degeneration. The emergence of this perspective hás come at the an algorithm is a fínite sequence of explicit, elementary instructions described in an
confíuence of different modes ofthinking, in alignment with various scientifíc, technofogícal exact, complete yet general manner. The application and execution of algorithms on a
and cultural advancements spanning several decades if not centuries. Spedfically, computer happens through programming languages/ which enable computing procedure.
computation in design represents an accumulation of multilayered concepts ranging from This is a fundamental property of computation as a technical achievement, but also as a
systems theory and cybemetics, to morphogenesis and developmental biology. This book theoretical framework in which systems of virtual and physícal existence are culminations
presents texts which characterise the initial foundations of these concepts in biology, of methodically developed informatlon sets.
mathematics, computer science, systems science and phifosophy, along with contemporary To fully clarify the defínition of computation, it must be placed within the context
translations of such to the application in architectural design. The intention of this outlay of architectural practice, theory and technology. The first graphical systems for computer
of concepts, methods and perspectives is to establish a multifaceted yet comprehensive design were developed in the 1960s. The initial premise was to mimic the predictive
description of what computational design thinking may constitute in architecture. possibilities that, at that time, were being well practised in the field of structural
engineering. What was soon realised was that optimisation, as a primary objective, was not
What is Computation? an applicabie approach glven the multiplidty of constraints, requirements and intentions
that envelope an architectural problem. Termed automated design, there was a holistic
The dominant mode of utiiízing computers in architecture today is that of vision of computer programs which could process complex design bríefs Ínto specific
computerization; entities or processes that are already conceptuaiized in the architectural solutions.7 This alluded to the question of how a computer can, or even
designeis mina are entered, manipuiated, or stored on a computer system. In should, mimic human thought processes and ingenuity. This carne at the same moment
contrast, computation or computing, as a computer-based desígn too! is generally at which cybemetic theory was popularised. Cybernetics advanced the notion of systems
Hmited. The probiem wsth thís situatíon is that designers do not take advantage theory to address the new existence of the man-machine relationship, stimulating the
ofthe computatíonat power of the computer. (KostasTerzidis)2 notion of how computers may be utilised to expand the human Íntellect, rather than
work redundantly to the processes which form such knowledge. What arose, in short, was
To understand computation, and its relevance for architectural design, one must a perspective at which architecture could be both perceived and pursued as a 'system7.
understand the distinction between computation and computerisation. In principie, thís This impiies that the formulation of architecture occurs through the understanding and
can be broken down as methods which elther deduce results from values or sets of values, development of the complex inten-elation of material parts, and social engagement/
or simply compile or associate given values or sets of values. One increases the amount shaping form, space and structure.8
and specificity of information, while the other only contains as much Ínformation as is Initially, graphical systems such as Sketchpad, GRASP and LOKAT developed as
initíally supplied.3 This basic conflictíng premise is ever present in the various methods early as the mid-1960s were modelled upon a 'systems-basecT approach. Sketchpad,

10 Computational Design Thinking 11 Introduction: Computational Design Thinking

Howthiswasrepresentedinarchitecturaltheon/wasashiftfromtheviewofarchjtectureasa
material objectto architecture as a system comprised ofand workingwith a series ofinterrelated
systems. Sketchpad unfolded a logic to capture interrelated geometries, and how associations
cause ripple effects in the development and manipulation of form. The description of design
was no longer symbolically the represe ntation ofthe physical elements as they lie against each
other but rather a summation of the forces, pressures and constraints which realise the form.
The first experiments with computation in design were typically based on specific, singular
design briefs. Overtime, astheseinitial programs were to address a largeraudienceofdesign
scenarios, they became more biased towards methods and objects which were standardised.11
Aithough this hás been the conventionai trajectory of software development in design, the
roots for a computational approach were seen already in these very early experiments.

It might turn out that ali that chãos is the natural consequence of informatíon
overioad, in whsch case the power of information-processing machínes might
AchimMenges, Morphogeneüc Design Experiment 01, 2003, Overseven growthsteps,
two correlated surfaces are computationally generated as part of a morphogenetic and prove usefuí. CMurray Mifne)12
evolutionary design process based on the construcüonal constraints ofthe system. The
computational growth process employs mathematical rewritmg mies based on Hemberg-
Extended Lindenmayer Systems. © Achim Menges.
5o what is computation, then? In reiation to design, computation is the processing of
information and interactions between elements which constitute a spedfíc environment,
thepivotaiwordbeíng/nteroctíons.Thisdefínitionpullscomputationoutofonlythe virtual

developed by Ivan Sutherland at MIT, applied the idea of constramts which could be realm. In a greater relevance to design and the designer, the most general application

varied to test and flex the relationships between geometries in the formation of an overall of computation is in produdng outcomes realised from the processing of internai and

system. Geometries could be drawn as parametric instances representing different aspects externai properties. Computation provides a framework for negotiating and influencing

of forrn, space and stmcture. Sketchpad was the first system to allow this method of the interrelation of these datasets of information, with the capacity to generate complex

variation and design formation to occur through a graphical input A light-pen would order, form and structure. Computation as a design methodology is to formulate the
spedfic. Where computer-aided processes begin with the specific and end with the objecf
register points of geometry creation and geometry dependencies. Eric Teicholz, working
with some of SutherlancTs graphical methods, developed GRASP at HarvarcTs Craduate computational processes startwith elemental properties and generative rulesto end with

School of Design. While the software was concemed, like others, primgrily with space infomnation which derives form as a dynamic system.

planning/ Ít was already a generative system utilising random form-making procedures

and constraints based in the specífíc interrelation of structure, solar exposure and SystemsThinking
programmatic organisation. LOKAT, also developed in conjunction with Harvard CSD,
interjected evaluation into the process defíning 'goodness of fíf according to mies of Every aspect of form, whether piecelike or patternljke, can be understooá as a

program associations and proximity. Sketchpad introduced an interface while Ínstítuting structure of components. Every object is a hierarchy of components, the iarge ones

a methodology for maintaining rule-based geometric assodations.9 Among these design spedfying the pattern of distribution ofthe small ones, the smaí! ones themseives,

environments, Sketchpad was exemplary in defíning such fundamental computational though at fírst sight more cleariy píecelíke, Ín fact again patterns spedfying the
arrangementand distribution ofstiil smalter components. (Christopher Alexander)13
methods as parametrics, associativity and rule-based systems generation. Its interaction
posed a breakthrough for the cybemetic view of the man-machine relationship, and
synthesis of process as a mirrar to the working ofthe resulting system. AristotleJnhisdefinitionofHolismJaysthefírststoneinunderstandingsystemsasawhole
which is more than the sum of its paris. He proceeds, though, to understand the parts in
isolation. Descartes added to this view the notion of understandlng processes via simple
Constmction of a drawing with Sketchpad Ís itself a model of the design process.
The locations ofthe poínts and Unes ofthe drawing mode! the varíabies ofa design, causality, with relationships deduced from/ at most, two interacting elements. In the mid-
20th century, Ludwig von Bertalanffy introduced general system theory as a response to
ana the geometric constraints applied to the points and Unes ofthe drawíng modei
the long evolution ofthe conception ofthe ywhole/offunctioning systems. Histheoryis
the design constraints which limit the values ofdesign variables. Clvan Sutherland)10

13 Introduction:ComputationalDesignThinking
12 Computational Design Thinking
based on a fundamental critique on dassic physics and its deductive methods and focus
on Ísolated phenomena. Bertalanffy considered such methods as unsuitable for biology,
reasoning that nothing in nature exists in isolation or simple dependencies, but rather
needs to be understood as complex systems of interactions and reciprodtíes. Beyond
simple and linear causal relations/ Bertalanffy proposed the concept and mathematical
operations to define systems based on nonlinear causalities. Examples of these are the
causal loops called 'feedback', that are the processes involving mutually stabilising
causes. Here, the effects react back on theír causes, and thus open up the possibility to
conceptualise what Bertalanffy defines as an yopen system', one of elemental matter and
processes in reciprocal exchange between organism and environment. This was a model
which could ascertain global behaviour by understanding the full collection of individual
elements and thelr Íntimate interactions.

Anopensystem isdefinedasasystemjnexchangeofmatterwithítsenvironment
presenting jmport ana export, buiiding-up and breaking-down of its material
components. (Ludwig von Bertalanffy)14

General system theory was formulated as a universal scientific discipline to dedpher

the laws which rule the definition of 'organised wholes'. Folded into architecture,
as exemplifíed by Christopher Alexanderrs writings on design process gnd physical
organisations/ form and functionalíty in the 1960s, the theory implied profound changes
to design thinking. Contraryto established design approaches, Alexander argued that an
overail design problem cannot be divided into sub-problems, and consequently, that it is
ímpossible to arrive at a novel design solution as a summary process of solving individual
problems one afterthe other.
Alexander describes a system as that which focuses on an overall behaviour
accomplished through the Interaction among partsr. The use of the words among paris
is criticai in this defínition as it states that the knowledge of the components hás to be
complemented with the knowledge of how the components Ínteract, whether it be in
competition ar in contiguity. As defíned initially byAristotle and later expanded upon by
Bertalanffy, systems do not function simpiy as the summation ofthe whole. The holistic
nature cannot be seen in the individual part, nor can it be seen with the addition of its
parts. The system behaviour emerges only in the dynamics of the Ínteractions of the
parts. This is not a cumulative linear effect but rather a cyclicai causal effect in whích the
complexity ofthe levei and amount of interaction cannot be directly deciphered.
The introduction of system theory in architecture and the consequential focus on
systems thinking in deslgn applied a twofold shift. First, the shift signifíed a dismissal of
Marco Baur Sebastian Kron, Christian Schmidts, Christian Weitzel, EvolutionaryDesign Strategies
Project02, ICD Stuttgart University CSean Ahlqmst, AchimMenges), Stuttgart, 2009. Emulating the view of archltecture as comprised of entities in static isolation for one which defines
processes ofbiological development and evolution, this scripted process utüises an iterafive formastheculmínationof5ystemswhichinteractwithitscontextinmatter,physicalityand
Boolean technique and applies fitness of área to volume ratio, solid to void ratio, and degree of
geometnc 'coníourmg' to evolve geometnes across mulüple generations. In stnct rules ofselection,
personal engagement Second, this shift in architectural thinking introduced fundamental
ranking and breeding, the populations ofgeometries converge towards a highly effective outcome concepts for how the computation of such interrelational, complex behaviour-based
given the criteria applied. © Institute for Computational Design, Stuttgart University.
systems could be achieved.

14 ComputationatDesignThinking 15 Introdurtion: Computationat Design Thinking

Where cybernetics, evolving in paraliel with general system theory, contributes to the
concept of cyclic-causality is the understanding of how machines may be able to resolve
such complexity. In particular, the notion of feedback and stability provides functions 'l !'^Í'Ï\»--'T1-'..-'"..'. lí .".-;*!';

for detemniníng how the whole is organised. Balanced with the idea of 'goodness of fit', J^.; /-/ \ .'-.. .\ '^SÍ^3SK&'^^.>

prediction and the accuracy of prediction serve as criticai factors in what Norbert Wiener, /-!:,"
jiïïtí—.s^'
who coined the term "cybemetics', calls the leaming machine'.15
Fundamental to computational design is the understanding of how systems, as form
and as mathematical ordering constructs, operate. Fundamental to understanding their /'^lï^^^^^fê^^^^/ •?Ï~ "' "~"^Ï*r:^3^^Íllt^^?Ít ^^1
operation is the levei of prediction contained in the model which envisions the system.
Initially, this was a primary consideration for the fírst graphical systems in architecture,
stated as: "Wíth the help of computer-aided models, the designer will be able to predict
the performance of any design alternative he may generate.r1GThe success ofthís depends
upon design paradigms which are oriented towards detining systems with concise
parameters, and using the effort of prediction and feedback to inform, further specify
and define overall the behaviour of the organised system.

Parametric Dependencies

Coetheï formalism, like ali rigorous ana ínteresting ones, actuatly marks a
turning away from the simpie structure of end-products and toward the active,
ever-changmg processes that bring them into beíng. (Sanford Kwinter)17

Ocean Norüi and Scheffler+Partner

Long predatíng systems theory was Goethe/s doctrine of morphology. In 1796 Goethe Arehitects, Czech National
introduced the notion of morphofogy, outlining the criticai distinction between form and Ubrary Intemaüonal Competition
Entry, 2006. Through an analytíc
formation. Conceiving of gestalt, the individual form, as inseparable from bíidung, the
computational procedure, the
contínuous unfolding of formation and transformatíon, morphology sought to illuminate stress distribution within íhe
the processes that govemed form ratherthan form itself. Goethe approached this through envelope ofthe CzechNaüonal
LLbrary's cantilevering volumes
the study of physícal attributes as a registration of its development, trying to decipher is evaluated and mapped as a
its metamorphosis, and understandíng the mechanics of constant development and vector fíeld of principal forces
(tOp). According to this stmctural
formation in nature. HÍs description of the 'vertical tendenc/ that occurs in particular
ínformation, together with view
plants elucidates a set of dependencies in form and structure in an organismo activation axes and spatial characterisücs,
a network of merging branches
and suppression of specifíc movements. The geometric relationship in certain plants, as
is computationaüy derived
Goethe cfiscovered, expressed verticality só that it may have connection to the roots, and (centre), which is developed into
spiralled só that it may blossom, allowing for the possibility of pollination. Although he a cantilevering structural envelope
(bottom), ©AchimMenges.
thought that the actual process of genesis was driven by a divine force, he clearly laid the
foundation for linking geometric behaviour with a functional logic, and understanding
the general principies mathematically. Through Coethe/s concept of natural morphology
being in a constant state of becoming, form and the processes of formation and re-
formation were ali a part ofthe same interdependent system.

We pass quickiy and easiíy from the mathematical concept of form in íts statical
aspect to form in Íts dynamic relations. We rise from the conception of form to an

16 Computational Design Thinking 17 Introduction: Computational Design Thinking

understanding of the forces which gave ríse to it; and in the representation of form
and in the comparison ofkmdred forms, we see in the one case a diagram of forces in
equHibríum, and Ín the other case we discem the magnitude and the directíon ofthe
forces which have suffíced to convert the one form into the other. (D/ArcyThompson)18

The mathematical understanding of the relationship between form and formation was
solidified in the work of D7\rcy Thompson. Similar to Coethe, D'Arcy Thompson/s quest
was to define the abstract mathematical systems which underlie organic stmctural form
and their transformations. In difference, Thompson sought to define form through the
understanding of how physical forces produced structure and pattem.Theattemptwasto
devise the underlying set of geometric laws which shape form in relation to externai force.
Thompson offers a criticai component to the understanding of form in that Ít is a system
which organlses Ítself in the presences of both internai and externai forces, and that the
organisations can shape patterns traced through mathematícal rules.
What Thompson lays the groundwork for in terms of computation and design is
parametrics and associative logics. Parametrics might be understood, conventionally, as
regarding solely the interdependendes of certain geometrical constraints. But, Íts deeper
understanding is that of establishíng methods for Ínterrelating particular behavíours of
forms and forces, and how they might be represented as associated mathematical and
geometric rules.19 Thompson uses the Cartesían grid as the ultimate parametric device.
It is extensible and transfonnational in multiple dimensions/ globally and locally. It is
a coordinate system, and, ultimately for Thompson, it supplies a measure to devise
comparison between the geometries whích he studies.
One particulariy striking aspect of Thompson's mathematical concept of form and
transformationisthefactthatitdoesnotremainmerelyanalytical butthatitsuggestshow
the mathematical operations can become generative.The passage on the development of
an undiscovered, intermediate pelvis type between Archaeopteryx and Apatomis in his
Theory ofTransformatÍons' marks the criticai point, at which the natural transition from
reconstmcting knownfonnstogenerating unknownformswith mathematical rigouroccurs.

Generative Morphogenesis
Natural morphogenesis, the sequential biological events which define the development of
an organism from initial generation to mature system, derives complex organisation, form
and structure from the interaction of system-intrinsic material capacities and externai
influences and forces. The resultant astounding functional integration, performative
capacity and material resourcefulness emerges through morphological differentíation, the
summary process of each elemenfs response and adaptation to its spedfic environment.

Achim Menges, Morphogenetic Design Expenment 03 - Paper Strip Morphologies, 2004. WhileGoetheandD/ArcyThompsoncontributedtotheunderstandingoftransformational
Morphological ciifferentiation (top left) of a parametnc system ciefmed by^the physical laws which Ínfluence the expression of form, the full knowledge of biological formation did
constramts ofpaper-stnps (bottom) leads to varying performative patterns ofstructural
not come into existence until the discoveries of genetics. Morphogenesis and evolution
behaviour (to]3 centre: contourplots offmite element analysis under gravity load) and
modulatíon oflight conditions (top nght: geographicaüy specific üluminance anaÍysis onthe provide both an overarching and intricate conception of the fonnation and functioning
system and a register surface). © Achim Menges.
natural systems. What this combination projects, to archltecture, is that computational

18 Computational Design Thinking 19 Introduction:ComputationatDesignThinking

designcanbedeveloped asa set of instructions regarding the process offormation, with
algorithms exist, but evolutionary searches are particularly relevant as the algorithms
the trigger being the interaction with internai and externarforces.
computationally evolve solutions without explicitly specifying the direct evolution
..Thl.worLOL^-Fraze^desc/lbed.'^^ aPtly tit^ book /In Evolutionary
process, which becomes an emergent property of the algorithm.23
^'r/"tec'ure'M-drta^the,transfer.°f the b'ol°g'°l Princip^ ofmorphogenesisiand
evo?mtoarch!tecture'His intention was that an «'^hitecturebasedon^he^bL^tness;
A large component of evolvability must be attríbuted to ínherent structural
lar!a^Tdromplex;ty°f natural systems must be thou9ht °fasbeingformed'a'b7tractTy
properties of features that originated (...) for one reason, but also manifest a
thro.ugh.the.same methods-what is also Pr°P°sed is that the producie3dlarchiterturlei'sla
capacity for subsequent recrustment (wíth mimmaí change) to substantiaiiy
port'?-o"t.ofthAnatura'system'.exhiblting'metabolism'and^
different and nove! functions, (Stephen J Gouid)24
towh'ch_'twasformed:.in exchanç)e with env'r°"ment, responsive'tolfeedback"and
m its own right.
The open-ended characteristic of evolutionary computation is of particular interest for
-,wha\mo.rpl°9eres's '"s°.portr;iys are the 'ntimate and irred"°ble relationships
architecture, as a design task can usually not be comprehensively described as a problem,
between formation and nnaterialisation. In any organic system the7e~islnolk'dïstinctio^
but rather as an opportunity for creating novel possibilities within the framework of a
fo.[ Tt3nceLbetween material and structure: They are in^parable -" alwauysLSGrLJnq
given brief. Here, evolution is more about finding than searching, about the continuai
the dual purpose/ at the most fundamental levei, of generating mass andbem7alselfy
extension of the search spgce through novel solutions that emerge as the by-product of
supporting system.As material is of a particular arrangement of matter and in semce-of
the evolutionary dynamics of selection, mutation and inheritance. In this way, evolutionary
^pecjfíc capacitie^ it is devoid of an a priori deftned way how that arrangement should be
algorithms, which were traditionally utilised for optimisation, also enable performative
formed. Manuel DeLanda cresses this in defíning Deleuze-s definÍtJo^muÍtÍplÏÏti^
morphological processes, adaptiogenesis and inherent novelty to be investigated In
as distinctly opposite to forms being defined by t-heir essence: -Unlike essencels"whïc'h
bounds greatly superseding hand-driven experimentation.25
assume that matter is apassive receptacle for externai forms, multjplic,tíes-areimmanent
EvolutionaryprocessËsaresignifícantastheyalwaysoperateonamultitudeorpopu/otíon
to^material processe^defíning their spontaneous capacity to generate pattemwithout
intervention/21 ofsolutionsratherthanjustone.Whatthisaddstotheknowledgeofdesigncomputation/and
simultaneously to architecture, is the complete displacement of the concept of type within a
\f there is no essence, or predicated driver for the arrangement of matter, then it
design process. Contrary to other approaches to design computation, as for example Frame
^clearthaMheproperties ofmatter and capacities to organise m response'to'internaÏ
Based Design26 or Case Based Design,27 the notion of historical precedents as registered in
externai infíuences are what drive formation. What DeLanda refers to'as"/extemaÍ
a formal language is not an active participant in this kind of computational generation. It is
m^ention^implies a directed, or top-down, development rather than responsiv^or
of criticai conceptual importance to understand that in this way evolutionary computation
n-up, formation. In computatíon, the product becomes distingujshedfrom^the
provides for truly explorative processes, rather than reinforcing existing knowledge (fused
processi. B_ecausethere is no Pre-existing knowledge of a specifíc formation, computation
into the description of a type) that is simply varied or recombined. This is also regístered in
can only capture the means and constraints to be Ínflected upon a specifíc5e7ofrmatuerial
the understanding of population within an evolutionary context.
properties. Computational processes contain everything about the material svstem'"-
In population thinking, type is a complete abstraction, something which only can
except forthe system itself. ThÍs is often referred to"as ;state space/.
be overlaid, not embedded. This works on the notion that acquired characteristics
Evolving Populations are not embedded in the genotype. They are repercussions from the interactions of
development and environment. As evolution Is a continuai process, the notion of type
as a static or suddenly shifting defínition also does not work. The forces of variation
Nothing in bioiogy makes sense except in the Hght of evolution. (Theodosius
Dobzhansky)23 and selection drive the formation and constant reformation of populations, not a pre-
defíned essential characterisation.28
In computation,the response is in the use ofvariational methods based on extensibility
Evolutionary computation hás been employed in various disciplines, and is often
rather than type. This is not to say that ali computational methods are intended to be
d"mbe<Las belng.a" °bout search''" comPuter scien^ aromputational pro"ble"m
universal. It is that natural systems function through the differentiation that deveiops
caribe defíned in terms of a search space. A search space is-a'n'otÍona"l"fiel7of'^l
in their morphogenetic processes. The differentiation occurs via the genotype, and the
existmg^parameters within a defíned problem. The iterative sampling of par^eters
reactions between the phenotype ancf the environment. Frazer'5 work, Darwin's description
generatespopü/otíonsofpotentJalsolutions.Amongthese particular solutions/a search
|,of natural seiection and evolution, and Mayr's detailing of popuiation thinking and
hm navigates this space to fínd the most suitable one. Many different search
|evolutionary development depict how such interrelations of form and formation occur.

20 Computationat Design Thinking

21 Introduction: Computationat Design Thinking
Achim Menges, Morphogenetic Desigrt SeanAMquist, Moritz Reischraann,
Experiment 02 - AA Strawberry Evoluüonary Spatial Subdivision, AA
Bar, 2003. Based on the dynamics of Emergent Technologies and Design
selecüon, mutation and reproduction, Programme (Michael Hensel, Achim
an evolutionary algorifhm generates Menges, MichaelWeinstock), London,
populations of individual morphologies 2008. Based upon a simple principie of
across mulüple generations. This volumetric subdivision, this algorithm
evolutionary process is driven by the evolves complex relationships withm
evaluation of fitness criteria related to a prescribed envelope between
the constructional aharacteristics ofthe distributions of programmatic allocations,
pneumatic structure, which is intended to spatial voids and overalt articulated
form a strawberry bar. © Achim Menges. massing. Fitness is determined through
fhe generation ofmultiple geometric
populations via conditions of externai
and internai spatial articulation arid
programmatic adjacencies, © Sean
Ahlquist/Moritz Fleischmann.

22 Compuíationat Design Thinking 23 Introduction:ComputationatDesignThinking

Emergent Formatíons

Emergence /s generally understood to be a process that leads to the appearance of

structure not directly described by the defining constrainïs and ínstantaneous forces
that controï a system. Over time 'something new' appears at scaies not directiy
specífíed by the equations ofmotion. An emergent feature also cannot be explicitiy
represented in the initía! ana boundary conditions. In short, a feature emerges when
the underlying system puts some effort ínto its creatíon. CJames P Crutchfíeld)29

What the summation of these concepts regarding computational desígn presents is a

generative approach to the formation of complex material arrangements. Underlying each
individual concept is the demand for a set of procedures to operate in a generally non-
hierarchícal and undirected fashion. Algoríthms serve to set the range of possibilities,
while analytical measures apply to sei the levei of 'fíf-ness of spedfic instances within
the set of possibilities. What this defines is a system by which emergence characterises the
resulting behaviours of forms generated from these computational means. Without the
presence or imposition of an envisioned yessence/ as mentioned previously, characteristics
which define a system can only emerge through the process of integration and formation.
It must be understood, though, that emergence is not only a property of pattem-
formation or physical self-organisation. Emergence is glso a factor in behaviour and
function.30This unfolds a particularly unique opportunity for the relation of designer,
design and computation. Novelty arises not only in formal arrangement (a typical
desíre in conventional computerised approaches), but Ít is more fruitfully applied in
the reali5atíon of multiple behavioural capacities - ones not initíally Ímagined duríng
the genesis of the process.
The most compelling aspect of systems which exhibit emergent behaviour is that
such behaviours are derivatives of simple condítions, often called yagents/.31 This depends
upon, again, the separation of properties and actions. An agent hotds a simpte set of
properties. The environment defines a set of mies in which the agents Ínteract. At neither
ofthe leveis does an indícation ofthe emergent behaviour exist.

To fina an emergem behavíour is to find the Ínteraction ofunit behaviours which

contributos to the global fitness. (J Poon and ML Maher)32

In computatíonal design, it is rather idealistic to imagine that a functional and highly

effective behaviour can emerge from a fundamental nondeterministic process. That
a design process evolves ímplies that the criteria for effectiveness shift as well. In
this effort, emergence is a property to not only be observed but to be deciphered, SeanAMquist, ICD Stuttgart University (AchimMenges), Deep Surfaces, Stuttgart, 2010.^
Based upon Hooke's Law of Elasticity, form is generated as an emergent cQHdriion tnrougnme^
traced and ultimately tabulated by means which can determine its relative capacíty.
determinaüon of tension-equilibnum in the system. The emergent benaviours Demg uduyu^^:L
While not traceable explicitly, the concept that a computational process relies upon beyond simple formal capacity, but also to include sucïi thinking as particular pattem formaton^n
self-shading and shadow casting, This method employs a Particle System and advanced me&ods
feedback provides the avenue by which emergent conditions can be honed in a non-
in topologicalmapping programmed m the Java-based environment Processing. © Sean ?lquist.
directed set of procedures.

i5 Introduction: Computatíonat Design Thinking

24 Computationat Design Thinking
Designing Computation

Before a gíven sdentífíc discipline can begin to gaín from the use of virtual
environments, more than Just casting oid assumptions ínto mathematical form Eïp^rimen-GI. Kw^vwr O;'
Eïcerlment 02 Kwimlb'" Cfftt »

is necessary. In many cases the assumptions themselves need to be modífíed.

(Manuel DeLanda)33

The approach for computational design is one which focuses on the execution of
variational methods for the purposefui Íntent of resolving the complexities that exist in
the interrelation and interdependendes of material stmctures and dynamic environments.
Design processes consist of two states: how they function/ and what they consist of. As
mentioned in relation to morphogenetic processes, one state describes the parameters
of materiality, while the other describes the infíuences which will activate those material
constraints to arrange in a particular manner. Non-existent is a state of computation
which explicitly ancf singularly describes the resultant condition. This occurs only in the
execution ofthe process as a whole.
Computation as defining processes for comparative, informational means hás the
potential to function as a universal application, but the mechanism works only in the
processing of specific, not symbolic, conditions relating to materiality, spatiality and
context. While the procedures define a vast state space of potentials, the result embodies
specífíc descriptions of an overall system. Computational processes are iterative and
recursive, but most importantly, expanding. They work by growing and specifying the
information which describes form through procedures which recursively generate form,
calling variable parameters within the state space. The variational condition is íntended
to address not only the complexity of emergent behaviour m active systems, but the
relation between the material and the adaptive/ customisable and controllable capabilíties
of current and future processes of materialisation.

In developing computer programs one is forced to question howpeople think ana

how design evotves. In other words, computers must be acknowledged not only
c?5 machines for imitatíng what is understood, but aíso as vehicles for exploring
what Ís not understood. (Kostas Terzídís)34

The explorative nature of desígn computation is not limited but enabled by the inherent
finiteness of algorithmic procedures. In his essay on 'Algorithm and Creativity', Peter
Weibel positsthat:
HorianKrampe, Christopher Voss, IntegratedUrbanMorphologies, ICD StuttgatíUmversity
The determinism of algoríthms leads to an indeterminism. In this way the (Prof. Achim Menges, Sean Ahlquist), Stuttgatí, 2011, The project implements an Evolutionary
Algorithm to negotiate constramts for spatial and cü-culatory organization with potentíals for
character of creativity is an open horizon, even though it is generated through climatic modulaüon in the deplo^rment of an urban housmg block. In this non-typologicaüy
a fínite number of rules. Creatívity means algorithmic design, whích needs to be based generative process, the methods evolve mutti-performmg morphologies as weU
as utilize scripted analyses to idenüfy novel emergent spatial formations. © Institute for
conceíved in a way that even the unpredíctable, that is occurrences for beyond the -
Computational Design, Stuttgart University,
subjective horizon of design, can be designed for. Creativity entaíls algorithmic

26 Computational Design Thinkmg 27 Introduction: Computatíonat Design Thinking

4 Paul Coates, Programming ArcMectm-e, Rouüedge (Abingdon), 2010, pp 1-4.
design plus the complementarity of designing additíona! system components. 5 ArtoSalomaa, Computatíon an.d Auíomata, Cambridge University Press (CamJ3ridge,UK), 1985, p 131.

C..) The creatíve figure wít! be both the designer ofalgorithms and the interpreter 6 Friedrich L Bauer, Historiscíie Noíizen zur InformatÏk, Epringer (Berlin), 2009, p 162.
7 RobHoward, Compufmgïn Consfrucüojs, ReedElsevier (Oxford), 1998, pp 19-38.
of their outcomes. (PeterWeibel)35
8 GordonPask, 'TheArchitecturalRelevance ofCybemetics', mAr-cAïtecfaíra/Desiga, 1969, pp 494-6,
9 Nick Chrisman, Charting the Unknown: How Computer Mapping at Harvsrd Became GIS, ESR[
Press (Redlands, CA), 2006, pp 69-79,
In designing the algorithms, Steven Coons addresses more dlrectly the criteria for the
10 Ivan Sutherland, 'Sketchpad: AMan-Machine Graphical Communication System', Proceedings of
algorithms themselves. Where the aim is simply to 'furnish information', Coons sees the ïhe AFIPS Spríngjomt Compuíer Conference, Detroit, 1963, pp 21-3,
expiidt difference between a computerised and computational approach: 11 Earl Mark, Mark Gross and Gabriela Goldschmidt, 'A Perspective on Computer Aided Design after
Four Decades', AMIT, 2008, pp 1-8.
12 Murray Mune, 'Whatever Became ofDesign Methodology?', in Computer Aids io Desigs and
There are two quite different phíiosophies of approach to the achievement of Arcliitecture, by Nicholas Negroponte, pp 30-6, Mason/Charter Publishers (London), 1975, p 31
these aíms. One approach would be to imbed in the computer a large set of 13 Christopher Alexander, Notes on the Synthesís ofForm, Harrard University Press (Cambridge,

spedal purpose packaged processes, each designed to perform some speciat tosk. MA), 1964, p 130.
14 Ludwig von Bertalanïïy, General System Theory: Foundaïions, Development, Applications, George
If the assembly of such a library of specía! routines couíd be made complete Braziller(NewYork), 1969, p 141,
enough, then the system would exhibit to the user on the outside an appearance 15 Norbert Wiener, Cybemetics: or Control ancï Communication in ïhe Animal and tïie MacSiine, MIT
Press (Cambridge, MA), 1965, pp 1-29.
of complete fiexibitity ana generality. This wouid be satisfactory só íong as
16 VladimirBazjanac, ;The Promises and Disappointments ofComputer-AidedDesign', m
the designer never called for a capability not already rígidly imbedded in the Negroponte, Compuïer Aids toDesign and Ãichitecture, pp 17-26.
mechamsm. But the design process is unpredictabte. Indeed part of the design 17 Kwinter, FarFrom Equüïbnum, p 149,
18 D'ArcyThompson, On Growth andForni, abridged edition, Cambridge University Press
process consists Ín dessgníng new ways to perform the design functíon Ítself. This
(Cambridge), 1961, p 270.
/s o higher orderofdesign activity, a sort of meta-design (like meta-mathematics) 19 Robert Woodbury, Elements ofParametncDesign^.o-uÜ.edge (Abingdon), 2010, p 24,

that clearíy is outside the scope ofany rigid sei- ofspecial processes that can be 20 ]oïm Prazer, An EvolutionaryArcïütectnre, AA Publications (London), 1995, pp 9-21.
21 Manuel DeLanda, Iníensive Science and Virtual Pivlosopliy, Continuum (NewYork), 2002, p 26.
anticipated at the begínnjng. (Steven Coons)36
22 Theodosius Dobzhansky, Amerjcan 5jbJogy Teacüer 55, 1972,pp 125-9.
23 PeterJBentleyandDavidWCome. Creaïíve EvolutionarYSysíenis.AcadermcPress (SanDiego),

In completing the mode of thinklng for computational design, the position of design- 2001, pp 1-78,
24 Stephenjay Gould, Tïie Structure ofEvolutionary Theory, Harvard University Press (Cambridge,
er must be addressed. Computation does not function with the explicit description of
MA), 2002, p 801.
fonn or finality. Therefore, the designer is posited as the author of the rules as implicit 25 M Hemberg, A Menges andU 0'Reüly, 'Evolutionary ComputationmArchitecture', in Archítectural

descriptions for the development of form. This is not a trivial matter, nor is it directly an Design, vol 74, no 3, Wiley-Academy (London), 2004, pp 48-53,
26 Marvin Minsky, Society ofMind, MIT Press (Cambridge, MA), 1986, pp 260-72.
abstract conceptual one. It is the detennination of conaete rules regarding material, and 27 Ann Heylighen and Herman Neuckermans, 'A Case Base of Case-Based Design Tools for
criteria for the engagement of architecture with its contextual systems. Frazer and Bentley Architecture',m Computer-Aided Desígn, vol33, no 14, Elsevier (Oxford), 2001, pp 1111-22.

describe the value in computational means for invention, though more importantly for 28 EmstMayr, What Evolution Is, BasicBoofe (NewYork), 2001, pp 83-7.
29 JamesP Crutchfield, 'The Calculi ofEmergence; Computation, Dynamics. and Induction', inthe
dedudng fitness and effectiveness in complexsolutions.Terzidis describes, more universally, Physica D (1994) special issue on the Proceedings ofthe Oji Intemational Seminar Complex
a general approach of authorship which eschews precedents and forms information about Sysíems— from Complex Dynamics to Artificial Realityh.eï.d 5-Q Aprïl 1993, Numazujapan, p 2.
30 JohnSGero (1996), 'Creativity, Emergence andEvolutionmDesign: Concepts andFramework',
functioning systems through the very criteria and logics in which it will exist. Design is not
Knowledge-Based Systems 9(7), pp 435-48.
only in the selection of criteria, but it is in the fashioning of the arguments themselves, 31 John Holland, Emergence: From Chãos to Order, Oxford University Press (Oxford), 2000, pp
in both a technical computational manner and a theoretical, material and architectural 115-241.
32 J Poon and ML Maher, ' Co-Evolution and Emergence in Design', in Artificial InteUigence in
vein. It is therefore importgnt that a criticai approach towards computational design exacts
Engmeering, vol 11, issue 3July 1997, pp 319-27, ElsevierScience, p 323.
knowledge from the historicai and practical foundations of these Ínterrelated fíelds. This 33 Manuel DeLanda, 'Virtual Environments and the Emergence of Synthetic Reason', in Joan
knowledge must also reflect a reciprocal maturation of both intellectual discourse and Broadhurst Dixon andEncJ Cassidy (eds), Virtual Futures, Routledge CLondon), 1998, pp 65-76.
34 Terzidis, Algoríthmic Architecture, p 59.
technícal advancement, fulfilling g vital positioning onhinking in computational design.
35 Peter Weibel, 'Algorithmus und Kreativitât', in Berka Walter, Emil Brüí and Christian Smekal Ceds),
WoberkommtdasNeue?Kreãtivitatm Wissenschafï und Kunsí, BõMauVeriag (Vienna), 2003, p 96,
Notes
36 StevenCoons, 'An Outline of the Requirements for a Computer-Aid.ed Design System', mAFIPS
l S^ordKvmter,FarFromEquilíbnw-n:EssaysonTechnologyandDesígnCultw-e,Actaï:
Joint Computer Conference Proceedíngs, 299-304, ACM (NewYork), 1963, p 301.
(Barcelona), 2008, p 147.
2 Kostas Terzidis, Algontïmiic Arcilitecture, Architectural Press (Oxford), 2006, p xi,
3 Kostas Terzidis, Expressive Form, SponPress (London; NewYork), 2003, pp 65-73,
© 2011 John Wiley & Sons Ltd.

29 Introduction: Computationat Design Thinking

28 Computational Deslgn Thinking
Formatíon and Transformation However, when we study forms, the organic ones in particular, nowhere do we tind
permanence, repose or termination. We find rather that everything is in ceaseless fíux.
]ohann Wolfgang von Goethe This is why our language makes such frequent use of the term 'Bildung' to designate what
hás been brought forth and likewise what is in the process of being brought forth.
In this brief text by Goethe, a translation from the original text /ldeen über In introducing a science of morphology, we must avoid speaking in temns of what is
organische Bildung' (1806) as part of his groundbreaking collection of fixed. Thus, if we use the term 'Gestalt/ at ali, we ought to have in mind only an abstract
'SchriftenzurlVlorphologie^thedefínitionofmorphologyisfírstestablished. idea or concept, or something which in actuality is held fast for but an instant.
Morphology, as Goethe describes, depicts the elicited 'character' ofsystems, What hás just been fomned is Ínstantly transformed/ and if we would arrive/ to some
of both an animate and inanimate nature, by the inextricable link of form degree, at a vital intuition of Nature, we must strive to keep ourselves as fíexible and
and formation. That a system is in constant transformation. Goethe breaks pliable as the example she herself provides.
from the Aristotelian view that the elements of a dynamic organism can
Notes
be isolated to dissect their functioning within the character of the whole. l Goethe's method ofthinkmgïiasbeendescribed as contemplative cognition, 'amode ofcognition
The characteristics of natural behaviour - conceptualised as interaction, which is at once sensory (grasping the phenomenon) and spiritual (perceiving the spirit which
maniíèsts itselfinthe phenomenon)', KarlVietor, Goethe íhe Tíimker, Harvard University Press
iteration, generation and variation - to which Goethe alludes. and the
(Cambridge,MA), 1950, p 12.
manner by which they are deciphered is fundamental to the exacting of 2 Goethe'sdiscussion in thefoüowingparagraphs revolves aboutthederivaüonofthenouns, Gestalt
computation in design. Important is not only the set of instructions, but and Biláimg. Gestaltïs derived from the Middle HighGermanpastparticiple oïstellen: to set,
place, put; Bildung is derived from the verb bílden, aüied in meanmg to, though not a translation of,
also the meía-instructions which describe the transformational activities
the English verb to buíld, The distinction between the two German nouns may be brought out by
giving rise to an overall functioning system. translating Gestaltas form, Bíldimgas formaïioji.

From Johann Wolfgang vou Goethe, translated by Bertïia Mueller,

Tormation and Transformation', Goethe 's Botanical Writings, Ox Bow Press
When our interest in natural objects/ especially the organic ones, is awakened to the
(Woodbridge, CT), 1989, pp 21-4. Reproduced by permission of University
extent that we desire to obtain an insight into relationships between character and
of Hawaii Press. © University of Hawaii Press. For this edition the notes
function, we believe ourselves best able to acquire such knowledge through analysis
have been renumbered.
of the parts. This method is indeed likely to take us far - it requires but a word or two
to remind friends of science what chemistry and anatomy have contributed toward an
intensive and extensive view of Nature.
But these analytícal efforts, if continued indefinitely, have their disadvantages.
To be sure, the living thing is separated into its elements, but one cannot put these
elementstogetheragain and givethem life.This istrueeven of inorganic bodies, to say
nothing of organic ones.
For this reason, the man of science hás always evincect a tendency to recognise living
forms as such/ to understand their outwardly visible and tangible parts in relation to one
another, to lay hold of them as indicia of the inner parts, and thus, in contemplation,1 to
acquire a degree of mastery over the whole. How closely this scientific aspiration is bound
up with the creative and imitative urges need not be dealt with in detail.
Hence several attempts are found in the progress of art, learning and sdence to
establish and develop a theory to which we should like to give the name 'morpholog/. The
vgried forms these attempts assume, will be spoken of in the historical portion of ourwork.
The German language hás the word 'Gestalt'2 to designate the complex of life
in an actual organism. in this expresslon the element of mutability is left out of
consideration: it is assumed that whatever forms a composite whole is made fast, is
cut off, and is fixed in its character.

30 Computationat Design Thinking 31 Formation and Transformation

On the Theory of Transformatíons, or the means of it in touch with Galileors aphorism (as oid as Plato, as old as Pythagoras,
as old perhaps as the wisdom of the Egyptians), that /the Book of Nature is written
Comparison of Related Forms in characters of Geometr/.1
We are apt to think of mathematical definitions as too strict and rigid for common
D'Arcy Weníworth Thompson use, but their rigour is combíned with ali but endless freedom. The precise definition
D/Arcy Wentworth Thompson's seminal book On Crowth ana Form (1917) of an ellipse introduces us to ali the ellipses in the world; the definition of a 'conic
lays the foundation for understanding the associated repercussions of section/ enlarges our concept/ and a 'curve of higher arder' ali the more extends our

environmental pressures and geometric formation. In this selected text, range of freedom.2 By means of these large limitations, by this controlled and regulated
Thompson establishes mathematics as a route to which these repercussions freedom/ we reach through mathematical analysis to mathematica! synthesís. We discover
can be mapped. Not only is it described as a means for dedphering an homologies or identíties whích were not obvious before, and which our descriptions
individual course of development or growth, but Thompson also outlays obscured rather than revealed: as for instance, when we leam that, however we hofd our
the interpolation between multiple morphometric mappings as a means chain, or however we fíre our bullet, the contour of the one or the path of the other is
to project potentials in form. This sets two fundamental branches of a always mathematically homologous.
conceptual framework for computational geometry: parametrics and Once more, and this is the greatest gain of ali, we pass quickly and easily from the
homologies. A parametric equation Es defíned as a constant equation in mathematical concept of form in its statical aspect to form in its dynamical relations:
whichrelationalparametersvary.ltresultsinprodudng f3/n/7/esofproducts we rise from the conception of form to an understanding of the forces which gave rise
where each instance will always carry a particular commonness with others. to it; and in the representation of form and in the comparison of kindred forms, we see
Thompson defines these embedded relationships as homologies. The in the one case a diagram of forces in equilibrium, and in the other case we discem the
framework that emerges encapsulates, within formal rules for geometric magnitude and the ctirection of the forces which hgve sufficed to convert the one form
organisation, the capacities of form, in physical stature and robustness, and into the other Here/ since a change of material form is only effected by the movement of
theirtransformation through externai influences. matter,3 we have once again the support of the schoolman's and the philosopher/s axiom,
ignorato motu, ignoratur natura.

In the foregoing chapters of this book we have attempted to study the inten-elations of There is yet another way - we learn from Henri Poincaré - to regarei the function of
growth and form, and the part which the physical forces play in this complex interaction; mathematics, and to realise why its laws and its methods are boundto underlie ali parts
and, as part of the same enquiry, we have tried in comparatively simple cases to use of physical science. Every natural phenomenon, however simple/ is really composite, and
mathematical methods and mathematical terminology to describe and define the forms of every visible action and effect is a summation of countless subordinate actions. Here
organisms. We have learned in só doing that our own study of organíc form, which we call mathematics shows her peculiar power, to combine and to generalise. The concept of an
by Goethe/s name of Morphology, is but a portion of that wider Science of Fomn which average, the equation to a curve, the descriptíon of a froth or cellular tissue, ali come
deals with the forms assumed by matter under ali aspects and conditions, and, in a still withín the scope of mathematics for no other reason than that they are summatíons
wider sense, wíth forms which are theoretically imaginable. of more elementary principies or phenomena. Growth and Form are throughout of this
The study of form may be descriptive merely, or it may become analytical. We composite nature; therefore the laws of mathematics are bound to underlie them, and her
begin by describing the shape of an object in the simple words of common speech: methods to be peculiarly fitted to interpret them.
we end by definlng it in the precise language of mathematics; and the one method In the morphology of living thlngs the use of mathematical methods and symbols
tends to follow the other in strict scientific order and historical continuity. Thus, hás made slow progress; and there are various reasons for this failure to employ a
for Ínstance/ the form of the earth, of a raindrop or a rainbow, the shape of the method whose advantages are só obvious in the investigation of other physical forms.
hanging chaln/ or the path of a stone thrown up into the air, may ali be described, To begin with, there would seem to be a psychological reason, lying in the fact that

however inadequately, in common words; but when we have learned to comprehend the student of living things is by nature and training an observer of concrete objects
and to define the sphere, the catenary, or the parábola, we have made a wonderful and phenomena and the habit of mínd which s/he possesses and cultivates is alien to
and perhaps a manifold advance. The mathematícai definition of a /formr hás a |that of the theoretical mathematician. But this is by no means the only reason; for in

quality of precision which was quite lacking in our earlíer stage of mere description; kindred subject of mineralogy/ for instance, crystals were still treated in the days

ít is expressed in few words or in still briefer symbofs, and these words or symbols Linnaeus as wholly within the province of the naturalist, and were described by him

are só pregnant with meaning that thought itself is economisecf; we are brought by ?er the simple methods in use for animais and plants: but as soon as Haüy showed the

33 On the Theory of Transformations, orthe Comparison of Related Forms

32 Computational Design Thinking
application of mathematics to the description and classification of crystals, his methods forget that the biologist is much more exacting in his requirements, as regarás form, than
were immediately adopted and a new science carne into being. the physidst; for the latter is usually content with either an ideal or a general description
A large pari of the neglect and suspicion of mathematical methods in organic of form, whiie the student of living things must needs be specific. Material things, be
morphoiogy is due ... to an ingrained and deep-seated belief that even when we they livíng or dead, show us but a shadow of mathematical perfection.6The physicist or
seem to discern a regular mathematicgl figure in an organism, the sphere, the mathematidan can gíve us perfectlysatisfying expressionsfortheform ofawave, oreven
hexagon, or the spiral which we só recognise merely resembles, but is never entirely of a heap of sand; but we never ask him to define the form of any particular wave of the
explained by/ Íts mathematical analogue; in short/thatthedetails in whichthe figure sea, nor the actual form of any mountain-peak or hill. [...]
dlffers from its nnathematical prototype are more important and more interestíng [...] For one reason or another there are very many organic forms which we cannot
than the features in which it agrees; and even that the peculiar aesthetic pleasure descr[be,stilllessdefine,inmathematica[terms:justasthereareproblemseveninphysical
wlth which we regará a living thing is somehow bound up wíth the departure from science beyond the mathematics of our age. We never even seek for a formula to define
mathematical regularity which it manifests as a peculiar attribute of life. This view this fish or that, or this or that vertebrate skull. But we may already use mathematical
seems to me to involve a misapprehension. There is no such essential difference language to describe, even to define in general terms, the shape of a snail-shell, the
between these phenomena of organíc form and those which are manifested in twist of a horn, the outline of a leaf, the texture of a boné, the fabric of a skeleton, the
portions of inanimate matter.4 The mathematician knows better than we do the stream-lines of físh or bird, the fairy lace-work of an insecfs wing. Even to do this we
value of an gpproximate result.5 The child's skipping-rope is but an approximation must leam from the mathematician to eliminate and to discará; to keep the type in mínd
to Huygens' catenary curve - but in the catenary curve lies the whole gist of the and leave the single case, with ali Íts accidents, alone; and to find in this sacrifíce of what
matter. We may be dismayed too easily by contingencies which are nothing short matters little and conservation of what matters much one of the peculiar excellences of
of irrelevant compared with the main issue; there is a principie of negligibiisty. the method of mathematics/
Someone hás said that if Tycho Brghé's Ínstmments had been 10 times as exact In a very large part of morphology, our essential task lies in the comparison of related
there would have been no Kepler, no Newton, and no astronomy. fomns rather than in the precise defínition of each; and the deformatíon of a complicated
If no chain hangs in a perfect catenary and no raindrop is a perfect sphere, this figure may be a phenomenon easy of comprehension, though the figure ítself have to
is for the reason that forces and resistances other than the main one are inevitably at be left unanalysed and undefíned. This process of comparison, of recognising in one
work. The same is true of organic form, but it is for the mathematician to unravel ihe form a defínite permutation or deformation of another, apart altogether from a precise
conflicting forces which are at work together. And this process of investigation may and adequate understanding ofthe original /type/ or standard of comparison, lies within
lead us on step by step to new phenomena, as Ít hás done in physics, where sometimes the Ímmediate province of mathematics/ and finds its solution in the elementary use of
a knowledge of form leads us to the Ínterpretation of forces, and at other times a a certain method ofthe mathematidan. This method is the Method of Coordinates, on
knowledge of the forces at work guides us towards a better insight into form. After which is based theTheory ofTransformations.3
the fundamental advance had been made which taught us that the world was round, l imagine that when Descartes conceived the method of coordinates, as a
Newton showed that the forces at work upon it must lead to its being imperfectly generalisation from the proportional diagrams of the artist and the architect, and long
spherical, and in the course of time its oblate spheroidal shape was actually verified. before the immense possibilities of thís analysis could be foreseen, he had in mind a very
But now, in turn, it hás been shown that its form is still more complicated, and the simple purpose; it was perhaps no more than to fínd a way of translating the form of a
next step is to seek for the forces that have deformed the oblate spheroid. As Newton curve (as well as the position of a poinü into numbers and into words. This is precisely
somewhere says, 'Nature delights in transformations". what we do, by the method of coordinates, every time we study a statistical curve; and
The organic forms which we can define more or less precisely in mathematical terms, conversely, we translate numbers into form wheneverwe /plot a curve', to illustrate a table
and afterwards proceed to explain and to account for in terms of force, are of many kinds, of mortality, a rateofgrowth, orthe dailyvariation oftemperature or barometric pressure.
as we have seen; but nevertheiess they are few in number compared with Nature's ali In precisely the same way it is possible to inscribe in a net of rectangular coordinates the
but infinite variety. The reason for this is not far to seek. The living organism represents, outline, for ínstance, of a fish, and só to translate it into a table of numbers, from which
or occupies, a fíeld of force which is never simple, and which as a mie is of immense again we may at pleasure reconstruct the curve.
complexity. And just as in the very simplest of actual cases wemeetwith a departurefrom But ít is the next step in the employment of coordinates which is of special interest
such symmetry as could only exist under conditions of ideal simplicity, só do we pass and use to the morphologlst; and this step consists in the alteratíon, or deformation, of
quickly to cases where the interference of numerous, though still perhaps very simple, our system of coordinates, and in the study of the corresponding transformation of the
causes Íeads to a resultant complexity far beyond our powers of analysis. Nor must we curve or figure Ínscribed in the coordinate network.

34 Computatíonal Design Thmking 35 On the Theory of Transformations, or the Comparison of Related Forms
í S ? 4 5- í; 7 7 a 3 * £ 6 7 S <?

^ y<
^
'4
&

c
'i
à'

efc ^

fê &t^
~Á-
k^
 ^

í y. 3 y 6 r s 9 x
TV 5^^6/99

PelvisofArchaeopteryx.D'ArcyThompson-fr'om original The coordinate systems ofthe first and second

publication. Reproduced by permission of Cambridge figures with three intermediate systems
University Press. © Cambridge University Press, interpolated. D'Arcy Thompson - from original
publlcation, Reproduced by permíssion of
Cambridge University Press. © Cambridge
UniversityPress,

8 f

Pelvis ofApatomis, D'Arcy Thompson - from The first intermediate coordinate network,
original publication. Reproduced by permission with its corresponding inscribed pelvis,
of Cambridge University Press. © Cambridge P'Arcy Thompson - from original pubücation.
UniversityPress. Reproduced by permission of Cambridge
Umversity Press, © Cambridge Umversity Press,

36 Computaíional Design Thinking 32 °ntheTheoryofTransformations,ortheComparisonofRelaíedForms

Let us inscribe in a system of Cartesian coordinates the outline of an organism, extraordinarily dífficult transformation; but, physically speaking, its achíevement may
however complicated, or a part thereof: such as a físh, a crab, or a mammalian skull. be extremely simpíe. The little round gourd grows naturally, by its symmetrical forces
We may now treat this complicated figure, in general terms, as a function ofx, y. If we of expansive growth, into a big, round, or somewhat oval pumpkin or melon.10 But
submit our rectangularsystem to deformation on simple and recognised lines/ altering, for the Moorish husbandman ties a rag round its middle, and the same forces of growth,
instance,thedirectionoftheaxes,theratioofx/y, orsubstituting for x and y some more unaltered save for the presence of this trammel, now expand the globular structure
complicated expressions, then we obtain a new system of coordinates, whose deformation into two superposed and connected globes. And again, by varying the posítion of
from the original type the inscribed figure will precisely follow. In other words, we obtain the encircling band, or by applying several such ligatures instead of one, a great
a new figure which represents the old figure under a more or less homogeneous stram, variety of artificial forms of /gourcT may be, and actually are, produced. it is cfear, l
and is a function of the new coordinates in precisely the same way as the old figure was think, that we may account for many ordinary biological processes of development or
ofthe original coordinatesxandy. transformation ofform bythe existence oftrammels or Unes ofconstraint, which limít
The problem is closely akin to that of the cartographer who transfers identical data ancf determine the action of the expansive forces of growth that would otherwíse be
to one projection or another;9 and whose object is to secure Cif Ít be possible) a complete uniform and symmetrical. This case hás a dose parallel in the operations ofthe glass-
correspondence, in each small unitofarea, between the one representation and the other. blower, to which we have already, more than once, referred in passing.11 The glass-
The morphologist will not seekto draw his organicforms in a new and artificial projection; blower starts his operations with a tube, which he first doses at one end só as to form
but, in the converse aspect of the problem, he will enquire whether two different but a hollow vesicle, within whích his blast of air exercises a uniform pressure on ali sides;
more or less obviously related forms can be só analysed and interpreted that each may but the spherical conformation which this uniform expansive force would naturally
be shown to be a transformed representation of the other. This once demonstrated, it wíll tend to produce is modifíed into ali kinds of forms by the trammels or resistances
be a comparatively easy task (in ali probability) to postulate the direction and magnitude sei up as the workman lets one part or another of his bubble be unequally heated or
of the force capable of effecting the required transformation. Again, if such a simule cooled. It was Oliver Wendell Holmes who fírst showed this curious paralle! between
alteration of the system of forces can be proved adequate to meet the case, we may fínd the operations of the glass-blower and those of Nature, when she starts, as she só
ourselves able to dispense with many widely cun-ent and more complicated hypotheses of often does, with a simple tube.12 The alimentary canal, the arterial system íncludmg
biological causation. For it is a maxim in physics that an effect ought not to be ascribed the heart, the central nervous system ofthe vertebrate, including the brain itself, ali
to thejoint operation of many causes if few are adequate to the production of it: frustra begin as simple tubular structures. And with them Nature doesjust what the glass-
fít per plura, quod fieri potest per pauciora. blower does/ and/ we might even say, no more than he. For she can expand the tube
* * * here and narrow it there; thicken its walls or thin them; blow off a lateral offshoot or
Before we pass from this brief discussion of transformations in general, let us glance at caecal diverticulum; bendthetube, ortwistand coil it; and infofd orcrimp Ítswallsas,
one or two cases in which the forces applied are more or less intelligible, but the resulting só to speak, she pleases. Such a form as that of the human stomach is easily explained
transformations are, from the mathematical point ofview, exceedingly complicated. when it is regarded from this point of view; it is simply an Íll-blown bubble, a bubble
The'marbled papers' ofthe bookbinderare a beautiful Íllustration ofvisible'stream- that hás been rendered lopsided by a trammel or restraint along one side, such as to
lines'. On a dishful of a sort of semi-liquid gum the workman dusts a few simple lines or prevent its symmetnca! expansion - such a trammel as is produced Ífthe glass-blower
patches of colouring matter; and then, by passing a comb through the liquid/ he draws lets one side of his bubble get cold, gnd such as is actually present in the stomach
the colour-bands into the streaks, waves and spiralswhich constitute the marbled pattem; itself in the form of a muscular bancf.
and which he then transfers to sheets of paper laid down upon the gum. By some such The Florence flask, or any other handiwork of the glass-blower, is always beautiful,
system ofshears, bythe effect of unequal traction or unequal growth in various directions beca use its graded contours are, as in its líving analogues, a picture ofthe graded forces by
and superposed on an originally simple pattern/ we may account for the not dissimilar which it was conformed. It is an example of mathematical beauty/ of which the machine-
marbled patterns which we recognise/ for instance, on a large serpenfs skin. But Ít must made, moulded bottle hás no trace at ali. An alabaster bottle is different again. It is no
be remarked, in the case ofthe marbled paper, that though the method of application of T an unduloid figure of equilibrium. Turned on a lathe, it is a solid of revolutíon/ and
the forces is simple, yet in the aggregate the system of forces set up by the many teeth not without beauty; but it is not nearly só beautiful as the blown fíask or bubble.
of the comb is exceedingly complex, and its complexity is revealed in the complicated The gravitational field is part of the complex fíeld of force by which the form of the
"diagram of forces" which constitutes the pattem. °rganísm is Ínfíuenced and determined. Its sharë is sçldom easy to define, but there is a
Totakeanotherandstifl more instructive illustration.Toturn one circle (orsphere) ^resultant due to gravity in hanging breasts and tired eyelids and ali the sagging wrinkles
intotwocircles(orspheres)would be,fromthe pointofviewofthe mathematician, an old. Now and then we see gravity at work in the normal construction of the body,

39 On the Theory ofTransformations, or the Comparison of Related Forms

38 Computationat Design Thnking
11 Where gourds are common, the glass-blower is still apt to take them for a prototype, as the
and can describe its effect on form in a general, or quaiitative, way. Each pair of ríbs in
prehistoric potíer also did, For instance, a taU, annulated Rorence oil-flask is an exact but no longer
man forms a hoop which droops of Íts own weight in front, só flattening the chest, and a conscious inútation of a gourd which hás been converted into a bottle in the manner described.
at the same time, twisting the rib on either hand near its poínt of suspension.13 But in 12 CíElsíe Venner, chap II.
13 See TF Anderson Stuart, 'How the Form of the Thorax is Partly Determined by Gravitation' , Proc,
the dog each costal hoop is dragged straight downwards, into a vertical instead of a JÏ,S.XLIX,pl43, 1891.
transverso ellipse, and is even narrowed to a point atthe sternal border.
From D'Arcy Wentworth Thompson, 'On the Theory ofTransformations, or
the ComparisonofRelatedForms', On Growth and Form: The Complete
Notes
Revised Edition, Dover Publications (NewYork), 1992, pp 1026-51.©
l CfPlutarch, Symp, viii, 2, onthemeanmgofPlatols aphorism (!ifit acíuaüywasPlato's'): TTÜS'
Cambridge University Press, reproduced wíth permission. For this edition
TJÁárcúï/ cAeyc ròv Qçòv ácí ye^erpsÏv,
2 Só said Gustav Theodor Fechner, the author ofFechner's Law, a hundred years ago. (Uber die notes have been renumbered.
mathematische Behandlung organischer Gestalten und Processe, Benchte d, k. sachs. Gesellsch.,
Maïh.-phys, Cl., Leipzig, 1849, pp 50-64.) Fechnerls treatment is more purely mathematical
and less physical in its scope and bearing than ours, and his paper is but a short one, but tïie
conclusions to which he is led differ little from our own. Let me quote a smgle sentence which,
together with its context, nms precisely on the lines which we have foüowed in this book: 'Só ist
also die mathematische Bestimmbarkeit im Gebiete dês Organischen ganz eben só gut vorhanden
ais in dem dês Unorganischen, und in letzterem eben solchen oder âquivalenten Beschrânkiaigen
untervrorfen ais iïi ersterem; und nur sofem die unorganischen Foimen und das unorganische
Geschehen sich einer einfacheren Gesetzlichkeit mehr nâhem ais die organischen, kann die
Approximaíion im unorganischen Gebiet leichter und weiter getrieben werden ais im organischen.
Dies wâre der ganze, sonach rein relative, Unterschied.' Here, in a nutshell, is fhe gist ofthe wbple
matter.
3 'We can move matter, that is ali we can do to it' (Oliver Lodge) .
4 Henri Bergson repudiates, wifh peculiar confídence, the application of mathematics to biology; cf
Creative Evoluïion, p 21, 'Calculationtouches, atmost, certain phenomena of orgamc desü'uction.
Organic creation, on the contrary, the evolutionary phenomena. whicü properiy constitute life, we
cannotmanywaysubjectto amathematicaltreatment.' Bergson thus foüows Bichat; 'C'estpeu
connaïtre lês fonctions animales que de vouloir lês soumettre au moindre calcul, parce que ïeur
mstabilité est extreme. Lês phénomènes restent toujours lês mêmes, et c'est ce qui nous importe;
maisleursvariations, enplus ou enmoins, sont sans nombre' (La VíeetlaMort, p 257).
5 Wïien we make a 'first approximation' to the solution of a physical problem, we usuaüy mean
that we are solving one part while neglecting others. Geometry deals with pure forms (such as a
straight line), defined by a single law; but these are few compared with the míxed forms, like the
surface of a polyhedron, or a segment of a sphere, or any ordinary mechanical construction or arty
ordinary physical phenomenon. It is only in a purely mathematical treatment of physics that the
'single law' can be dealt with alone, and the approximate solution dispensed witti accordingly.
6 Cf Haton de la GoupiUière, op cit: 'On a souvent 1'occasion de saisir dans la nature un reflet dês
formes rigoureuses qu'étudie la géometrie,'
7 CfWHYoung, 'TheMathematicalMetïiod andits Limitations', Congresso dei Mâtematicí, Bologna,
1928.
8 ThemathematicalTheoryofTransformationsispartoftüeTheoryofGroups, of great importance
m modem mathematícs. A disttnction is drawn between Súbstitution-groups and Transformation-
groups, the former being discontinuous, the latter continuous - in such a way that within one and
the same group each transformation is mfímtely little different from another. The distinction among
biologists between a mutation and a variation is curiously analogous.
9 Cf(eg) Tissot, Mémoire sur la représeníaíion dês sui 'faces, et lês projections descartes
géograpíiiques, Gauthier-ViUars (Paris), 1881,
10 Analogousstructuraldifferences, especiaüyinthe fibrovascularbundles, helpto explainthe
differences between (eg) a smooth melon and a cantaloupe, or between various elongate, flattened
and globular varieties. These breed true to type, and obey, when crossed, the laws ofMendelian
inheritance. CfEWSínnett, 'Inheritance ofFruit-shapein Cucurbita', Bofan. Gazette, LXXTV, pp
95-103, 1922, andotherpapers.

40 ComputationaL Design Thinking the TheoryofTransformations, orthe Comparison of Related Forms

Variatíonal Evolutíon The assumptions of population thinking are diametricaliy opposed to those of
the typologist. The populationist stresses the uniqueness of everything in the
Ernst Mayr organic world. What is true for the humgn species - that no two individuais are
alike - is equally true for ali other species of animais and plants. Indeed, even
Ernst Mayr hás made a significant contribution to definition ofthe 'modern the same individual changes continuously throughout its lifetime and when
synthesis', which combines the theories of heredity/ evolution and natural piaced into different environments. Ali organisms and organíc phenomena are
selection. In this excerpt, he defines speciation and emphasises the composed of unique features and can be described collectively only in statistical
fundamental necessity ofdistinguishing populations from types or classes. terms. Individuais, or any kind of organíc entities, form populations of which we
Extracted from his book What Evolution Is (2001), this text describes the can determine the arithmetic mean and the statistics of variation. Averages are
roots of the typologist as being bound to an essentialist point of view, merely statistical abstractions, only the individuais of which the populations are
stating that al] phenomena are qualities which are absolute. Evolution composed have reaiíty. The ultimate condusions of the population thinker and
follows a path from heredity to genotype to environment to variation to of the typologist are precisely the opposite. For the typologist, the type Çeidos)
phenotype. With the imposition of environment, the possibility of wilfully is real and the variation an illusion, while for the populationist the type (average)
unbending traits, held within the genotype, is shown as less feasible. is an abstraction and only the variation is real. No two ways of looking at nature
With the ultimate goal being fítness within a particular environment, the could be more different. ÇMayr 1959)1
identifícation of traits can only occur with the phenotype; a population
is being defíned by common characteristics at the phenotypic levei. Darwin/sVariational Evolution
Mayr's concept of populatíon thinking separates the direct link between ItwasDarwinwhointroducedthis newwayofthinking into sdence. His basicinsíghtwas
genotype and phenotype. One handles the pool of information while that the living world consists not of invaríable essences (Platonian classes), but of highly
the other deafs with organising that information in response to natural variable populations. And it is the change of populations of organisms that is designated
selection and environment. This offers a distinct understanding of the as evolution. Thus, evolution is the tumover of the individuais of every population from
basis of formation - as a repercussion of contextual interaction rather generatlon to generation.
than a direct lineage of initial intent. When Darwin, in 1837, became an evolutionist, he asked himself, how can the process
of evolution be explained? Could he adopt one of the already proposed explanations?
[...] He realised eventually that neíther transmutationism nor transformationism nor any other
VARIATION AND POPULATION THINKING theory based on essentialism would do. And he was right. Ali essentialistic theories of
Darwin showed that one simply could not understand evolution as long as one accepted organic evoiution are badly flawed, as was convincingly established during the post-
essentíalism. Spedes and populations are not types, they are not essentialisticaliy defíned Darwinian controversies.
classes, but rather are biopopulations composed of genetícally uníque individuais. This Darwin had to develop an entirely new kind of explanation that accounted for
revolutionary insight required an equally revolutionary explanatory theory of evolution: the abundance of variation in nature. This led him to his theory of natural selection,
Darwin'5 theory of variation and selection. Two sources of evidence led Darwin to this which was based on population thinking. The same theory was found independently
new concept. One was the empirical study of variable natural populations (particularly byAlfred Russel Wallace.
during his study of the bamacles), and the other was the observatíon by animal and Although Darwin pubiished On the Origín of Spedes in 1859 (actually Wallace and
plant breeders that no two individuais oftheir herds or breeding stocks were identical. Darwin published a fírststatement in 1858), the explanatorytheory ofvariational evolution
These individuais were not members of essentialistic classes, and, as we now know, ali was not universally adopted until c 80 years later. It is g theory based on the variability
individuais in a sexual populafion are geneticglly unique. of populations. There were two sets of practitioners who had already appreciated this
Apparently, most people fínd it diffícult to grasp the significance of this uníqueness. variability for a long time, the taxonomists and the animal and plant breeders, and Darwin
Let them rememberthat no two individuais among the six billíon humans are idêntica!, had dose connections to both of them.
not even so-called Ídentical (monozygous) twins. An understanding of the fundamental When sorting out the collections he had made cm the voyage of the Beagle,
difference between a class of essentially identical objects and a biopopulation of unique Darwin encountered the same question again and again: are some slightly different
individuais is the foundation ofso-called "popuiation thinking', one ofthe most important speclmens merely variants within a population or are they different species? Indeed/
concepts of modern biology. in the 1840s when he wrote his monographs on the classifícation of the bamacles,

42 Computational Design Thinking 43 VanatíonaLEvotutíon

Darwin carne to the condusion that no two specimens in a collection from a single
population were exactly identical. They ali were as uniquely different from each other
as are human individuais. And the animal and plant breeders/ wíth whom Darwin was
associated since his Cambridge student days, told him the same. They always knew
which individuais in their herds they should select as the breeding stock for the next
generation. Indíviduaíity made this possible.
Box 1 Darwin/s Five Major Theories of Evolution
SÍnce the terms 'transmutationism' and 'transformationism' are not suitable
for this new theory, Darwin/s theory of evolution through natural selection is best
1. The nonconstancy of species Cthe basic theory of evolution).
referred to as the theory of vanational evoíutíon. According to this theory, an
2. The descent of ali organisms from common ancestors (branching evolution).
enormous amount of genetic variation is produced in every generation, but only a few
3. The gradualness of evolution (no saltations, no discontinuities).
individualsofthevastnumberofoffspringwill surviveto producethe nextgeneration.
The theory postulates that those individuais with the highest probability of surviving 4. The multiplication of species (the origin of diversity).
5. Natural selection
and reproducing successfully are the ones best adapted, owing to their possessíon of
a particular combination of attributes. Since these attributes are largely determined
by genes, the genotypes of these individuais will be favoured during the process of
selection. As a consequence of the continuous survival of individuais Cphenotypes)
with genotypes best able to cope with the changes of the environment, there will be
a continuing change in the genetic composition of every population. This unequal
survival of individuais is due in part to competition among the new recombinant
genotypes within the population, and in part to chance processes affecting the
Box 2 Rejection of Some of Darwirfs Theories by Early Evolutionists
frequency of genes. The resulting change of a population is called evolution. SÍnce
The following table shows the composition of the evolutionary theories of various
ali changes take place in populations of genetically unique individuais, evolution is
evolutionists. AU of these authors accepted a fífth theory, that of evolution as
by necessity a gradual and continuous process.
opposed to a constant, unchanging world. They dfffered in acceptlng or rejecting
some of Darwin's four other evolutionary theories.
Darwin/sTheories of Evolution
Darwin'5 views on evolution are often referred to as The Darwinian Theory. Actually
Common Populationa! Natural
they consist of a number of different theories that are best understood when clearly
distinguished from each other. The most important of Darwir/s theories of evolution
Descent Graduai Selection
are discussed opposite Csee Box 1). That they are indeed fíve independent theories is
documented by the fact that the leading 'Darwinians' among DarwÍn/s contemporaries
Lama rck No Yes No No
Darwin Yes Yes Yes Yes
accepted some and rejected others Csee Box 2).
Haeckel Yes Yes ? In pari
Two of these fíve theories, evolution as such and the theory of common descent,
Neo-Lamarckigns Yes Yes Yes No
were widely accepted by biologists within a few years of the publication of the Origin.
TH Huxley Yes No No No
This represented the first Darwmian revolution. The acceptance of man as a primate in the
de Vries Yes No No No
animal kingdom was a particularly revolutionary step. Three other theories, graduaiism,
TH Morgan Yes No No Unimportant
speciation and natural selection, were strongly resisted and were not generally accepted
until the evolutionary synthesis. This was the second Darwmían rei/o/utíon.The Darwinism
proposed byAugustWeismann and Wallace, in which an Ínheritance of acquired characters
is rejected, was named Neodarwmism by George John Romanes. The Darwinism accepted
since the evolutionary synthesís is best simply called Darwínísm, because in most crudai
aspects it agrees with the origina] Darwinism of 1859, while the belief in an inheritance of
acquired characters is by nowtotally obsolete.

VariationaLEvoLution
44 Computationat Design Thinking
Darwin's theory of gradualism fítted well into the thinking of the transformationists, of polymorphism in a population over long periods. A balance of selection pressures
but the resistance of the saltationists wgs sufficiently great that the universal acceptance of is usually assumed, but it may be reinforced by some superiority of the heterozygous
the graduainess of evolution had to await the evolutionary synthesis. Darwin/s concept of carriersthatfavoursthe retention ofthe rarergene in the population. In a highly diverse
gradualness,however,wasofanentirelydifferentnaturefromthatofthetransformationists. envíronment, phenotypic diversity may be selected, as in the case ofthe banded snails.
Their graduainess was due to the gradual change of an essential type, whereas Darwinian
gradualism is due to the gradual restructuring of populations. This makes ít quite clear THESOURCEOFVARIABfLITY
why Darwinian evolution, being a populational phenomenon, must always be gradual. A What is the source of thís vgriability? Where does it come from? How Es Ít maíntained from
Darwinian must be able to show that every seeming case of saltation or discontinuous generation to generation? This is what puzzled Darwin ali of his life, but in spite of ali his
evolution can be explained as being caused by a gradual restructuring of populations. effortsheneverfoundtheanswer.Anunderstandingofthenatureofthisvariabilitywasfinally
made possible, after 1900, by advancements in genetics and molecular biology. One can never
VARIATION fuliy understand the process of evolution unless one hás an understanding of the basic facts
The availability of variatíon is the indispensable prerequisite of evolution/ and the study of inheritance, whích explain variation. Therefore the study of genetics is an integral part of
of the nature of variation is therefore a most important part of the study of evolution. the study of evolution. But only the heritable part of variation plays a role in evolution.
Variation, the uniqueness of every individual, is, as we said, characteristic of every sexually
reproducing spedes. To be sure, at first sight al! the individuais of a species of snail or Genotype and Phenotype
butterfly or físh might seem identicgl, but a closer study of these individuais will reveal As early as the 1880s Ít was recognised by perceptive biologists that the genetic material
ali sorts of differences in size, proportions, colour pattem, scaling, bristles, and whatever (germ plasm) was something different from the body of an organism (soma), and this
characteristic one studies. Further studies have shown that variability affects not onty distinction was satisfíed when the early Mendelians introduced the terms genotype
visible characters, but also physiological traits, patterns of behaviour, aspects of ecology and phenotype. But the prevailing opinion at that time was that the genetic material
Ceg, adaptation to dimatic conditions) and molecular patterns, ali ofthis reinforcing the consisted of proteins like those that make up the body. It carne as a real shock when
conclusion that in one way or another every individual is unique. And it is this always Avery demonstrated in 1944 that the genetíc material consisted of nucleic acids. The
avaiiable varíability that makes the process of natural seíection possible. terminological distinctfon between an organism and its genes now acquired a new
Although the variability of the phenotype was appreciated by naturalists as far back meaning. The genetic material ítself is the genome (haploid) or the genotype (diploid),
as Darwin's day, the early geneticists treated the genotype as rather uniform. When the which contrais the production of the body of an organism and ali of íts attributes, the
studies by the population geneticists from the 1920s to the 1960s revealed the presence phenotype. This phenotype is the result of the interaction of the genotype with the
of a great deal of cryptic variation, this was questioned by some of the classical authors. environment during development. The amplitude of variation of the phenotype produced
Yet not even the most enthusiastic Darwinians suspected the amount of genetic variation byagivengenotypeunderdifferentenvironmentalconditionsiscalied \tsnormofreaction.
in populations that was eventually revealed by the methods of molecular genetics. Not only For instance, a given plant may grow to be larger and more luxuriant under favourable
was Ít discovered that much of the DNA consists of noncoding DNA Cjunk'), but it was conditions of fertilising and wateríng than wíthout these environmental factors. Leaves
also found that many, perhaps the majority of alleles are 'neutra!', that is, their mutation of the Water Buttercup ÇRanunculus fiabeilaris') produced under water are feathery and
does not affect the fítness of the phenotype Csee below). As a result, it is now realised that very different from the broadened leaves on the branches above water. As we shall see,
seeminglyidentical phenotypes mayconceal considerablevariation at the levei ofthegene. it is the phenotype that is exposed to natural selection/ and not individual genes directly.
ithasbeen heatedlyargued inthe pastwhethera particular propertyof an organism
Polymorphism wasdue to'nature/(Íts genes) or'nurture'(itsenvironment). Ali research ínthelastlOO
Sometimesvariation falis Ínto defínite classes/ a phenomenon referred to aspoiymorphism. years indicates that most characteristics of an organism are affected by both factors.
In the human spedes we have polymorphisms for eye colour, hair colour, straightness or This is partícularly true for characters that are controlfed by multiple genes. There are
curliness of the hair, different blood groups and many other of the genetic variants of two sources of variation in a sexually reproducing population, superimposed on each
our species. The study of polymorphisms hás greatiy contributed to our understanding other: the variation of the genotype Cbecause in a sexual species no two individuais are
ofthestrength and direction of natural selection, aswell as the causal factors underlying genetically idêntica!) and the variation ofthe phenotype (because each genotype hás
variation. Two outstanding studies are those on the colour polymorphism of banded its own norm of reaction). Different norms of reaction may react rather differently to
snails (Cepoea) by Cain and Sheppard and on chromosome arrangements in Drosophsla the same environmenta! conditions.
* * *
by Dobzhansky. In most cases Ít is unknown what is responsible for the maintenance

46 CompLftational Design Thnking 42 VariationatEvotution

THE IMATURE OFVARIATION d grthropods (crustaceans). Selectíon seems to be able to recruit genes
In Darwin/s day, the nature of variation in populations was not yet understood. Thís cpo c ^l.plnnmental processes that previousiy had seemed to have otherfuncíions.
understanding was possible only following developments in the late 19th and 20th innÏi^Tbeen"shown by morphological-phylogenetic research that photoreceptor
centuries. What Darwín did know as a naturalist, taxonomist and student of natural _L fl^ ihaddeveloped at least 40 times independently during the evolution of
populations was that variation in natural populations seemed to be vírtually ínexhaustible. organs ^y^tv'."A developmental genetidst, however, showed that ali animais with eyes
!t.-pro_Y'dcs abundant material for natural selection in ali organisms, at least in sexually anir"ü.',_u'7^rpnijlatorv aene, Pax 6, which organises the construction of the eye. It
reprodudng specíes of animais and plants. The visible characterístícs of an organism'', "at fírst concluded that ali eyes were derived from a single ancestral eye
its phenotype, are due to instructíon during development by their genes and by the was L^tP^Ygene. But then the geneticist also found Pax 6 in species without eyes
genotype ínteracting with the environment. rósecKhat they must have descended from ancestors with eyes. However, thls
rioTurned out to be quite improbable and the wide distribution of Pax 6 required a
The Impact ofthe ÍVlolecuIar Revolution explanation. It is now belleved that Pax 6, even before the origin of eyes had
Although the basic principies of ínheritance were worked out between 1900 and the 'unknown functíon in eyeless organisms, and was subsequently recruited for its role as
1930s the real understancfing of the nature of inheritance was achieved only through
an eye organiser.
the motecutar revolutlon. It began in 1944 (Avery et al) when it was established that the
genetic material consísted not of proteins but of nudeic acids. In 1953 Watson and Crick
CONCLUSIONS ^ ^ ^ _^ ^ ^_ .^._^.^
discovered the structure of DNA, and after this one major discovery followed the other, 1'Tisshown in this chapter that Darwin, by making biopopulatíons the
cutminating in the discovery ofthe genetic code by Nirenberg in 1961 ... Finally, every evolutíonary theorising, rather than Platonic types, had found an entirely nove! solution
step in thetranslatíon of the genetic mformation in the course ofthe developing organism for the explanation of evolutíon. He postulated that the inexhaustibie genetic varíation
was understood in principie. Unexpectedly, the basic DarwinÍan concepts ofvanation and of a population, together with selection Celimination), is the key to evolutionary change.
setectionwere not affected in any way. Not even the replacement of proteins by nucleic To understand howthís is implemented one must understand inheritance ... The genetic
acids as the carriers of genetíc information required a change in the evolutionary theory. material is constant and does not permít an inheritance of acquíred characteristics. The
Onthecontrary, an understanding ofthe nature ofgeneticvariation greatlystrengthened genotype, interacting with theenvironment, producesthe phenotype during development
Darwinism, for it confírmed the fínding of the genetidsts that an ínheritance of acquired Mutation$ continually replenish thevariability ofthe gene pool. However, the variatlon of
characters is impossibte. the phenotypes that provide the material for selectíon is produced by recombination in
Molecular biolog/s greatest contributíon to evolutionary bíology was the creation of meiosis, a process of restructuring and reassorting the chromosomes.
the fíetd of developmental genetícs. Developmental biology, which^had só long reslsted
Note _ „
the ewlutionary synthesis, now adopted Darwmian thinkíng and analysed the functional
l E Msyr, Typological versus Population Thmkmg', in Evohition anci Antïu-opology. A
role of the genotype. This led to the discovery of regulatory genes (hox^ pax, etc-) and Appraisal. pp 409-12, Anthropological Sociely ofWashmgton (Washmgton), 1959.

thus vastly enlarged our understanding of the evolutlonary aspects of development,

From Ernst Mayr, 'Variational Evolutíon', What Evolution Is, Basic Books
(NewYork) - Perseus Books Croup, 2001, pp 83-144. Copyright © 2002
Evolutionary Developmental Biology
Ernst Mayr. Reprinted by permission of Basic Books, a membeï of the
One of the most ímportant discoveries of molecular genetics was that some genes are Perseus Books Group. The note hás been added for this edition.
very old. This means that the same gene (essentially the same sequence of base pairs) is
found in organisms that are only very distantfy related, say m Drosophsla and mammals.
A second discovery was that certain genes, often referred to as regulatory genes, contrai
such basic developmental processes as the determination of anterior vs posterior or of
dorsal vs ventral, These fíndings shed consíderable light not only on previously completely
puzzling developmentaí processes, but also on the causation of fundamental events
(branching poínts^ in phylogeny.
Sáenfists had always assumed that the same gene, no matter where found, always
had the same phenotypic effect. But developmental geneticists have now shown thatthís
is not necessarily só. The same gene may have rather different expressions in annelíds

48 CompuíationatDesignThinking 49 VariationalEvolution
The Meaning of General System Theory system and of the elements involved. These considerations lead to the postulate of
a new scientifíc discipline whích we call general system theory. Its subject matter is
Ludwig von BertaLanffy formulation of principies that are valid for 'systems' in general, whatever the nature of
their component elements and the relations or 'forces' between them.
In this text, published as a part of the book General System Theory: General system theory, therefore, is a general sclence of/wholeness'which uptill now
Foundations, Development, Applícations (1969), which was based upon was considered a vague, hazy and semi-metaphysical concept. In elaborate form it would
the essay "An Outline of General System Theory/ (1950), Bertalanffy be a logico-mathematical discipline/ in itself purely formal but appiicable to the various
provides a universal description of the means and outcomes by which empirical sdences. For sciences concerned with 'organised wholes', it would be of similar
dynamic systems operate. Citing a confíuence of studies across many fíelds signifícance to that which probability theory hás for sciences concerned with 'chance
of science, Bertalanffy specifically discusses two criticai conditions that events'; the latter, too, is a formal mathematical discipline which can be applied to most
apply to ali scenarios: equifínality and feedback. Equifinality states that diversefields, such asthermodynamics, biological and medicai experimentation, genetics,
a systenfs functioning state can be achieved from various initial starting life ínsurance statistics, etc.
concfitions. Feedback, whereby the system is in constant transformation,
poses a mechanism by which information is reinvested into the system to This intíicates major aims of general system theory:
provide a constant rebalancing and recalibration of its functioning state.
As Bertalanffy explains, both rely upon the determination of a system to 1 There is a general tendencytowards Íntegration in the various sdences, natural
achieve a steady, or "homeostatic", state. Of particular relevante to design and social.
is the understanding that the result of a functioning system is a "state" 2 Such íntegration seems to be centred in a general theory of systems.
and that given inputs/ outputs and feedback, this is one of many states in 3 Such theory may be an important means for aiming at exact theory in the
constant fíuctuation. nonphysical fields of science.
4 DevelopÍng unifying principies running /vertically/ through the universe of the
individualsciences/thístheorybringsusnearertothegoaloftheunityofscience.
Aims of General System Theory 5 This can lead to a much-needed integration in scientifíc education.
We may summarise these considerations as follows. Similar general conceptions and
víewpoints have evolved in various disciplines of modern science. While in the past, science A remark as to the delímitation of the theory here discussed seems to be appropríate.
triedto explaín observable phenomena by redudng them to an interplayofelementary units The term and programme of a general system theory was introduced by the present
investigatable independently of each other, conceptions appear in contemporary science author a number of years ago. It hás turned out, however, that quite a large number of
that are concemed with what is somewhat vaguely termed "wholeness", ie, problems of workers in various fíelds had been led to similar conclusions and ways of approach. It is
organisation, phenomena not resolvable into local events/ dynamic interactions manifest suggested/ therefore, to maintain this name which is now comíng into general use, be
in the dífference of behaviour of parts when isolated or in a higher confíguration, it only as a convenient label.
etc; in short, 'systems' of various orders not understandable by investigation of their It looks, at first, as Íf the definition of systems as /sets of elements standing in
respective parts in isolation. Conceptions and problems ofthis nature have appeared in ali inten-elation' is só general and vague that not much can be learned from it. This, however,
branches of science, Írrespective of whether Ínanimate things, living organisms or social is not true. For example, systems can be defíned by certain families of differential
phenomena are the object of study. This correspondence is the more striking because the equations and if, in the usual way of mathematical reasoning, more specifíed conditions
developments in the individual sciences were mutually Índependent, largely unaware of are íntroduced, many important propertíes can be found of systems in general and more
each other, and baseei upon different facts and contradicting philosophies. They indicate special cases.

a general change in scientifíc attitude and conceptions. The mathematical approach followed in general system theory is notthe only possible
Not only are general aspects and viewpoints alike in different sciences; frequentiy or most general one.There are a number of related modern approaches, such as Ínformation
we find formally Ídentical or isomorphic laws in different fields. In many cases, theory, cybemetics/ game, decision and nettheories, stochastic models, operations research,
isomorphic laws hold for certain ciasses or subclasses of 'systems', irrespective of tomentiononlythemostimportantones. However/thefactthatdifferentialequationscover
the nature of the entities Ínvolved. There appear to exist general system laws which ^extensjve tields in the physical biological, economical, and probably also the behavioural
apply to any system of a certain type, in-espective of the particular properties of the ^saences, makes them a suitable access to the study of generalised systems.

SI The Meaning of General System Theory

50 Computationat Design Thinking
Information and Entropy
Another developmentwhich is dosely connected with system theory is that ofthe modern
theory of communication. it hás often been sald that energy is the currency of physics,
justaseconomicvaluescan be expressed in dollarsorpounds.Thereare, however, certain
fields of physics and technology where this currency is not readily acceptable. This is the
case in the field of communication which, due to the development of telephones, radio,
radar/ calculating machines, servomechanisms and other devices, hás led to the rise of a
new branch of physics.
The general notion in communication theory is that of information. In many cases,
the flow of information corresponds to a fíow of energy, eg, if light waves emitted by
some objects reach the eye or a photoelectric cell, elicit some reaction ofthe organism
or some machinery, and thus convey information. However, examples can easily be given
where the fíow of information is opposite to the fíow of energy, or where infon-nation is
transmitted without a flow of energy or matter. The first is the case in a telegraph cable,
where a direct current is fíowing in eme direction, but information, a message, can be sent
in either direction by interrupting the current at one poínt and recording the interruption
at another. For the second case, think of the photoelectric door openers as they are STIMÜLU; ^ESSAôl ^ÊSSAOS ^SPONSC
installed in many supermarkets: the shadow, the cutting off of light energy/ informs the

^
photocell that somebody is entering, and the door opens. Só information, in general,
cormot
cannot be expressed in terms of energy. BECEPTOfí "''"""' ^
emcroR ~1>
APPÂRÂTü$
There is, however/ anotherwayto measure information, namely, interms ofdecisions.
Take the game of Twenty Questions, where we are supposed to find out an object by
receiving simple 'yes/ or /nor answers to our questions. The amount of information
conveyed in one answer is a dedsion between two alternatives/ such as animal or
F££08ACK
nonanimal. With two questions, it is possible to decide for one out of four possibilities,
eg, mammal - nonmammal, or fíowering plant - nonflowering plant. With three answers,
it is a dedsion out of eight/ etc. Thus, the logarithm at the base 2 ofthe possíble dedsions
can be used as a measure of information/ the unit being the so-called binary unit or bit.
The information contained in two answers is log 4=2 bits, of three answers, log^, 8=3
bits, etc. This measure of information happens to be similar to that of entropy or rather
negative entropy, since entropy a!so is defined as a logarithm of probability. But entropy,
as we have already heard, is a measure of disorder; hence negative entropy or information
is g measure of order or of organisation since the latter, compared to distribution at
random, is an improbable state.
Figure l. Simple feedbackscheme,
A second central concept of the theory of communication and control is that of
Ludwig von Bertalanffy - reprinted fi-om
feedback. A simple scheme forfeecfback is the following CFigure 1 ). The system comprises, ongmal publicaüon © George Braziller,
first, a receptor or 'sense organ', be it a photoelectric cell, a radar screen, a thermometer,
or a sense organ in the biological meaning. The message may be, in technological devices,
a weak current, or, in a living organism, represented by nerve conductíon, etc. Then there
is a centre recombining the incoming messages and transmltting them to an effector,
consisting of a machine like an electromotor, a heating coil or solenoid, or of a muscle

52 Computatíonat Design Thinking 53 The Meaning of General System Theory

which responds to the Íncoming message in such a way that there is power output of high diminish and eventually abolish Íts equipotentiality. Thus, dynamics is the broader
energy. Finally/thefunctioningoftheeffectoris monitored backtothe receptor, and this aspect, since we can always arrive from general system laws to machinelike function
makes the system self-regulating, ie, guarantees stabitisation or direction of action. by introducmg suitable conditions of constraint, but the opposite is not possibfe.
Feedback arrangements are widely used in modern technology for the stabilisation
of a certain action, as in thermostats or in radio receivers; or for the direction of actions Causality and Teleology
towards a goal where the aberration from that goal is fed back, as infonnation/ till the goal Another point l would like to mention is the change the scientifíc world picture hás
or target is reached. This is the case in self-propelled missiles which seek their target, anti- undergone in the past few decades. in the world view called mechanistic, which was
aircraftfire control systems, ship-steering systems, and otherso-called servomechanisms. bom of clássica! physics of the 19th century, the aimfess play of the atoms, governed by
There is indeed a large number of biological phenomena which correspond the inexorable laws of causality, produced ali phenomena in the world, inanimate, living
to the feedback model. First, there is the phenomenon of so-called homeostasis, and mental. No room was left for any directíveness, arder or telos. The world of the
or maintenance of balance in the living organism, the prototype of which is organisms appeared a product of chance, accumulated by the senseless play of random
thermoregulation in warm-blooded animais. Cooling of the blood stimulates certain mutatíons and selection; the mental world as a curious and rather inconsequential
centres in the brain which 'turn on/ heat-producing mechanisms of the body, and the epiphenomenon of material events.
body temperature is monitored back to the centre só that temperature is maintaíned The only goal of science appeared to be analytical, ie, the splitting up of reality into
at a constant levei. Similar homeostatic mechanisms exist in the body for maintaining ever smaller units and the isolation of individual causal trains. Thus, physical reality was
the constancy of a great number of physicochemical variables. Furthermore, feedback split up into mass points or atoms, the living organism into cells, behaviour into refíexes,
systems comparable to the servomechanisms of technology exist in the anima! and perception into punctual sensations, etc. Correspondingly, causality was essentially one-
human body for the regulation of actions. If we want to pick up a penei!, a report is way: one sun attracts one planet in IMewtonian mechanics, one gene in the fertilised
made to the central nervous system ofthe distance by which we have failed to grasp ovum procfuces such and such Ínherited character, one sort of bacterium produces this
the penal in thefírst instance; this information isthenfed back to the central nervous or that disease, mental elements are lined up, like the beads in a string of pearls, by
system só thatthe motion is controlled till the aim is reached. the law of association. Remember Kanfs famous table ofthe categories which attempts
Só a great variety of systems in technology and in living nature follow the feedback to systematise the fundamental notions of dassical science: it is symptomatic that the
scheme, and it is well known that a new discipline, called cybernetics, was introduced by notions of Ínteraction and of organisation were only space-fillers or did not appear at ali.
Norbert Wiener to deal with these phenomena. The theory tries to show that mechanisms We may state as characteristic of modern science that this scheme of isolable units
of a feedback nature are the base of teleological or purposeful behaviour in man-made acting in one-way causaiity hás proved to be insufficient. Hence the appearance/ in
machines as well as in living organisms, and in social systems. ali fíelds of science/ of notions like wholeness, holistic, organismic, gestalt, etc/ which
It should be borne in mind, however, that the feedback scheme is of a rather ali signify that, in the last resort, we must think in terms of systems of elements in
speciai nature. It presupposes structural arrangements of the type mentioned. There mutual interaction.
are, however, many regulations in the living organism which are of an essentially SÍmilarly, notions of teleology and directiveness appeared to be outside the scope of
difíerent nature, namely, those where the order is effectuated by a dynamic interplay science andto be theplaygroundofmysterious^supematuraloranthropomorphic agencies;
of processes. Recall, eg, embryonic regulations where the whole is re-established or else, a pseudoproblem, intrinsically alien to science, and merely a misplaced projection
from the parts in equifínal processes. It can be shown that the primary regulations ofthe observeis mind Ínto a nature governed by purposeless iaws. Nevertheless, these
in organic systems, íe, those which are most fundamental and primitive in embryonic aspects exist, and you cannot conceive of a living organism/ not to speak of behaviour
development as well as in evolution, are of the nature of dynamic Ínteraction. They and human society, without taking into account what variously and rather loosely is called
are based upon the fact that the iiving organism is an open system, maintaining adaptiveness, purposiveness, goal-seeking and the like.
itself in, or approaching a steady state. Superposed are those reguiations which It is characteristic of the present view that these aspects are taken seriously as
we may call secondary, and which are controiled by fixed arrangements, especially a legitimate problem for science; moreover, we can well indicate models simulating
of the feedback type. This state of affairs is a consequence of a general principie such behaviour.
of organisation which may be called progressive mechanisation. At first, systems - Twosuch modelswe have already mentioned. One is equifinality,thetendencytowards
biological, neurological, psychoiogical or social-are governed bydynamicinteraction ^a characteristic final state from different initial states-and in different ways, based upon
of their components; later cm, fixed arrangements and conditions of constraint are ^dynamic interaction in an open system attaining a steady state; the seconcf, feedback,
established which render the system and its parts more effident, but also gradually homeostatic maintenance of a characteristic state or the seeking of a goai, based

54 Computational Design Thmkmg S5 The Meaning of General System Theor^

upon circular causal chains and mechanisms monitoring back infomnation on deviations resort, ali is vanilla, ali mind or spirit. The unifying principie is that we find organisation
from the state to be maintained or the goal to be reached. A third model for adaptive at ali leveis. The mechanistic world view, taking the play of physical partides as ultimate
behaviour, a "design for a brain', was developed by Ross Ashby, who Íncidentally started reality, found its expresslon in a civilisation which glorifies physical technology that hás
with the same mathematical definitions and equations for a general system as were used led eventualiy to the catastrophes of our time. Possibiy the model of the world as a great
by the present author. Both writers have developed theír systems independently and, organisation can help to reinforce the sense of reverence for the living which we have
following different lines of interest, have arrived at different theorems and condusions. almost iost in the last sanguinary decades of human history.
Ashby'5 model for adaptiveness is, roughly, that of step functions detining a system,
ie, functions which, after a certain criticai value is passed, jump Ínto a new family of
From Ludwig von Bertalanffy, General System Theory: Foundatíons,
differential equations. This means that, having passed a criticai state, the system starts
Development, Applícatíons, George Braziller (NewYork), 1969. © George
off in a new way of behaviour. Thus, by means of step functions, the system shows
Brazíller.
adaptive behaviour by what the biologist would call trial and error: it tries different ways
and means, and eventually settles down in a fíeld where it no longer comes Ínto conflict
with critica! values of the environment. Such g system adapting itself by trial and error
was actually constructed by Ashby as an electromagnetic machine, called the homeostat.
i am not going to discuss the merits and shortcomings of these models of teleological
or directed behaviour. What should be stressed/ however/ is the fact that teleological
behaviour directed towards a characteristic final state or goal is not something off
limits for natural sdence and an anthropomorphic misconception of processes whích,
in themselves, are undirected and accidental. Rather it is a form of behaviourwhich can
well be defined in scientific terms and for which the necessary conditions and possible
mechanisms can be Índicated.
* * *

General System Theory and the Unity of Science

Let me dose these remarks with a few words about the general implications of
interdisdplinary theory. The integrative function of general system theory can perhaps be
summarised as follows. Só far, the unitication of science hás been seen in the reduction
of ali sdences to physics, the final resolution of ali phenomena into physical events. From
our point of view, unity of sdence gains a more realistic aspect. A unitary conception
of the worlcf may be based, not upon the possibly futile and certainly far-fetched hope
finally to reduce ali leveis of reality to the leve! of physics, but rather on the isomorphy of
laws in dífferent fields. Speaking in what hás been called the 'formal' mode, ie, looking at
the conceptual constructs of science, thís means structural uniformities of the schemes
we are applyíng. Speaking in ^mgterÍaF language, it means that the worid, ie, the total of
observable events, shows structural uniformities, manifesting themselves by isomorphic
traces of order in the different leveis or realms.
We come, then, to a conception which in contrast to reductionism, we may call
perspecíivism. We cannot reduce the bioiogical, behavioural and social leveis to the
lowest levei, that of the constructs and laws of physics. We can, however/ fínd constructs
and possibly laws within the individual leveis. The world is, as Aldous Huxley once put
it, like a Neapolitan ice cream cake where the leveis - the physical, the bioiogical, the
social and the moral universe - represent the chocolate/ strawberry and vanilla layers. We
cannot reduce strawberry to chocolate - the most we can say is that possibiy in the last

Sï The Meaning of General System Theory

56 Computational Design Thinking
Systems Generatíng Systems In a properly functioning building, the building and the people in it together form a
whole: a social, human whole. The building systems which have só far been created do
ChristopherALexander not in this sense generate wholes at ali.

In his earlier writings, Christopher Alexander established the notion of 1. There are two ideas hidden in the word system: the Ídea ofa system as a whoie ana the
the unselfconscious process. Initially, this refers to the notion of process, jdea o f a genemtíng system.
in a social context, as being interrefational by which culture, building and The word system, like any technical word borrowed from common use, hás many
environment are concurrently formed. In this selected text, published in meanings and is impredse. This lack of precision in a technical word might seem dangerous
an issue of Architectural Design (1968), it is extended into the notion of at fírst; in fact it is often helpful. It allows new Ídeas to fíourish while still vague, ít aliows
how architectural problems may be solved through an analogous process connections between these ideas to be expiored, and Ít allows the ideas to be extended,
where design forms through the iterative readings and responses to Ínstead of having them cut short by premature definition and precision.
interrelational conditions, with the intention of producing environments Theword 'system' isjustsuch a word. Itstill hás many meanings hidden in it.Among
synchronouswiththeir cultural settings.lnherent in thispositionregarding these meanings there are two central ones: the Ídea of a system as a whole, and the idea
process and form is the substantial step that Alexander makes to transfer of o generating system.
the conception of form as an observed object to one as an externalised These two views, though superfícially similar, are logically quite different. In the fírst
operating system. In this article, he argues howsuch a system is bom, itself, case the word 'system' refers to a particular holistic view of a single thing. In the second
of a generative system, establishing the duality between the object as a case, the word /system/ does not refer to a single thing at afl, but to a kít of parts and
computing agent and the method as a computational process. Alexander combinatory mies capable of generating many things.
lays out particular aspects of such a process, describing three conditions:
the global behaviour, the components that form such behaviour and the I.Asystemasawholeisnotanobjectbutawayoflookíng atanobject. Itfocuseson
types of local relationships among those components. While these three some holístic phenomenon which can only be understood as a product of interaction
aspects characterise a system, this does not extensively describe the among parts.
process by which a system can be achieved. Alexander explains, through Let us consider some examples of holistic phenomena which need to be viewed
defíning four characteristics of a system, the distinction between the as systems.
behaviour, as a collection of actions, and the system which generates that The great depression is an obvious example of a holistic phenomenon. We cannot
behaviour, as a series of interactions, and at what levei of abstraction understand the depression, except as a result of interactíon among rates of consumption,
such complexity can be understood. capital investment and savings: the ínteractions can be specified in the form of equations;
if we follow these equations through to their conclusion, we see that under certain
1 There are two ideas hidden in the word system: the idea of a sysfem 05 o whole conditions they must always lead to a depression.
and the idea of a generating system. The stability of a candle flame is another example of a holistic phenomenon. Why
does it maintain approximately the same size and shape throughout its fíickering? In this
2 A system as a whole is not an object but a way of looking at an object. It case, the /parts/ are flows of vapourised wax, oxygen and burnt gases - the processes
focuses on some holistic property which can only be understood as a product of of combustion and diffusion give the interaction between these flows - and these
ínteraction among parts. ínteractions show us at what size and shape the flame will be approximately stable.
The strength of a rope is another example of a holistic property. This strength is a
3 A generating system is not a view of a single thing. It is a kit of paris, with rules result of interaction among the individual strands, caused by the twisting of the rope:
about the way these parts may be combined. untwisted, the ropers strength is governed by the weakest strand; twisted, the strands act
:together and increase their strength.
4 Almost every /system as a whole' is generated by a 'generating system'. If we wish Another example of a holistíc property is the relation between .input and output in
to make things which function as /wholes/ we shall have to invent generating |any computer In the toy computer called ThinkADot, a bali dropped Ínto one of three
systems to create them. 's comes out cm one of two sides. The output side is not determined by the Ínput hole,
ïut by the input hole and the internai state of the machine, which is itself determined by

58 Computatíonat Design Thinking 59 Systems Generating Systems

the sequence of past inputs. in arder to understand this behaviour, we must understand
the machine as a whole, considering the pgst inputs and the internai states as parts, and
the way that different sequences of inputs and internai states create specifíc new internai
states and outputs as interactions. IUI^IS AW KITO W N»ARTO
Another kind of holistic behaviour is that instability which occurs in objects that are *4~1"- ts-'' '' :-1>

very vulnerable to a change in one part: when one part changes, the other parts change
also. We see this in the case of erosion: cutting down trees robs the sói! of the roots which íí i; ;. " .* '•

holdittogethe^sothatwindandwatercanstripthesoilofallremaining pia nts, a nd make

a desert. We see it again in the death of the traditional farm: when the combine harvester
:.^ "^r^:. 4 ^ JL \0 ^
replaced traditional harvesting, the entire balance of scale economies was destroyed, the y^S^ ^4^ ÜAL.

little farms collapsed, and gave way to giant farms.

^ ^ F ^ L^
.'^ &,/=-' : V. ï <S . It

Let us summarise the content of these examples. !n every case we are confronted
with an object which displays some kind of behaviour which can only be understood as
a product of interaction among parts within the object. We cal! this kind of behaviour,
íiá ? ^ T t
<wmai AUU^I
holistic behaviour.
The central point ofthewhole argument can bestated verysimply. The mostimportant THISUÍOI PAitTS
properties whsch anything can have are those properties that deai with its stabílíty. It is
stability which gives a thing Íts essential character. The strength of an arch, the even

^DEFGHIJK
burning of a fíame/ the growth of an animal, the balance of a forest ecology, the steady
flow of a river, the economic security of a nation, the sanity of a human individual, the
health of a society: these are ali, in one way or another, concerned with stability.
Stability, no matter in which of Íts many forms, is a holistic property. It can only be
understood as a product of interaction among parts. The essential character of anything
whatever, since it must at heart be based on some kind of stability, must be understood
ORSTUV\
as a product of Ínteractions within the whole. When we view a thing in such a way as to
reveal its character in holistic temns, we speak of it as a system.
<WH« H AUKI^f
[n order to speak of somethíng as a system, we must be able to state clearly:
(1) the holistic behaviour which we are focusing on; C2) the parts within the thing, THISBKITOr PARTO
and the interactions among these paris, which cause the holistic behaviour we have
defined; C3) the way in which this interaction, among these parts, causes the holistic
behaviour defíned.
If we can do these three, it means we have an abstract working model of the holistic
behaviourinthething. Inthiscase, wemay properlycallthething a system. [fwecannot
do these three, we have no model/ and it is meaningless to call the thing a system. The
idea of g system is synonymous with the idea of an abstract model of some spedfíc hoiistíc
behaviour. We may speak of the economic system in a country, because we can construct
a system of equations which reproduce important holistic phenomena like depressions or
infíation. ff we couldn/t do this, it would be meaningless to speak of economic systems. <mBU:H MAILN^ Xt;VÏI^tl^
We must not use the word system, then, to refer to an object. A system is an
abstraction. It is not a special kind of thing, but a special way of looking at a thing. It is Image from original publication. Reprinted by pennission ©f Christopher Alexander.
a way of focusing attention on some particular holistic behaviour in a thing, which can © Christopher Alexander,

only be understood as a product of ínteraction among the paris. Everything under the

61 Systems Gen&rating Systems

60 ComputationalDesignThinking
sun may be viewed as a system: a man smoking a dgarette may be viewed as a system;
só may a leaf drifting in the wind; só may a brick; só may mankind on earth. But it only
becomes a system if we abstract from Ít some special holistic property, which we cannot
explain except in terms of interactions within the whole. Without a spedfic statement of
what holistic behaviour we have in mind, what interactions among what parts cause this
behaviour, and how they do só, calling a thing a system is no more than saying: This is a
pretty complicated thing, and l donrt understand it very well/ The ways in which man hás viewed the soiar system have resultfed in many ideas about its
The idea that a system is an abstraction needs emphasis. ThÍnk of a fíower as a system. structure. Á singíe set of objecte may be thought of as a system in a number of different ways.

If we want to understand the fact that the flower buds, and swells, and blooms -that we
must certainly do by looking at the flower as a system. In this case it is the Ínteraction
among the paris which aeates the behaviour of the whole. But the same flower hás other
properties which are not helped at ail by thinking of the fíower as a system: if it is used
as a projectile, then its trajectory cannot be explained as a result of interactions among
Íts parts,and if it is given as a gift, there is nothing that the fíower does, no matter how
complex the situation, that needs to be understood as a result of interactions among the
flower's paris. The idea of a system is helpful only in understanding kinds of behavíour
which result from interactions among parts.
Furthermore, even though we cail a thing a system when we try to view it as a whoie,
this does not mean that we ever really view the thing in Íts entirety.
When we look at an airline from g systems point of view, we may focus on its
scheduling - and we shall learn that because the aírline only hás a limited number of
aircraft, the schedule of a fíight from New York to Chicago turns out to be dependent on
the schedule of another fíight from IVIÍnneapolis to Salt Lake CÍty. In this instance, we are
looking at the airline 'as a whole', because we are looking at the interactions among parts,
but we are not concemed with the last button on the last mechanic's cap. The notion of
'whole' refersonlytothe breadth ofvlsion/ nottothe inclusion ofdetail: itisstill abstract.
Most often common Ignguage obscures this very badly. When we speak of the solar
system, or a hi-fi system, or an airline system or of a plumbing system, the words are used
in such a way as to suggest that the 'system^ is synonymous with the objects. But just
occasionally the word is used correctly, even in common language. For instance, when
we speak of the Ptolemaic system as opposed to the Copernican system, in each of the
cases the word system' is used correctly: it refers to an abstract way of looking at the
interaction among earth, planets, sun and stars - not to the objects themselves.
The discipline of abstraction hás one drawback. Occasíonally we are confronted with The ways m which man hás viewed the solar system have
resulted in many ideas about its stmcture, A single set of
phenomena which are clearly the products of interactions - but the interactions are só
objects may be thought of as a aystem in a number of different
complex that we cannot seç them clearly, and we cannot make the effort of abstraction ways. Image from original publication. Reprinted by permission
of Christopher Alexander. © Christopher AIexander.
successfully. Tgke, for instance, the baffíing complexity of a seagull landing, or of an
ecstatic, screaming, laughing girl. In these cases a too rigid insistence on the idea that a
system is an abstract model, might easily iead us to abstract out some facile inessential
system - at the cosi of the wonder which is really there.
This is exactly what happens when a systems analyst looks at g building: he/she
manages to describe the circulation, the acoustics, the heating ana the load-bearing

62 Computational Design Thinking 63 Systems Generating Systems

structureassystems-andfailsto Ídentifythe mostinteresting human and social systems, are numbers, variables, and signs like + and =. The rules specify ways of combining these
because he/she can't describe them in explidt terms. parts to form expressíons, ways of forming expressions from other expressions, ways
Thus, there is a second lesson to be leamed. The fírst lesson said: don't call a of forming true sentences from expressions, and ways of forming true sentences from
thing a system unless you can identify the abstract system you are talking about. The other tme sentences. The combinations of parts, generated by such a system, are the
second lessem says: learn the first lesson, but don/t let it railroad you into making true sentences, hence theorems, of mathematics. Any combination of parts which is not
facile abstractions. formed according to the rules is either meaningless or false.
When we are confronted with a complex thing, we often begin with nothing more A generating system, in thís sense, may have a very simple klt of parts, and very
than a feeling or a 'sense' that it functions as a system. Driven by this feeling, we then simple rules. Thus the system of triangles which may be put together to form a square,
try/ painstakingly, to abstract outjustthat holistic behaviourwhich seems essential, and is a generating system, Its rulesgenerateallthewaysofputtingthesetrianglestogether
those ÍnteractÍons which cause the behaviour. This is an active process. It begins wíth to form a square. It is typicat of a system that the mies mie out many combinations of
feeiing, and sensing, and only turns to thinking later. Start with some aspect of life só the paris. Thus these triangles could be put together in an infínite variety of ways ~ but
interwoven that you feel in your bonés it must be a system, only you can/t state it yet - most of these ways are ruled out, because the outside perimeter is not a square and
and then, once you can feel it dearly, try to pln the system down, by defíning the holistic this thing is not connected.
behaviour you are discussing and which interactions among which parts create Ít. But feel Another example of a generating system, is the system of language. Here we have
it clearly fírst, before you try to think it. rules at several different leveis. At one levei, the letters are the parts, and there are rules
Thesystems point ofview is not neutrgl. Itwill changeyourwholeviewoftheworld. which govern the way that letters may be put together to form words. In English there
It will lead you to realise that the most important characteristics of human individuais are could be no word beginning with Rx. The mies of phonology prohibit Ít. At another leve!,
products of their interactlons with other people. It will iead you to realise that the life of the words are themselves parts, and there are mies which govem the kinds of sentences
nations - though these nations may seem self-suffícient •" is produced by Ínteractions in which may be made from words.
the whole worfd, and that they only get their strength from their position in this larger Perhaps the most interesting and important generating system in the woríd is the
whole. It will lead you to see that the health of cities is produced by interactions among geneticsystem. Every animal in the animal kingdom Ísgenerated byasetofchromosomes
interdependent parts, induding houses, cafés and theatres, yes, but also equally induding spedfic to that animal. Each chromosome in tum is generated by four bases Clike a necklace
siums and graveyards. which uses only four kinds of bead). The four bases form a kit of parts which generates the
The system viewpoint is a modem, disciplined, version of the sense of wonder. chromosome. These chromosomes themselves provide the rules for building amino acids
!t is that view of things which man takes when he becomes aware of oneness and (another kit of parts), protelns from amino acids (another kit of paris), cells from proteins
wholeness in the world. (another kit of parts) and then build the animal from cells. The kit of parts fonned by
the four bases, and their mies of combínation, indirectfy generates every animal there is.
3. A generating system is not a view ofa singíe thing. tt is a kit ofparts, with rutes about A building system is a generatlng system in this sense. It provides a kit of parts
the way these parts may be combíned. - columns, beams, paneis, windows, doors - which must be put together according
This is a different use ofthe word system from the fírst one. In coiloquial English we to certain rules.
often use the word system to mean 'a way to do something': thatrs what a betting system
is; thafs what the Montessori system is; thafs what the democratic system is. 4. Aímost every 'system as a whoíe' is generated by a generating system. tf we wish to
Each of these systems is, at heart, a system of mies. A betting system tells you how make thmgs which function as /who!esf we shaií have to snvent generatíng systems to
to place your bets, the Montessori system lays down rules to be followed by children and create them.
teachers in nursery school, the democratic system of government lays down certain rules There is a relationship between the two ideas of system which have been defined.
about the nature of representation, the choice of representatives and the conduct of Almost every object with behaviour that depends on some 'system as a whole/ within the
elections. In ali these cases, the rules are designed to generate things. A betting system object, is Ítseif created by a generating system.
supposedly generates winning bets, an educational system generates well-educated Take an obviousand simple case: a hi-fi system. Its purityof performance can only be
pupils, the democratic system supposedly generates freedom and good govemment. understoodasa productofthecombined effectofallthevariouscomponents,working as
We mgy generafise the notion of a generative system. Such g system will usually a whole. The same hi-fi system is also generated by a generating system: the kit of ail the
consist of a kit of paris (or elements) together with rules for combining them to forrn parts on the market, and the rules governing the eléctrica) connections and impedance
allowable 'things'. The formal systems of mathematics are systems in this sense. The parts matching between these parts.

64 Computational Design Thinking 65 Systems Generating Systems

To take a more complicated case: the railroad switch-yard. It plainly functions as a system; yet the objects it produces do nothing: they have no holistic system propertles
whole. In order to understand it as a device for breaking up and making trains, we must whatever. In the same sense, those building systems which have só far been conceived
focus on the sequence of switches, and on the fact that the length of track in front of make buildings, but they do not make buildings with any really important holistic system
the switches depends on the iength of track behind the switches and on the length of propertíes. In a properly functioning building, the building and the people in Ít together
trains. At the same time, the switch-yard is also plainly generated by a generating system. form a whole: a social human whole. The building systems which have só far been created
The pieces of track, swítches, couplings, cars, together with the rules for putting them do not in this sense generate wholes at ali. While Ít is Ínherent in the generating system of
together, form a kit of parts which generates properly functioning switch-yards. an animal that the fínished animal will work as a whole, it is not inherent in any of todayrs
The most complicated case of ali, and the dearest, is that of an animal. A landing building systems that the buildings they produce will work as social or human wholes.
seagull certainly needs to be seen as a system: só does almost everything else that Creating building systems in the present sense is not enough. We need a new, more subtle
seagulls do. At the same time, this seagull is created by a generating system: the kind of building system, which doesn't merely generate buildíngs, but generates buildings
genetic system. An animal is both something which needs to be seen holistically, and guaranteed to function as holistic systems in the social, human sense.
genergted by a generating system.
The relationship between holistlc systems and generating systems is easy to ChristopherAlexander, 'Systems Generating Systems', Arclïítectural

understand. If an object hás some holistic property caused by interaction among parts Design, December issue No 7/6, John Wiley & Sons Ltd (London), 1968,
pp 90-1- Originally published in Systemat, ajournal ofthe Inland Steel
then Ít is clear that these particular parts and these particular interactions will only come
Products Company. Reproduced by permission of Christopher Alexander.
into being if the parts have very constrained reiationships to one another. The object
© Christopher Alexander.
then, must be generated by some process which assembles parts according to certain
constraints, chosen to ensure the proper interaction of these paris, when the system
operates. This is exactly what a generating system is.
The generating system need not be consdous (as in the case of the switch-yard),
nor even always explidt Cãs in the genetic case). Sometimes the processes which make
up the generating systems are integral with the object being formed - thus the candle
fíame is generated by chemical processes which are the same as those processes which
then maintain the systemrs equilibrium and make up the interacting parts, when we view
the fíame as a holistic system.
li is true then, that almost every 'system as a whole' is generated by a generating
system. This axiom contains a remarkable lesson for designers. Man as a designer is
concemed with the design and constmction of objects which function as whoies. Most of
the important properties a city needs to support life, for instance, are holistic properties.
Our axiom means this: to ensure the holistic system properties of buildings and
cities, we must invent generating systems, whose parts and mies will create the necessary
holistic system properties oftheir own accord.
This is a radical step in the conception of design. Most designers today think of
themselves as the designers of objects. If we follow the argument presented here, we
reach a very different conclusion. To make objects with complex holistic properties, it
is necessary to invent generating systems whích will generate objects with the required
holistic properties. The designer becomes a designer of generating systems - each capable
of generating many objects - rather than a designer of individual objects.
Afinalword ofcaution.Aswe havealreadyseen, a building system is an example ofa
generating system. It is a kit of paris with mies ofcombination. But not every generating
system necessarily creates objects with valuable holistic properties. The generating system
which makes squares out of triangles is an example. It is a perfectly good generating

66 Computational Design Thinking 62 Systems Generating Systems

The Architectural Relevance of Cybernetícs The argument rests upon the idea that architects are fírst and foremost system
designers who have been forced, over the last 100 years or só, to take an increasing
Gordon Pask Ínterest in the organisationai (ie/ nontangible) system properties of development,
communication and contrai. Design problems were coped with as they cropped up,
The discipline of cybernetics, of which Gordon Pask was a leading figure but for some time it hás been evident that an underpinning and unifying theory is
in the advancement of this intellectual domain, states/ in principie, that required. Cybernetics is a discipline which fílls the bill insofar as the abstract concepts
systems are based on regulation, control, adjustment and purpose, fíltered of cybemetics can be interpreted in architectural terms (and, where appropriate,
through means of feedback. The transdisciplinary domain, permeating such identified with real architectural systems), to form a theory (architectural cybemetics,
disparate fields as engineering, biology, sociology, economics and design, the cybernetic theory of architecture).1
institutes a paradigm for thinking which emphasises circular reasoning,
interrelating output, adaptation and self-organisation. Pask expresses Historical Roots2
these notions in this article, fírst published in Architectural Design In or before the early ISOOs 'pure/ architecture existed as an abstraction from the art
(1969), as the underpinnings to the comprehension of architecture as a of building. Its ruies were essentially condensed statements of what could be observed
compilation of active systems, in contrast to the perception of a building by looking at builders working on a site/ and by looking at buildings constructed during
as simply a static material object, where the engagement of the human is, different periods and in different places. Architects added a modicum of engineering
most critically, lost. He locates a source for this distinction in the contrast practíce and of historical or aesthetic sensibility to their discipline and created new
between design formulation based on language", defíning a fínite, often structures with stabilíty and síy/e. On the whole, their structures were judged, within
historically based, set of possibilities being inherently limited by style, 'pure' architecture, according to these canons.
and one derived of environment and action, responsive to the prospects Even in those days, of course, architects were asked to solve problems entailing the
of behavioural changes. Cybernetics, intended as a unifying theory and regulation and accommodation of human beings; hence, to design systems. But, in a
related to architecture here by Pask, contributes to computational thinking sense, their brief was quite narrow. The problems could ali be soived by the judicious
a logic for perceiving architecture as an environmental, social and cultural application of pure archítectural rules. The form of the artefact Chouse, college or
device, and proposes the fundamental components of design processes theatre) was largely determined by the quite rigid codes of architecture Cdictating,
which may calculate, determine and predict such systems. for example, its acceptable whole-part relationships) and by the conventions of
society or the individual practitioner. Speaking technically, there were well-accepted
It is easy to argue that cybernetícs is relevant to architecture in the same way communication media for conveying instructions, directives and ideas (style manuais
that it is relevant to a host of other professions; medicine, engineering or law. and só cm). Further, there was a metaíanguage for talking about these instructions,
PERT programming/ for example, is unequivocally a /cybemetic' technlque and directives and ideas, for comparing them, criticising them and evaluating them (as in
ít is commonly employed in construction scheduling. Computer-aided design is a statements of stability or style). Indeed, when interpreted, the body of metalinguistic
/cybemetic/ method and there are several instances of its application to architecture statements formed the theory of pure architecture. Consequently, architects did not
(for example, the W5CC's planning scheme in which the designer uses a graphic need to see themselves as systems designers/ even though they designed systems, and
display to represent the disposition of sti-uctural modules on a grid and in which the the evidence suggests that they did not do só.3 instead the professional Ímage was that
computer summarises the cosi effort consequences of a proposed layout). Of these of a sophisticated house, college or theatre builder.
cases the first (PERT programming) is a valuable but quite trivial application of In the course of the Victorian era new techniques were developed too rapidly
cybernetics; the second is likely to have a far-reachlng influence upon architectural to be assimilated ínto pure architecture and new problems were posed and could
design. But neither of them demonstrates more than a superficial bond between no longer be solved by applying the rules of pure architecture, for example, make
cyberneticsand architecture. Ifwe leavethe matteratthis levei, then archítects dive a 'railway station' or make a 'great exhlbition'. The solution to such (in those days)
intoa cyberneticbagoftricksanddrawoutthosewhichseemtobeappropriate.That outlandísh problems clearly depends upon seeing the required building as a part of
is a perfectly reasonable thing to do, of course. But cybemetics and architecture the ecosystem of a human society. Of course the problems were solved and the novel
really enjoy a much more Íntimate relationship; they share a common philosophy of techníques were mustered for this purpose (Temple Meads, the Tropical House at
architecture in the sense that Stafford Beer hás shown it to be the phiiosophy of Kew, the Crystal Palace). To my own taste the solutions are exceptionally beautiful.4
operational research. Nevertheless, they are individual and idiosyncratic solutions because, in the new

68 ComputaíionatDesignThinking 69 The ArchHectural Relevance of Cybernetics

context, there was no way of carrying on a general and criticai discussion. Let us Architectural Holism
be clear about this polnt. There obviously was a great deal of discussíon over IK Once a mdimentary version of the functional/mutu3ljstic hypothesis hás been
Bmnei, D Burton and J Paxton's use of glass and ironwork; technical discussion and accepted, the integrity of any single system is questíonable. Most human/stmctural
aesthetic discussion. But nobody seems to have appreciated the full signifícance of systems rely upon other systems to which they are coupled via the human components.
their structures in the context of the architectural potentialities of the age, ie, as By hypothesis, there are organisational wholes which cannot be meaningfully dissected
examplesofsystem design.The reason isfairlyobvious.Whereasthe purearchitecture into parts. Holism is of several types;
of the early ISOOs had a metalanguage, aibeit a restrictive one which discouraged
innovation, the new (augmented) architecture had not yet developed one. Another o A functionally interpreted building can only be usefully considered in the context
way of putting it is to say there was no theory of the new architecture.5 of a city (notíce that the city is also functionally interpreted and, as a result, is a
dynamic entity).
Architectural Sub-Theories
In place of a general theory there were sub-theories dealing with isolated facets of b A Cfunctionally Ínterpreted) structure, either a building or an entire city, can only
the fíeld; for example, theories of materiais, of symmetry, of human commitment and be meaningfully conceived in the context of íts temporal extension, ie, its growth
responsibility, of aaftsmanship and the like. But (it is probably fair to say) these sub- and development.
theories developed more or less independently during the late 1800s.
Naturally enough, each sub-theory fostered a certain sort of building or a certain c A (functionally interpreted) structure exists as part of an Íntention/ ie, as one
sort of socio-architectural dogma; for exampie, Futurism. However, the point of productofap/on.
immedíate interest is that many of the sub-theories were system orientated; aithough
they anticipated the invention of the word they were, in an embryonic sense, d If (assumed dogma) man should be aware of his natural surroundings, then
/cyberneíic/ theories and the thinking behind them made a valuable contribution to the buildings should be wedded to or arise from these surroundings (Wrighfs
development of cybernetics as a formal science. organicthesis).

Architectural Functionalism and Mutualism It is a corollary of o, b and c that the structure of a city is notjust the carapace of
A structure exísts chiefly to perfomn certain functions, for example, to shelter its society. On the contrary, its structure acts as a symbolic control programme on a
occupants or to provide them with services. At this levei, a 'functíonal' buílding is par with the ritual constraints which are known to regulate the behaviour of various
contrastedwitha/decorative/building;itisanausterestructure,strippedofexcrescences. tribes and which render this behaviour homeostatíc rather than dívergent. Hence,
But, the concept of functionalism can be usefully refined in a humanistic direction. the architect is responsible for building conventions and shaping the development
The functions, after ali, are performed for human beings or human societles. It follows of tradítions (this comment simply elevates the idea that a building contrais its
that a building cannot be viewecf simply in isolation. It is only meaningful as a human inhabitants to a higher levei of organisation).
environment. It perpetually ínteracts with its inhabitants, on the one hand serving them
and on the other hand controlling theirbehaviour. In otherwordsstructures makesense Evolutionary Ideas in Architecture
as parts of larger systems that include human components and the architect is primarily Systems, notably cities, grow and develop and, in general, evolve. Clearly, thís
concernedwiththeselargersystems; ttíeyCnotjustthe bricksand mortar part) arewhat concept is contingent upon the functionalist/mutualist hypothesis (without which it
architects design. l shall dub this notlon architectural 'mutualism' meaning mutualism is difficult to see in what sense the system itself does grow), though the dependency
between structures and men or societies. is often unstated. An immediate practical consequence of the evolutionary point of
One consequence of fuhctionalism and mutualism is a shift of emphasis towards the viewisthatarchitectural designsshould have rulesforevolution builtintothem iftheir
form (rather than the material constitution) of structures; materiais and methods come growth is to be healthy rather than cancerous. In other words/ a responslble architect
into prominence quite late in the design process. Another consequence is that architects must be concerned with evolutionary properties;,he cannot merely stand back and
are required to design dynamic rather than statíc entities. Clearly/ the human part of the observe evolution as something that happens to his structures. The evolutionary
system is dynamic. But it is equally true Cthough less obvious) that the structural part is closely related to holism, type c, but it is a carefully specialised version of c
must be imaged as continualfy regulating its human inhabitants. as manifest in the work of the Japanese.

70 CompuíationalDesignThmking 71 The Ardittectural Retevance of Cybernetics

Symbolic Environments in Architecture development of educational technology and to provide for their impact upon whatever
Many human activities are symbolic in character. Using visual, verbal or tactile symbols, structure is erected. By token of this the architect quite often comes into the picture at
man "talks with/ his surroundings. These consist in other men, ínformation systems such the time when a higher educational system is being contemplated, without commitment
as libraries, computers or works of art and also, of course, the structures around him. to whether or not it is called a uníversity. The Fun Palace project, by Joan Littlewood and
Buildings hgve always been dassified as works of art. The novel sub-theory is that Cedric Price, was an early entry project of this type in the fíeld of entertaínment and Ít is
structures may be desígned (as well as intuited) to foster a productive and pleasurable not difficult to find examples in áreas ranging from exhibition design to factory building.
dialogue. This way of thinking is most clearly manifest in connection with the literary art The point l wish to establish is that nowadays there is a demand for system orientated
forms, notably Surrealism which relles upon a variety-(novelty-) producing juxtaposition thinking whereas, in the past, there was only a more or less esoteric desire for Ít. Because
of releasers and supernormal stimuli (evoking inbuilt emotive responses) within a thematic of this demand/ it is worthwhile collecting the isolated sub-theories together by forming a
matrix. At the architectural levei, this type of design appears in the vegetable surrealism generalisation from their common constituents. As we have already argued, the common
ofsomeoftheArt Nouveau. Butit reaches maturityin Gaudí'swork, especiallythe Parque constituents arethe notions of control, communication and system. Hence the generalisation
Güell which, at a symbolic levei, is one of the most cybernetic structures in existence. is no more nor less than abstract cybernetics Ínterpreted as an overall architectural theory.
As you explore the piece, statements are made in temns of releasers, your exploration It would be premature to suggest that the necessary interpretation and consolidation
is guided by specially contrived feedback, and variety Csurprise value) is íntroduced at is complete. But a creditable start hás been made by a number of people; citing only
appropriate points to make you explore. those with whom l have personal contact, Christopher Alexander, Nicholas Negroponte,
It is interesting that GaudFs work is often contrasted with functionalism. Systemically many students and ex-students from the AA School of Architecture and from Newcastle.
ít/s functionalism pure and simple, though St is aimed at satisfying on/ythe symbolic and
informational needs of man.6 Status ofthe NewTheory
In commonwiththe pure architectureofthe ISOOs, cybernetics provides a metalanguage
The MachineryofArchitectural Production for criticai discussion. But the cybernetic theory is more.than an extension of 'pure'
Just as a functionally interpreted building constitutes a system, só aiso the construction architecture. As we noted somewhat earlier, pure architecture was descriptive (a
of this building is a system. The new techniques developed in the last century and the taxonomy of buildings and methods) and prescriptive Cãs in the preparation of plans)
general mechanisation of production facilities led to sub-theories concerned with the butitdid littleto predict orexplain. In contrast, the cybernetictheory hás an appreciable
achievement of forms Cthe most important centred around the Bauhaus) and these, in predictive power.7 For example, urban development can be modelied as a self-organising
turn, restricted the forms that could be produced. system (a formal statement of 'evolutionary ideas in architecture') and in these terms it is
possible to predict the extent to which the growth of a city will be chaotic or ordered by
TheWidening Brief differentiation. Even Íf the necessary data for prediction is unavailable we can, at least,
Asa resultofthese, essentiallycybernetic,sub-theoretical developments, manyarchitects pose and test a rational hypothesis. Much the same comments apply to predictions in
wantedto design systems but, on thewhole, they were expected to design bulldÍngs.To a whichtimeisnotofprimaryimportance;forinstance, in predictíngtheinfluenceofspatial
large extent this is still (quite reasonably) true. Al! the same, there is a sense in which the and normative constraints upon the stability of a (functionally interpreted) structure.
brief given to an architect hás widened during the last decades. The cybernetlctheory can also claim some explanatory power insofar as it is possible to
In part this is due to a spate of problems for which no conventional solution exists mimic certain aspects of architectural design by artificial intelligence computer programs8
(structures connected with aerospace developments, industry, research, entertainment, Cprovided, Índdentally, that the program is able to learn about and from architects and by
the use of oceans, etc). Here, the architect is in much the same position as his Victorian experimenting in the language of architects, Íe, by exploring plans, material specifícations,
predecessorwhen askedto build a railwaystation. In part, however,therestraintshave been condensed versions of clients" comments, etc). Such programs are clearly of value in
relaxed becauseofthegreatêrprevalenceofsystem orientated thinking among clients and their own right. They are potential aids to design; acting as intelligent extensions of the
public sponsors. It is, nowadays, legitimate to enter the design process much earlier, even tool-like programs mentioned at the outset. Further, they offer a means for integrating
for a conventional project. For example, it is quite commonplace to deslgn (or at least to the constructional system Cthe 'machinery of prQduction') with the ongoing design
plan) cities as a whole with provision for their evolution. A university need not be conceived process since it is quite easy to embody the constraints of current technology in a special
as a set of builctings around a courtyard with living accommodation and lecture theatre. part of the simulation. However, l believe these programs are of far greater importance
The educational system might, in certain drcumstances, be spatially distributed ratherthan às evidendng our theoretical knowledge of what architecture is about. Insofar as the
localísed. In any case, architects are positively encouraged to anticipate trends such as the Program can be written/ the cybemetíc theory is explanatory.

72 ComputationalDesignThinkmg Ï3 The Architectural Relevance of Cybemetics

Speculations a human being in contact with Ít. If there is a human being in the environment,
It seems likely that rapid advances will be made in at least five áreas guided by the the computer, material and ali, engages him in dialogue and, within quite wide
cybemetic theory of architecture. limits, is able to learn about and adapt to his behaviour pattern. There is thus one
sense in whích the reactive environment is a controiler and another in which Ít is
1 Various computer-assisted (or even computer-directed) design procedures will be controlled by its inhabitants.
developed into useful instruments.
ASimple Cybernetic Design Paradigm
2 Concepts in very different disciplines (notably social anthropology, psychology, In the context of a reactive and adaptive environment, architectural design takes place in
sociology, ecology and economics) will be unifíed with the concepts of several interdependent stages.
architecture to yield an adequately broad view of such entities as /civilisation/,
'cit/ or 'educational system'. 1 Specification ofthe purpose or goal ofthesystem (with respecttothe human
Ínhabitants). It should be emphasised that the goal may be and nearly always
3 There will be a proper and systematic formulation of the sense in which witt be underspecifíed, ie, the architect will no more know the purpose of the
architecture acts as a social contrai Cie, the germ of an idea, mentioned under system than he reatty knows the purpose of a conventional house. His aim is
/Holism/, will be elaborated). to provide a set of constraints that allow for certain, presumably desirable,
modes of evolution.
4 Thehigh pointoffunctionalism isthe conceptofa house as a 'machineforliving
in'. But the bias is towards a machine that acts as a tool serving the ínhabitant. 2 Choice ofthe basic environmental materiais.
This notion will, l believe/ be refíned into the concept of an envircmment with
which the inhabitant cooperates and in which he can externafise his mental 3 Selection ofthe invariantswhich are to be programmed intothesystem. Partlyat
processes, ie, mutugiism will be emphasised as compared with mere functionalism. this stage and partly in C2) above, the architect determines what properties will
For example, the machíne for living in will relieve the inhabitant of the need be relevant in the man-environment dialogue.
to store infomnation in memory and the need to perform calculations as well
as helping out with more obvious chores líke garbage disposal and washing up 4 Specification of what the environment will learn about and how it will adapt.
dishes. Further, it will elicit his interest as well as simply answering his enquiries.
5 Choice of a plan for adaptation and development. In case the goal of the
5 GaudÍ Cintentionally or not) achieved a dialogue between his envíronment and its system is underspecifíed, as in CD, the plan will chiefly consist in a number of
Ínhabitants. He did só using physically static structures (the dynamic processes evolutionary principies.
dependinguponthemovementofpeopfeorshifts in theÍrattention^.The dialogue
can be refined and extended with the aid of modern techniques which allow us Of course, this paradigm applies to systems which adapt over rather short time intervals
to weave the same pattern in terms of a reactive environment. If, in addition, the Cminutesorhours). In contrast, the adaptation in a projectsuch asthe Fun Palacesystem
environment is malleable and adaptive the results can be very potent indeed. i took place over much longer time intervals (for instance, an eight-hourly cycle and a
have experimented along these lines myself9 but the work of Brodey and his group weekly cycle formed part ofthe proposal). Depending upon the time constraints and the
atthe environmental ecology laboratory is a project on a much more Ímpressive degree of fíexibility required, it is more or less convenient to use a computer (for example,
scale. As a broad statement of what is going on, a computer controls the visual the weekly cycle is more economicaily programmed by a ffexible oftice procedure). But
and tactile properties ofenvironmental materiais (which are available in sufficient exactly the same principies are involved.
diversityformostarchitectural purposes^.These materiais contain sensors, tactile Urban planning usually extends over time periods of years or decades and, as
or visual as the case may be, which retum messages to the computer at several currently conceived/ the plan is quite an infíexible spedfication. However, the argument
leveis of generality. In the absence of a human inhabitant, the feedback leads to Just presenteei suggests that it need not be inffexible and that urban development could,
stabilisation with respectto certain preprogrammed ínvariants Cfor example, that perhaps with advantage, be governed by a process like that in the dialogue of a reactive
a body of material shall maintain mechanicaf stability and occupy a prescribed environment Cphysical contact with the Ínhabitants giving place to an awareness of
value), and to a search process in which the material actively iooks for signs of tiieir preferences and predilections; the inflexibie plan to the environmental computing

74 Computational Design Thinking Ï5 The Architectural Retevance of Cybernetics

Cleariy, in other respects, it would be uncomfortabïy prickly to live m.
machine). Ifso, the same design paradigm applies, since in ali ofthe cases só far considered The impact of cybemetics upon architecture is considerable just because the theory does liave
much more predictive power tban pure architecture had. Cybemetics did relatively litüe to alter
the primary decisions are systemic in charactec ie, they amount to the delineation or the
the shape ofbiochemistryforinstance, because although these concepts are bound up with
modifícation of a contrai program. This unlversality is typical of the cybernetic approach. everything Erom enzyme organisation to molecular biology, the discipline of biochemistry already
One final manoeuvre will indicate the fíavour of a cybemetic theory. Let us turn the had a predictive and explanaïoryfheorf ofits own, I made the same point for engineering in an
eariierfootnote.
design paradigm in upon itself; let us apply ít to the interaction between the designer and
I have the work of Negroponte's group (see pp 78-85) chiefly in mind, though there are other
the system he designs, rather than the interaction between the system and the people exemplais,
who inhabit it. The glove fíts, almost perfectiy in the case where the designer uses a For example, the Colloquy of Mobiles project and the Musicolour system, A Commení, A Case
History and a Piau in ComputerArt, (ed) Jasia Reichardt.
computer as his assistant. In other words, the relation 'controller/controlled entity' is
10 The cybemetic notions mooted in this article are discussed in An Approacli to Cybemeíics,
preserved when these omnibus words are replaced either by 'designer/system being Hutchinson (London), 1961 (paperback 1968) and, in a lighter vein, in 'My Predictions for 1984' m
deslgnecT or by 'systemic environment/inhabitants' or by "urban plan/dty'. But notice Prospecï, TIieScIiweppesBookoftlieNewGeneration, Hutehmson (London), 1962,

the trick, the designer is controlling the construction of control systems and consequently
Gordon Pask, 'The Archítectural Relevance of Cybernetics', Arcliítectural
design is contrai of contrai, ie/ the designer does much the samejob as his system, but he
Desígn, September issue No 7/6, John Wiley & Sons Ltd (London), 1969, pp
operates at a higher levei in the organisational hierarchy.
494-6. © John Wíley & Sons Ltd.
Further the design goal is nearly always underspedfied and the 'controller' is no
longer the authoritarian apparatus which this purely technical name commonly brings to
mind. In contrast the controller is an odd mixture of catalyst, crutch, memory and arbiter
These, l believe, are the dispositions a designer should bring to bear upon his work (when
he professionally plays the part of a controller) and these are the qualities he should
embed in the systems Ccontrol systems) which he designs.10

Notes
l Very similar comments apply to engineering, since engineers, like architects, prescribe artefacts.
Surely, also, some engineers make use of a cybemetíc fheory. But the requirement is not só
ubiquitous in engineermg; nor is the impact of cybemetics só great because a creditable body
of engineering theory, a predictive and explanatory theory, existed long before the cybernetic
concepts carne along as daring innovations. Moreover, whüe ali architects design systems that
interact closely with human beings and societies, most engineers (there are obvious exceptions)
are not forced to do só, Human interaction is a major source of difficulties which can only be
overcome by cybemetic thinkmg,
2 The choice of a historical origin is somewhat arbítrary and depends upon the author's emphasis,
For example, Alexander, preoccupied with the logic ofform, traces essentially cybemetic
concepts back to Lodoli and Laugier, In the present article I am anxious to foUow the pragmatic
development of cybemetic ideas and to see them emerging in the history of modem architecture.
3 There are two important sorts ofexception:
(i) Architects of genius, with a breadth ofvision that impeis fhem to see things in a systemic and
interdisciplmary fashion. Theyhave existedovertheyears: Sir Christopher Wren and Sir Jolm
Soane, for example.
(ii) Men like John Nash, whose talents lay in conceiving an urban development as a functional and
aesthetic whole. But, within the tenets ofthe early 1800s, such men are probably 'organisers with a
vision', ratherthan 'architects'.

4 I have chosen these examplçs partly because Üiey are weU known in the textbooks but mamly
because I am impressed by their sysíemic qualities and the way in which they convey their
designeis purpose to the occupant, Two ofthem stül exist. I just recaïL the Palace. Even in its
tawdry reincamation it was a remarkable structure. Since ü was one of the fest mstances of a
prefabricated building it also counts as a piece of sysíem design at the engineering levei.
5 Lack of an adequate metalanguage was not toe only factor. As Nikolaus Pevsner points out, ±e engineers
and üie artists pursued divergent paths ofdevelopment more or less in conflict with one another and
this accounted for at least some of the architectural idiosyncrasy. However, if a metalanguage bad
existed, fhen the synthesis ofthe present century could have been achieved much earlier.

76 Computational Design Thinking H The Architectura^ Relevance of Cybemetics

Towards a Humanism Through Machines We shall consider only the third possibility and shall treat the problem as the intimate
association of two dissimilar species: man and machine; and two dissimilar processes:
Nicholas Negroponte design and computation. We shall further define our concern as the acquaintanceship of
two intelligent systems, the architect and the Architecture Machine. By virtue of ascribing
With the emergente of the computer and computerised methods in design intelligence to an artefact or the artificial, the partnership is not one of master (smart,
in the 1960s, Nicholas Negroponte, a leading exponent of the infusion of leader) and slave (dumb, follower), but rather of two associates which each have the
computer processes to architecture, focused particularly on the relationship potential for self-improvement
between designer, computer and the media of instruction, feedback and Imagine a machine which could respond to local situations in the physical environment
intelligence. His conception of what he calls "computer-aidecT is more aligned Ca family that moves, a residence that is expanded, income that decreases). Such a device
with a computational approach, proposing the transformation ofthe computer could report on and concern itself specifícally with the unique and the exceptional. In
from the "subservient machine to the intelligent environment, described in effect, it would concentrate on the particulars. The human designer cannot do this. He
Negroponte's Refíections on ComputerAids to Design and Architecture (1975). cannot accommodate the particular; he obliges the general. Britton Harris suggests, /He
The context for Negroponteï proposal of an Architerture Machine' is to counter is forced to proceed in this way because the effectuation of planning requires rules of
the designeis tendency to, within his own methods, eschew the complexity of general applicability and because watching each sparrow is too troublesome for any but
large-scale problems, and generalises the specifícity of small-scale concerns. As Cod\ The reader surely needs little reminder of the results.
described in this text from an issue of Architectural Design Cl 969), the dialogue
between designer and machine functions only through recursive exchange and Prelude to an Architect-Machine Dialogue
reply, moving towards specifidty of procedures, conditions and information. Considerthat you are in a forelgn country, do not knowthe language, and are in desperate
While Negroponte refers to the physicality of computers, their components and need of help. At first your hand movements and facial expressions carry most of your
drcuitry, the notion of transparency, exchange and contextual specifícity are meaning to the silent observer. Your behaviour uses a language of gestures and strange
ali relevant aspects of a computational approach. Of particular importance is Cto the observer) utterances to communicate your purpose. The puzzled listener searches
the consideration that the language of a dialogue between man and machine is for bits of content he can understand and link to his own language. You react to his
particularly foreign to that of conventional design processes. This accentuates reactions and a language of pantomime begins to unfold. This new language hás evolved
the distinction between computational processes and computerised procedures, from the mutual effort to communicate. Returning to the same person a second time,
seen basically as simple translations of standardised means. with a new need, you will fínd that the roots of a dialogue already exist. But this second
conversation might be gibberish to a third party observing the exchange for the fírsttime.
Given that the physical environment is not in perfect harmony with Everymanrs life style; A designer-to-machine introduction should be a similar linguistic evolution.
given that architecture is not the faultless response to human needs; given that the Each should track the other^s design manoeuvres, evoking a rhetoric that cannot be
architect is not the consummate manager of form and use; let us conslder the evolution antidpated. The event is circular since the designer-machine unity provokes a dialogue
of physical environments. In particular lei us consider an evolution aided by the use of a and the dialogue promotes a stronger unity. This progressively Íntimate association of
specific dass of machines, as Warren McCulloch calls them, ethical robots. In the context two dissimilar specíes is the symbiosis. It evolves through mutual training - in this case/
of architecture, we shall call them Architecture Machines. through the dialogue.
A man-machine dialogue hás no history. The historical and presently antagonised
There are three possible ways of having machines assíst the deslgn process: mismatch (non-dialogue) between man and machine hás generated a great deal of
preoccupation for Ít. In less than a decade, the term 'man-machine communication/ hás
1 Current procedures can be automated, thus speeding up and reducing the cost of passed from concept to cliché to platitude.
existing practices. The theory, however, is important and straightforward: in order to have a cooperative
2 Existing methods can be altered to fít within the specifícations and constitution of interaction between a designer of a certain expertise and a machine of some scholarship,
a machine, where only those Íssues are considered that are supposedly machine- the two must be congenial. They must, for example, share the labour of establishing a
compatible. common language. Thus a designer, when addresslng a machine, múst not be forced to
3 The process, considered as being evolutionary, can be introduced to a mechanism (also resort to machine-oriented codes. And in spite of computational effíciency, the paradigm
consldered as evolutíonary), and a mutual training, resilience and growth can be developed. forfruitful conversations must be machínes that can speak and respond to natural language.

78 Computationat Design Thinking Ï9 Towardsa HumanismThrough IVlachines

With direct, fíuíd and natural man-machine díscourse, two former barriers between ripening of ideas and ways. At fírst, Jobs in which the man is particularly inept would
architectsand computing machines could be removed. First, thedesigners using computer- stimulate a nontrivial need for cooperation. Subsequently, each interlocutor would opt
aided design hardware would not have to be specialists. WÍth natural communication, out of situations notably clumsy for his constitution, while at the same time he would
the /this is what l want to dor and 'can you do Ít?/ gap can be bridged. The design task pry into issues which were originally outside his scope of concem Cor the concern of his
would no longer be described to a knobs and dials person to be executed in his secret profession). Eventually, a separation of the parts could not occur; the entire /symbiotic/
vernacular. Instead, the job would be formulated and executed in the designeis own system would be, as Gordon Pask described, 'an artifícial intelligence that cannot be
idiom, wíth simple negotiations. A vibrant stream of ideas could be directly channelled partitioned'. Itwould be computer-aided design.
from the designer to the machine Cand back).
Thesecondobstructionovercomebysuchdosecommunionwouldbethepotentialof Computer-Aided Versus Computerised
re-evaluating the procedures themselves, In a direct exchange, the designer could exercise /ComputerisecT operations are too often misnamed 'computer-aided/. The computerised/
his proverbial capriciousness. At fírst, a desígner might have only a meagre understandíng computer-aided distinction is too often confused with or solely embodied in the mode of
of his specifíc problem and thus require machine tolerance and compatibility in his search machine usage.
for the consistency among critería and form and method/ between Íntent and purpose. The conventional mode of computer usage forthe past 20 years, 'batch processing',
The progression from visceral to intellectual could be articulated in provisional statements entails a computation centre to which a user delivers a 'progranY Ca deck of cards,
that converge on both design and methods. magnetic tape, paper tape) to be /runr. Then severa! hours or days later the user retums to
But the tête-à-tête must be more direct and fluid; it is gestures, smiles and frowns receive his 'output'. More recently, real-time computation depending upon /time-sharing/
which turn a conversatíon into a dialogue. In an intimate human-to-human dialogue, techniques allows the user a prompt machine response and permits terminais (usualiy
hand-waving often carnes as much meaning as text. The 'manner' carries cultural teletypes) to reside in the office or at home. These terminais are connected to the large
informatíon; the Arabs use theír noses, the Japanese nod their heads. Customarily, in central machine, and they can be Ínterconnected with each other. The rapid switching of
machine communication studies, such "manners7 are ignored and frequently are referred users' programs in and out of the large machine provides each user with the illusion of a
to as 'noise'. But such silent languages are not noise; Warren Brodey and Nilo LÍndgren dedicated machine and pemnits him continuai use of his terminal.
submit that a dialogue is composed of 'whole body involvement ~ wíth hands, eyes, It is commonly suggested that the online nature of the interaction in a time-sharing
mouth, facial expressions - using many channels simultaneously, but rhythmised into a system is in itself a dialogue and transforms computerised procedures into computer-
hamnoniously simple exchange'. aided emes. This is simply not true. For exgmple, let us suppose you desire the average
Imagine a machine that could foilow your design methodology and at the same apartment-to-parking-space distance for some design project. In a batch processing
time discern and assimilate your conversationaí idiosynaasies. ThÍs same machine, mode Cpresuming the program exists), you supply as data the description of your design
after observlng your behaviour, could build a predictive modei of your conversational and the answer returns hours later - indeed a computerised procedure. On the other
performance. Such a machine could then reinforce the dialogue by using the predictive hand, in a 'real-time' environment, the project description resides in the machine, and
model to respond to you in a manner that is in rhythm with your personal behaviour. you simplytypeonyourteletype terminal theapartment-to-parking distance command.
This dialogue would be só intimate (even exclusive) that only mutual persuasion and But just because the answer comes back in three seconds rather than three days/
compromise would bring about perceptions and ideas - ideas/ in fact, unreaiisable by /computerised/ does not become /computer-aided/. It simply becomes more convenient.
either converser alone. In such a symbiosis, it would not be solely the desígner who would Computer-aidedness demands a dialogue; events cannot be merely a fast-time
decide when the machine is relevant manifestation of cause and effect.
The overlaying of a specific design character upon a generalised machine is not Online communication, therefore/ is not a sufficient (though necessary) condition for
fanciful; computer programs exist that illustrate some primitive attempts. An anonymous a computer-aíded environment. Computer-aided design requires three further features:
machine, after identifying a speaker, can transform itself into an exclusive apparatus that mutual interruptibility for man and for machine, local and decficated computing power
reflects previous encounters with that speaker. The extent ofthe metamorphosis depends within the terminal and a machine ÍntellÍgence.
an the length of acquaintance. Atthe onset ofthe partnership, the machine gathers gross Interruptibility gives a dimension of interaction that allows the process, as well as the
features; later it avails ítself of subtleties. product, to be manipulated. In a computer-aided system, the machine may interruptthe
It might be argued that we are proposing the creation of a design machine that is user and present unsolicited Ínformation - for example, that the cost of his low-income
an extension of and in the Ímage of a designer who, as he stands, hás already enough housing project is $38 per square foot. The apparent high cost might have been due to
error and fault. However, we have Índicated that the maturation would be a reciprocal 'nadequacies in the original estimating routine provided for the computer; for Ínstance,

80 ComputatíonalDesignThinking 81 Towards a Humanism Through Machnes

substantial indirect savings might have been overlooked. In this case, the designer could
tamper with the estimating routine and incorporate hitherto neglected parameters. Thus
the architect might welcome the remark, ignore it, or take offence and request that such
jnterludes of fínance be restricted.
Unfortunately, the present time-sharing philosophy fosters a cause-and-effect
conversation. Time-sharing assumes that a designeis explicit manipulations will occupy
between 1 and 10 per cent of any sitting; the remaining time represents his deliberations
and distractions. Each user's moments of contemplation are in effect another user's
Ínstants of computation. A designer can interrupt his own program, but a routine cannot
easily ínterrupt Íts partner-in-thought. This is because, in arder to leave the computational
utility available for other users, each routine resides in the machme only when expllcitly
called into service by its particular user. In otherwords, the user/s machines are encouraged
to listen, but not to ínterrupt.
To retain assets of 'time-sharing', while avoiding the anathema of batch processing,
and to acquire mutuaf interruptibility, we must adjustthe allocation ofcomputing power.
Some information processing power and a certain manipulative and storage capacity must
be transferred to the terminal that was originally a teletype transmission and receptive
device. This semi-autonomous terminal (possibly portable) would become a small
computerthatwould be a "machine in resÍdence'.AnArchitecture Machinewould besuch
a machine. The designer would speak directly to this satellite machine. In turn, this small
remate computerwould interactively converse wíth larger parent machines. (Sendíng out
work to a central mechanism would not be apparent to the designer; his Architecture
Machine would elect this course for reasons of speed, memory, information, or ali three.)
It is the Architecture Machine in residence/ located in habitation wíth the designer,
which would undergo the personalisation. It would be composed of additive and
subtractive pieces of hardware as determined by the discipline of its partner. The local
consortium of parts would do the interrupting/ the dialoguing and the evolving. Observe
that the interrupting and the re-intermpting would be dependent on the nature of the
designeis activities, on the context of his efforts. Through familiarity with the specific
designer'5 idiosyncrasies/ the appropriateness of the machine/s Ínterruptions can be
suitably reinforced by context; this is the inception of an intelligent act.
A mechanical partner, as we have suggested, must have intelligence. Customaríly,
computer-aided design studies and intellígent automata studies have been antipodal
efforts. On the one hand, we are told to render unto each their respective design
functions and talents: man (inteiligent) thinks and the machine (dumb, fast) calculates.
UEBAN5, a primitive effort at man-machine
dialogue useful in architectural design. (IHustration On the other hand, we are told that, Anything you can do, a machine can do better/.
is from a motion picture, URBAN5, by Professor The two outlooks are not necessarily contradictory. There is a real issue whether machine
Negroponte in cooperaüon with Lêem B Groisser.)
Image from original publication, Reprinted by intelligence can be independent of human intelligence. In computer-aided design, only
permission ofMITTechnologyReview. © MIT the combination of mechanical amplification and mechanical imitation will validate
Tecïmology Review.
the dialogue. The dialogue will evolve an intelligence, this intelligence wiil stimulate
a more profound dialogue, which in turn will promote further intelligence/ and só on.
Furthermore, the concurrence of 'extended designei and 'artificial designer' will force a

82 Computationat Design Thinking 83 Towards a Humamsm Through Machines

design redundancyandanoverlappíngoftaskswhicharenecessaryfortheunderstanding Forfurther reading on computer processes in design and architecture see
of Íntricate desígn couplings. Perpetuai cross-examination of ideas by both the man and Nicholas Negroponte (ed), Refíectíons on ComputerAíds to Design ana
themachinewill encouragecreatívethoughtthatwould otherwise beextinguished bythe Architecture, Petrocelli/Charter, NewYork, 1975.
lack of an antagonistic (thus challenging) environment. Computer-aided design concerns
an ecoiogy of mutua! design complementation, augmentation and substitution. Nicholas Negroponte, 'Towards a Humanism Through Machines',
Archítectural Design, September issue no 7/6, John Wiley & Sons Ltd
Vis-à-vis IVlachine Intelligence (London), 1969, pp 511-12. Originally published by Technology Review
Intelfigence is a particularly difficult behaviour to emulate in machines because of its Inc in Technology Review, vol 71, no 6, April 1969. Technology review by
Nicholas Negroponte Copyright 1969. ReproducedwithpermissionofMIT
extreme dependence on context - time, locality, culture, etc. A sophisticated set of
Technology Revíew in the formai Otherbookvia Copyright Clearance
sensors, effectors and processors is needed to viewthe real world Cdirectfy or Índirectly)
Center. © 2002 MIT Technology Review.
andtodiscemchangesin meaning broughtabout bychanges in context-in otherwords,
to be intelligent.
For example, the meaning of a metaphor (in the physical environment, in a story,
in a painting) is conveyed through context, and the assessment of such meaning is an
example of an intelligent act. CNote that a literary metaphor characterises the era and the
culture in which Ít was written.) One mightjudge a machine's intelligence Cnot necessarily
its maturity, wisdom or knowledge) by its abifity to apprecíate a joke - a joke being a
funny story with a punch Une that is an about-face in context As humans we exhibit an
intelligence by tracing back through previous metaphors of the tale and deriving pleasure
from the new meanings revealed and brought on by a shift in context. ÇNote that people
of dífferent cultures have difficulty in understanding each other'sjokes.)
Some architects might propose that machines cannot design unless they can think,
cannot think unless they want, and cannot want unless they have bodies and, since
they do not have bodies, they therefore cannot want, thus cannot think, thus cannot
design ~ quod erat demonstrandum. The argument, however, is usually an emotional
issue ratherthan a logícal conclusion. Nonetheless, the reader must recognise (Íf he is a
'machíne-intelligence' enthusiast) that theories on machine Íntelligence can, at this time,
best be supported with such examples as computers playing a superb game of checkers
and a medíocre game of chess. And furthermore, architecture, unlike checkers Cwith fíxed
rules and a fixed number of pieces) and much like ajoke (determined by context), is the
croquet game in Alice in Wonderiand, where the Queen of Hearts Csociety, technology,
economics) keeps changing the rules.
Let us not be misled. We are not interested in a machine that will simply parrot
a human desígner, nor are we Ínterested in a machine that wíll have an autonomous
existence bywhich to mimic and replace an archítect. An Architecture Machine will feature
a dependence. An artificia! intelligence is in fact an Ínterdependence.
You might summarise the proposition here as being: for machínes to contribute to
an environmental humanism, they must have a 'natural dialogue' with a human designer,
natural because they need his metaphors and natural because they need his ideas
unmutated. The dialogue, in turn, must reside within a computer-aided Cnot computerised)
system. The system, consequently, must Ínclude a /resident processar', some real-world
sensors and effectors, and an intelligence.

84 Computational Design Thinking 85 Towardsa HumanismThrough Machines

r

A New Agenda for Computer-Aided Design ambíguíty and discontinuity in shape interpretation, instabílity in the rules for carrying
out shape computations, and nonmonotinicity in criticai reasoning to determine whether
WiLliam]MitcheLl or not a design proposal is complete and satisfactory. My examples will be taken from the
sphere of architecture, but l shall provide a general theoretical treatment that applies to
In this extract from the introduction of the book The Electronic Design Studio many other áreas of design as well.
* * *
(1990), William IVlitchell establishes a succession from earlier concepts of
computational design - the primary break focusing on the moment at which Formalising Design Rules
information arises within a computational process. Historically, processes had The earliest idea about computational treatment of design rules was that they could be
been executed to exact a solution from a spedfíc and complex set of initial expressed as procedures which would accept design requirements as input and produce
inputs. In this instance, the information exists a priori. As Mitchell points out, appropriate shape Ínformation as output. Such g procedure might, for example, accept
this works in a disjunctive manner with a design paradigm which uses search a list of rooms together with área and adjacency requirements and produce a plan that
and iteration to establish specificity, rather than preset determinism. Vitality in satisfied those requirements. This was consistent with the dogma of early Modernism
computation depends upon malleable functions and operations which deduce that architectural design was a matter of satisfying an established set of requirements
information rather than simply dispense it. Mitchell cites this as a criticai as closely and effíciently as possible, and it was also consistent with early, procedurally
characteristicforanycomputational procedure.This also refíects into Mitchell/s oriented approaches to programming in languages like Fortran.
consideration of the means for terminating a computational procedure. Much useful research was done wlthin this paradigm/ and sometimes this resulted in
Mitchell addresses the contrast between a purely analytical approach and the programs that actually worked. In the mid-1970s, for example, Philip Steadman, Robin
more circuitous nature which underlies architectural design. With either case, LÍggett and l published an effícient, rigorous procedure for producing small rectangular
the procedures of a computational process/ the design system, respond to the floor plans that satisfíed adjacency, área and dimensional constraints, and that minimised
shifting, evolving, contradicting and coinciding nature of architecture, the total fíoor área CMitchell, Steadman, and Liggett1976; Steadman 1983).
designedsystem, as dynamic in shape, material and function. A fundamental Hmitation of this approach, however, is that it requires design rules to
be expressed in a very cumbersome and artifícial way ~ strictly in terms of the constructs
Design is the computation of shape information that is needed to guide fabrication provided by procedural programming languages. A dever programmer can certainly do
or construction of an artefact. This information normally specifies artefact topology this, but the archítectural content of the resulting code is very diffícult to comprehend/
Cconnections of vértices, edges, surfaces and closed volumesX dimensions, angles and and this makes it diffícult to subject the mie system to criticai scrutiny. Furthermore, the
tolerances on dimensions and angles. There may also be assodations of symbols wíth rules are inextricably intertwined with information that specifies a strategy for applying
subshapes to specify material and other such properties. them, só Ít Is usually very difficult to Ísolate and modify them.
The process of design takes different forms in different contexts, but the most usual But human designers learn. They see the work of others, become sensitive to new
computational operatlons are transformations Çunary operations) and combinations (binary issues, become aware of new possibilities, respond to critidsm and constantly modify
operations) of shapes in a two-dimensional drawing or a three-dimensional geometric model. An the mies that they apply. We usually call this stylistic evolution. It may take the form of
initial vocabulary ofshapes, togetherwith a repertoire ofshape transformation and combination sweeping stylistic change, or that of refínement and elaboration of an established style.
operators, establishes the shape álgebra within which the computation takes place. In any case it is an essential component of aeative design, and we must provide for it in
The computation terminates successfully when it can be shown that certain CAD systems that are seriously intended to support creative design. Further evidence for
predicates are satisfied by a shape produced by recursively applying the transformation this need is provided by the experience of those who have attempted to implement mle-
and combination operators to the inítial vocabulary. These predicates are usually stated in based design systems, and have typically found that they must spend a great deal of time
symbolic (verbal or numerical) form. Thus determination ofwhether a predicate is satisfied tinkering with the mies to get them to produce the right sorts of results.
usually involves producing a numerical cr verbal Ínterpretation of a drawing/ then deriving Adoption of a more modular and declarative style of programming is an important
inferences from thls Ínterpretatíon by applying rules orformulae. step in the right direction. If rules are expressed as productions, for example, they are
This definition may seem provocatively reductionist and to leave little room for decoupled from contrai information, and it becomes convenient to add rules/ delete
creativity. But l shalt argue, in this paper, that taking a computational view of design can rules and modify rules. It is possible to adapt and tune a rule system in incrementai,
reveal predsely where creativity enters and why we intuitively take it to be characteristíc experimental fashion. Some sort of general inference engine can be used to apply the rule
of ali but the most trivial design processes. In particular, l shall focus on the roles of system in whateverform it currently exists.

86 Computational Design Thinking 8ï A New Agenda for Compuíer-Aided Design

^
There are now many waysto set up production systemsthat encode design ruies. You can
build them in LÍsp, or you can take advantage of Prolog/s high-level fadlities for processing
rules expressed in the formai of first-order logic. Numerous shells for knowledge-based
systems providefacilities for expression of rules in if... then formal Ashape grammar is a!so
a production system, with the additional advantage that it expresses rules directly in terms
of shapes rather than in terms of some symboiic calculus. The following rule, for example,
shows (opposite) one thing you can do with a square - one possible compositional move.
By sketching you can rapidly discover other things to do with squares - some of them
interesting and some ofthem not. By remembering the interesting ones you can establish a
grammarfor composing squares, then you can go on to explore the language specified by
thatgrammar. Ifyouthinkof some more interestingthingsyoucan add corresponding mies
tothegrammar, andthusstructure a newten-ainfordesign exploration. (See KnightC1986)
for a detailed discussion of the way that changing the rules changes a designeis output.)
Shape rules can be thought of as stereotyped responses to stereotyped situations,
and therefore as constraints on the free play of the imagination. This was certainiy
the position of the Romantics, who opposed the hegemony of classical rules. But the
imagination needs something to play with. Though the consequence of applying a given
rule to a given shape is plain to see, the consequences of recursivety applying a rule
system to that shape can be very surprising. The mies of a grammar are not limiting
prescriptions, but tools for constmcting a path from the known to the unknown - tools
that can be changed if they do not seem to get you to the right place.
What sorts of shape rules should a designer use to structure paths into the unknown?
Useful and interesting sets of shape rules are, l think, those that yield a lot for a little.
Any set of designs can be generated in an uninteresting way by a grammar that hás a
ru!e to produce each of the designs in the sei, but it is no advantage to a designer to
know this sort of grammar. It /s advantageous, however, to know a concise grammar, with
few but powerful rules/ that specifíes an extensive and interesting set of designs. It hás
been shown that such concise yet powerful grammars cgn be provided for interesting and
Ímportant bodies ofarchitectural design work CStiny and Mitchell 1978; Mitcheil 1989).
Graphically expressed shape rules are much easier for a designerto understand and
critídsethan, say, pages of Prolog, só there is good reason to suggestthatshells for building
knowledge-based design systems should take the form of shape grammar interpreters
that can be programmed graphically, and that aliow quick and easy modífícation of rules.
Some interesting prototype systems of this sort have been developed CMitchell, Liggett,
andTan 1989; Nagakura 1989;Tan 1989).
ReprintedbypemüssionofIvüTPress. ©MTTPress.
The interface details of systems for building and applying seis of design rules are
less important, however, than the general point that we should focus on the expression
of design mies in declarative, modular, easily understood and easiiy modifiable formai.
In other words/ we should think of design systems as open, fíexible, constantly evolving
knowledge-capture devices rather than static collections of familiar tools and dispensers
of established wisdom. When we can do this l think we will see the emergence of design
systemsthatdonotjustmechanicallyassemblebanalíties/ butthat have reai styleandflair.

88 Computationat Design Thinking 89 A New Agenda for Computer-Aided Design

some way of knowing which parts of the indefinitely extensive entailment of a design
Knowing When a Design is Finished are worth attention and which are not, but precisely how the critic knows this is a very
There is a famous paradox in Plato/s Meno which seems to suggest the impossibiiity of diffícult Cand l think unsolved) problem.
truly creative thought. The protagonist asks Socrates how it is ever possible to attain It might seem that the issue could be settled by setting forth an a priori agenda of
new knowledge. If you know what you are looking for, he suggests, Ít will not really relevant issues for consideration. In well-defíned problems, such as chess problems and
be new to you when you find it. But if you don/t know what you are looking for/ if you theorem-proving problems, it is fairly straightforward to provide a complete spedfícation
can/t put forth somethlng definite as the subject of inquiry, you will have absolutely ofwhatis required. But in creative design, characteristically, this is impossible: issuesand
no way of ever knowing that you have found it, and the search will never terminate. solution criteria are evoked contextually as the design takes shape (Reitman 1965; Akin
Turing provided us with a modem resolution. We can establish a process for generating 1986). You do not necessarily know what you want until you see what you can have. Nor
Ínfomnation to consider (propositions, or numbers, or shapes), and we can specify a test do you know what to avoid until you hgve seen some failures. 5o the criticai agenda may
which determines whether a given piece of information is that for which we are looking evolve and change as the design possibilities structured by shape rules are made explidt,
-the solution to our problem. When we have generated some information that passes and may not, in fact, stabilise until the point of termination hás aimost been reached.
the test we have solved the problem, and the process terminates. More specifícally, in The most original and indsive critidsm is often só because it departs from established
the context of design, we can set forth the requlrements that a shape must satisfy, then agendas, establishes new Íssues for consideration/ and reshapes the discourse.
attempt to instantíate a shape that does só. The design is successfully completed when Furthermore, different presumptions about what a thing is meant to be, or needs
we can show that we have a shape which satisfies the requirements. We may never have to be, or just might be/ lead to different diagnoses of what it lacks. If you see that
seen this shape before we make it expficit for consideratlon: Ít may be genuinely new a shape might be a duck you can, from the rule Ali ducks have feathers, derive the
and surprising to us. criticism that the proposal lacks feathers, and suggest that feathers must be added to
In the early days of CAD this was seen as a matterofstating programmatic and technical yield a complete and satisfactory design. But if you are more interested in its potential
requirements, then applying analysis procedures to a database to determine whether or rabbithoodyou can,fromthe mie A/o/-oüib/'tóhovefeot/7er5 derive the criticai conclusion
not these requirements were satisfíed (Maver 1988). A fíoor plan layout program, for that the addition of feathers would noí be an appropriate way to complete the design.
example, might incorporate procedures to testa proposed plan for compliancewith given In general, a designer'5 sketch is an incomplete, ambiguous and possibly inconsistent
room adjacency and área requirements. This approach was a very direct extension of the depiction of a possible artefact. Thus it allows contradictory observations to be made
traditions of engineering analysis that had begun with GalEleo's analysis of designs for a and contradictory criticai inferences to be drawn from these. Design development is,
cantilever beam to determine whether they would prove strong enough. in large part, a matter of identifying and resolving these inconsistencies: hence the
Clearly this approach was, and continues to be, very useful. But analysis procedures frequent criticai comment that a design in its early stages is still 'unresolvecT. It is only
certaínly have notsuppianted human critics. Why? in the endgame of design, when the final details are being worked out within a well-
One reascm is that architectural desígns (as opposed to some sorts of engineeríng established overall framework, that there is value in the mechanisms for automatically
designs) often have extraordinarily complex entailments. Buildings perform subtle maintaining semantic Íntegrity of designs that have só frequently been proposed for
economic, social and cultural roles, and these can only be understood adequately by CAD systems (Borkin 1986; Eastman 1978; Kalay 1 989).
reasoning about them in the light of extensive economic, social and cultural knowledge. A final problem, and perhaps the most important one, is that analysis programs
It might be possible, in principie, to build artificial critics that could apply such broad- workwithin a framework ofstrictly monotonic reasoning.Theydraw criticai conclusions
basect knowledge to produce detailed, useful evaluations of design proposals, but no from observations of the design together with some fairiy stable, consistent body of
such systems have emerged só far. The evaluations and appraisals that CAD systems do facts and rules about the world. Adding facts and rules, or new observations about the
successfully produce are extremely narrowly focused. design, should never Ínvalidate criticai conctusions that have already been drawn. But
A second problem may be thought of as a special case of the weil-known /frame interesting criticai discussion, as heard at architectural juríes for example, is not really
problem' in artifícial Íntelligence CDennett 1984), How do you know what possible much like that. In fact, it exhibits the classic hallmarks of nonmonotonic discourse
consequences ofa design proposal should be given attention and explored?There are CGinsberg 1987). Conclusions are frequently modifjed or retracted when critics notice
Índefinitely many ways that a design does not fail, but most of these are Írrelevant things about the design that they had not noticed before, or evoke knowledge of
Why doesn/t a critic spend time carefully demonstrating, to everybody's satisfaction, the world that had not previously seemed relevante or revise their beliefs in the face
that painting the walls yellow won't cause the bathtub to explode? This is a perfectfy of argument. From the observation that a room hás no windows a critic might, for
valid line of inference, but not one that is worthwhile to pursue. Clearly the critic hás example, condude that it would be intolerably stuffy. A second critic might challenge

90 ComputatíonalDesignThinking 91 A New Agenda for Computer-Aided Design

r
Maver and Han-y Wagter (eds), CAAD Futures 87, Elsevier (Amsterdam), 1988.
this conclusion by observing that the room could easily be served by air conditioning. A MitcheU,WiüiamJ, TAeLogic ofArcMecfure, MITPress (Cambridge, MA), 1989,
Miteheü,WütiamJ, Robin S Liggett, and Milton Tan, 'Top-DownKnowledge-BasedDesign', Tbe
third critic might respond that air conditioning would be too expensive, and só on. Ali
Electronic Design Studio, 1989.
these criticai conclusions hold in the absence of information to the contrary - but such MitcheU,WilliamJ, Philip Steadman and Robin S Liggett. 'Synthesis and Optimization of SmaU
informgtion might be on the tip of another critica tongue. RectangularFloorPlans', £'nvironmezifandPJanamgB3, no l, 1976, pp 37-70.
Nagakura, Takehiko, 'Shape Recogniüon and Transformation - A Script-Based Approach' , The
In sum, it is unrealisticto assumethat a definitive, consistent set ofdesign requirements
Electronic Design Studío, 1989,
can always be estabiished ahead of time in orderto provide an irondad testfora solution. Reittnan, Walter R, ' Creative Problem Solving: Notes from the Autobiography of a Rigue'; in Cogniïion
A creative design proposal may, in fact, provide a challenge to established beliefs and andïïiougTitJohnWüeyCNewYork), 1965,
Steadman, Philip, Archiíecíural Morphology, Hon (London), 1983.
criticai agendas. We should recognise that architectural designs are tested against a rich
Stiny, George, andWilliamJMitcheü, 'ThePalladianGrammar', Enviro^menfaj-idP^amung-^S, no l,
and complex interpretational discourse that develops in parallel with the processes of 1978, pp 5-18.
estabiishing design rules and making design possibilities explidt Stiny, George, andWiïliamJMitcheH, 'CountíngPaüadianPlans', Environment and. Planning B 5, no 2,
1978, pp 189-98.
Tan, Milton, 'Saying Wïiaí it Is byWIiatit IsLike', The Electronic Design Studio,
Conclusion
Ivan SutherlancTs idea of structured design representation in computer memory, the From Malcolm McCullough, William J Mitchell and Patrick Purcell, eds,
Galileantraditionofdesignvalidationbyanalysisforcompliancewithpredeflnedcriteria, The Electronic Desígn Stuáio: Arcïiitectural Education in the Computer
and faith in stable, universal design rules, provided the foundation on which computer- Era, 3400 word excerpt from pages 1-2,6-8,11-13,13-15.© 1990

aided architectural design was initially built. But we should not remain prisoners of Massachusetts Institute of Technology, by permíssion of The MIT Press. ©
MIT Press.
these ideas. Close consideration ofthe phenomenology of design exploration and the
epistemology of critidsm suggests that we must embrace the possibilities of designs
that have ambiguous and unstable structural descriptions, of constructive mie systems
that are provisional, fíuid and mutable as we discover what they can produce/ and of
criticai reasoning thgt is not bound by assumptions of monotonicity. These Íssues are
not, l suggest/ ones that arise under anomalous conditions that can safely be ignored
in mainstream, 'practical' CAD systems. On the contrary, their centrality is characterístic
of creative design processes.
l do not see the emergence of these complexities as cause for pessimism. The great
achievement of pioneering work in CAD hás been to construct a suffíciently rigorous and
comprehensive theoretical framework to allow dear identifícation of these Íssues and
appreciation of their importance. The challenge now is to build a new generation of CAD
systems that responds to them in sophisticated ways. The language games that architects
play are subtle, and require commensurately subtle instruments.

References
Alan, Omer, Psycïiology ofArchiíecíural Design, Pion (London), 1986,
Borkin, Harold, 'SpatialandNonspatial ConsistencymDesignSystems',£z;v3ronmenfanrfP^aímmgB 13,
1986,pp 207-22,
Dennett, Daniel, 'CognitiveWheels: The FrameProblemofAT, m Christopher Hookway (ed), Minds,
Macliínes and Evolution, Cambridge UniversítyPress (Cambridge), 1984.
Eastman, Charles M, 'The Representation ofDesignProblems and Maintenance oftheir Stmcture',
in Jean-Claude Latombe (ed), Aitificial Intelligence and. Pattem Recognition m Computer-Aided
Design, North-HoUand (Amsterdam), 1978.
Gmsberg, Mathew L (ed), Readings in Nonmonotonic Reasonmg, Morgan Kau&nann (Los Altos,
CA),1987,
Kalay,YehudaE, Modelmg-Objects andEnvironments,]oïmWÏie-y (Ne-v/^ïoTïi), 1989.
Knight, Terry Weissman, Transfonnatíons ofLanguages ofDesigns, PhD dissertation, Graduate School
ofArchitecture and Urban Planning, University of Califomia (LA), 1986.
Maver, Thomas W, 'Soflware Tools for the Technical Evaluation of Design Altematives', m Thomas W

93 A New Agenda for Computer-Aided Design

92 Computational Design Thinking
Algorithmk Form that of Íntuition and talent, where styfistíc ídeas are pervaded by an individual, a 'star', or
a aroup oftalented partners within the practice. In contrast, an algorithm is a procedure,
the result of which is not necessarily credited to its creator. Algorithms are understood as
KostasTerzidis
abstract and universal mathematical operations that can be applied to almost gny kind or
quantity of elements. For instance, an algorithm in computational geometry is not
Criticai to the understanding of computation as a mode of thinking is
about the person who invented Ít but rather about its effíciency, speed and generality.
its distinction from computerisation. In his writings, Kostas Terzidis hás
Consequently, the use of algorithms to address formal problems is regarded suspidously
elaborated on the expanse of digitally based processes, in the realms
some2 as an attempt to overlook human sensitivity and creativity and give credit
of practice and theory, describing the mechanisms, functionality and
instead to an anonymous, mechanistic and automated procedure.3
relations to the balance between intention and emergence. In this text,
While most algorithms are tailored to automate tedious manual methods, there is a
from the book Expressive Form (2003), he clarifies the distinction between
^ertain category of algorithms that are not aimed at predictable results. Their inductive
computerised and computational processes in terms of the relationship
trategy is to explore generative processes or to simulate complex phenomena. Such
between authorship/ abstraction and specifícity. Terzidis explains, in
Juctive algorithms can be regarded as extensíons to human thínkíng and therefore
computerised methods specificity exists as a precursor, where the action
>w oneto leap into áreas of unpredictable, unimagínable and often inconceivable
of design is a matter of fíltering, reorganising and recompiling information
ítential. For instance, artificial neural networks are systems of algorithms that simulate
to present different fields of outcomes. Computerisation mainly m*mics
human brairTs functions. They are being used in classificatíon, forecasting and
the actions of designing in rationalising the information of a predictable
llelling applícatíons. In making determínations, neural networks use several principies,
outcome. Computation, on the other hand, involves the action of deducing
liding gradient-based training, fuzzy logic, genetic algorithms and Bayesian methods.
spedfidty, that which cannot be captured given the initial dataset of
|t distinguishes these algorithmic processes from common algorithms is that their
properties and conditions. It is argued that this, in fact, does mime the
Iviourisoften non predictabfe andthatfrequentlythey produce pattems ofthought
logic of design, as a pursuit of evolving precision from initially abstract
lesults that amaze even their own creators. Similarly in design, shape grammars,
instructions. Terzidis poses a valuable point in describing computatíon as
Imatical models, topological properties, genetlc systems, mappings and morphisms
not vitally bound to the digital. This frees it from the whole of geometric
lorithmic processes aimed at exploring uncommon, unpredictable and uncharted
and mathematically generated styles and particular modes of software
Iproperties and behavíours."
engagement Cthe interface). Computation serves as the means to which an
|o opposing human activities that are central to algorithmic composition as a
explorative 'search' can be undertaken.
|>f thought are invention and discovery. Invention is defíned as the act of causing
jiing to exist by the use of ingenuity or imagination; it is an artificial human
An algorithm is a computational procedure for addressing a problem in a fínite numberj
1. In contrast, discovery is the act of encountering, for the fírst time, something
steps.lt involves deduction, induction, abstraction, generalisation and structured logicj
iady existed. Both invention and discovery are about the origin of ideas and their
is the systematic extraction of logical principies and the development of a generic soi^
ïe in the context of human understanding. These two intellectual mechanisms
plan. Algorithmic strategies utilise the search for repetitive pattems, universal pr^napj
|om a logic which tends to argue whether the existence of certain ideas or
interchangeable modules and inductive links. The intettectual power of an ali
|s is one of the following: either a human creation or simply a glimpse of an
in its ability to infer new knowledge and to extend certam li^tsof^e^^'l^|
'xisting universe regardless of the presence of humans. The most paradigmatic
While many algorithms have been invented and implemented forarchitectura11
|of this polemic is that of geometry itself: the existence of geometry can be
in space allocation and planning problems,1 /theirr implementation in aesthetK
|as either a descriptive revelation of properties, measurements and relatíonships
formal theoríes hás been/generaÍÍy, iimited. Most of the theories related to form^
|g forms or as an arbitrary, postuiate-based mental structure that exists only in
maïnlyto subjective interpretation and perception. In contrast, algorithmic log ici^vj
|n mind. Forinstance, Eudidean geometrywas developed originallyto measure
deterministic approach toform and its shaping forces; itsuggests rationality, con^j
íon the surface of earth and yet, in Eudidean geometry/ Platonic primitive
coherency/ organisation and systemisation. What makes algorithmic logicso^P^U^
^ch as squares/ círcles and triangles, do not exist per se in nature though they
for architects is that they have maintained an ethos of artistic senslbllity an |
approximations of natural objects. Likewise, architecture can be
playfulness in their practice. In contrast, because of its mecham6t'c niat^ag^
|s either a simulation of the laws and structure of nature or as a world offantasy
is perceived as a non human creation and therefore is consideredl^llstah^a^
lation.s |p g^y ^^ algorithms are encapsulations of processes or systems of
TraditionaHy, the dominant mode for discussing creativity in architecture hás

|orithmicForm
94 ComputationaL Design Thinking
•3-Ï

-.SM
r
M
• •^11
processes that allow one to leap and venture into the world of the unknown whether
natural or artificial. They are not the end product, but rather a vehicle for exploration.
the current discourse, that mouse-based manipulations of control points on nonuniform
rationa! B-spline CNURBS)-based surfaces are considered by some theorists to be acts of
computing.9 Whi!e the mathematical concept and software implementation of NURBS as
Computation is a term that differs from, but is often confused with, computerisation.
surfaces is a product of applied numerical computation, the rearrangement oftheir control
While-computation is the procedure of calculating/ ie, determining something by
points through commercial software is simpiy an affíne transformation, Íe, a translation.
mathematical or logical methods, computerisation is the act of entering, processing
An alternative choíce is being formulated that may escape these dialectically opposed
or-storing information in a computer or a computer system.6 Computerisation is about
strategies:algorithmicdesign.ltinvolvesthedesignationofsoftwareprogramstogenerate
amtomation, mechanisation, dígitisation and conversion. Generally, it involves the
space and form from the rule-based logic inherent in architectural programs, typologies,
digitisation'of entities or processes that are preconceived, predetermined and well
building code and language itseff. Instead of direct programming, the codifícation of
defíned. In contrast, computation is about the exploration of indeterminate/ vague,
design intention using scriptíng languages available in three-dimensional packages (ie,
undear and often ill-defined processes; because of Íts exploratory nature, computation
Maya Embedded Language (MEL), 3dMax5aipt and FormZ) can build consistency,
aims at emulating or extending the human intellect. It is about rationalisation, reasoning,
stmcture, coherency, traceability and intelligence into computerised three-dimensíonal
logic, algorithm,"deduction/ induction, extrapolation, exploration and estimation.h íts
form. By using scripting languages designers can go beyond the mouse, transcending the
manifoldimplications, Ít involves problem solving, mental structures, cognition, simulation
factory-set limitations of current three-dimensional software. Algorithmic design does
and rule-based Íntelligence, to name but a few.
not eradicate differences but incorporates both computational complexity and creative
Digitisation is the conversion of analog information into digital Ínformation.
use of computers. For archítects, algorithmic design enables the role of the designer to
Computerisation, by definition, Ínvolves digitisation. This is not necËSsarily the case with
shift from architecture programming to programming architecture, Rather than investing
computation. Because many mental processes can be analysed, codified^ systematised
in arrested confíicts, computational terms míght be better exploited by this aiternative
or even synthesised without the use of computers, computational methods do not have
choice. For the fírst time perhaps, architectural design might be alignecf with neither
to involve digitisation. Nonetheless, their implementation on a computer system allows
formalism nor rationalism but with Íntelligent form and traceable creativity.
explorations of complexities that extend the limits of human prediction. Formstance, even
There are still some misconceptions about the role of the computer in the process
though the works of Gaston Julia7 in the 1920s and, consequently, BenoTt Mandelbrot3 in
of design. Design, like many other mental processes, at the information-processing
the 1970s were conceived and expressed on paper, they would have never been visualised,
levei hás nothing specifícally neural about it. The functional equivalence between brains
understood and explored further without the use of computational graphics.
and computers does not imply any structural equivalence at an anatomical levei Ceg,
Computing is a term used to denote the act of making a mathematical calculation
equivalence of neurons with circuits). Theories of infonnation processes are not equivalent
ar a computation. Computing is often equated with computation since bothemploy the
to theories of neural or electronic mechanisms for information processing.10 Even though,
same methods. Essentially, both terms share the same meaning. Grammatically, the term
physically, computers may appear to be a set of mindiess connectíons, at the information
computation involves the suffíx y-tion/ that denotes a state, condítion o^quaijty of a
levei they are the materialisation of mathematical and syllogistic procedures.11
procedure. Similarly, the tem computing employs the suffíx^-Íng^hat Ímplies an acüon
The word /toolr is often used to describe the synergistic interaction of designers with
of implementing a procedure. While the two terms are linked through a state-actíon
computers. A tool is defíned as an instrument used in the performance of an operation.
relationship, the noun -computation- implíes a set of theories and methods, whereas the
The connotative notion of a tool Ímplíes control, power, dominance, skill and artistry.
noun /computing/ suggests an active investigation within the sphere of computation.
A pen, for instance, is a device that allows one to perform or facilitate the manual or
Inany case/ these twoterms are entirely different, distinguished and set apart from the
mechanical work of writing or drawing. The capabilities, potency and limitations of a
terms computerisation or digitisation.
tool are known or estimated in advance. This is not the case with computers performing
The dominant mode of utilising computers in architecture today is that
inductive atgorithmic computations. Neither is their capacity or potency understood,
computerisation; entities or processes that are already conceptualised in the designei
nor can their limitations be pre-estimated. Indeed, designers are frequently amazed by
mind are entered, manipulated or stored in a computer system. In contrast, computation
processes performed by algorithmic procedures, overwhich they have no control and of
orcomputing, as a computer-based design tool, is generally limited.The^problemwith
which they often have no prior knowledge.
this situation is that designers do not take advantage of the computational power of the
Since the mid-1970s, beginning with shape grammars and computational geometry
computer. Instead some venture into manipulations or criticisms of computer models asj
and continuing through topologicaf properties and morphism, designers and theorists
theywere products of computation. While research and development ofsoftware involves
have been concerned with the use of algorithms as a mechanism for exploring formal
extensive computational techniques, mouse-based manipulations of three-dimer
compositions. These theories have attempted either to automate and enhance existing
computer models are not necessarily acts of computation. For instance, Ít appears, from

97 Algorithmic Form
96 ComputationalDesignThinking
manual techniques or to explore new uncharted territories of formal behaviour. Varioi D/or/es,14 Peter Eisenman's concept of an architectural diagram as an
methods have been employed in the search for new forms: formal analysis involves analytical, generative or representatíonal device is directly linked to the
the investígation of the properties that describe an architectural subject. Composition_ 1^^ of human understanding and interpretation. Thls human-centric approach is
geometrical attributes ana morphological properties obeying Calilean, Newtonian Ïliirïf within the sphere of subjective phenomena and personal interpretations. Within
and, lately, molecular and organic principies are extracted from figurai appearances of ÍÏ.Í+ rM m, any lcl9ic that c*ea's wit^ ^e evaluation or production of form must be, by
an object. In contrast, structural analysis deals with the derivation of the motivations both understandable and open to Ínterpretation. The problem with this approach
and propensities which are Ímplicit wíthin form and which may be used to extract their 3titdoesnotallowthoughtstotranscendbeyondthesphereofhumanunderstandÍng.
generative processes. Morphísm employs algorithmic processes for the interconnection while Ít praises and celebrates the uniqueness and complexity of the human mind,
between seemingly disparate entíties and the evolution from one design to another. becomes resistant to theories that point out the potential limítations of the human
Unlíke computerisation or digitisation, the extraction ofalgorithmic processes is an act In contrast, algorithmic form is not about perception or interpretation but rather
of high-level abstraction. It is often equated with rationalism/ determinísm orformalísm. the process of exploration, codifícation and extension of the human mind. Both
but more importantly these resources are ultimately in the servíce of transcendency. the algorithmic Ínput and the computer's output are inseparable within a computational
Transcendency is the quality of lying beyond the ordinary range of perception. It is svstemofcomplementarysources. Inthissense, adiagram becomesthe embodiment ofa
the quality of being above and beyond in response to timelessness and spacelessness. obtainable through a logic of mutual contributions: that of the human mind and
AlgorithmÍc structures represent abstract patterns that are not necessarily associated wlth that ofthe machine's extendibllity.
experience or perception. Furthermore, the observable outputs of algorithms should not Simiiarly/ Eudidean geometrywasunderstood asan extension ofhuman perception.
be equated with the processes that created them. Marcos Novak makes a distinctíon The divinity of íts nature can be ultimately linked to its ability to infer abstract concepts
between topology and curved surfaces. Topology, he points out, 'means simply the study that appeal to the mind rather than the eye. LÍke religíon, it was the revelation of an
ofthoserelationsthatremain invariant undertransformationsand deformations.A notíon abstractsystemofrelationshipsthattranscendedaboveand beyond the material objects
of continuity is indeed Ímplied in this defínition, but the contínuity is abstract.'12 Símilarly, it represented. Similarly, algorithmic form is an extension of human understanding. The
in Architectomcs of Humanism Líonel March /seeks an arder beyond appearances' as mere exístence of the term 'unimaginable' can be linked to the abilíty of algorithms to
he attempts to /uncover the //many modes of numbering"' and looks for the "warring surpassthesphereofhuman contrai and prediction. Like meta-structures,15algorithmic
and opposing elements", which go to make an original microcosm echoing universal formsaremanifestationsofinductive processes thatdescribe, extend and oftensurpass
harmony/13 Algorithmic processes result from events that are often nelther observable nor the designeis intellect.
predictableandseemtobehighlyintuitive.Theseeventsaremadepossiblebyabstraction There is often confusion about ownership of algorithmic fomns. Intellectual property is
and ingenuity. For instance, the discovery (or invention) of 'hyperspace' resulted from defined as the ownership of ideas and contrai over the tangible or virtual representation
an algorithmic ÍnductÍve process of projections that map three-dímensional poínts Ínto of those ideas. Traditionally, deslgners maíntain full ínteflectual property rights over their
four-dimensional ones/ yet both the projectíons and the results are neither predictable designs or manifestations thereof, based on the assumption that they own and control their
nor observable. In this sense, algoríthmic processes become a vehide for exploration that ideas. This is not always the case with algorithmic forms. While the hints, clues or suggestions
extends beyond the límits of perception. for an algorithm may be Íntellectual property of the designer-programmer, the resuiting
When comparing contemporary prgctlsing architects such as Thom Mayne/ Frank tangible or virtual representations of those ideas are not necessarily under the control
Gehry and Peter Eisenman Ít is necessary to overlook many signifícant and distinguishing of their author. Algorithms employ induction, regression, randomness, recurslon, cellular
differences in order to Ídentify at least one common theme: the use of the computer as automata, probability, Markovchains orquantum computatíon, to namea few, the outcomes
an exploratory formal tool andthe increasing dependencyoftheirworkoncomputational of which are unknown, unpredictable and unimaginable. If there is an intellectual root to
methods. The most paradigmatic examples of the last 10 years Ínvest in computationally these processes it must be sought in a world that extends beyond human understanding.16
generated parts and diagrams. Through computation, architecture transcends itself Both the notions of /unknown/ and 'unimaginable' escape from human understanding since
beyond the common and predictable. In contrast, computerisation provokes Whorfían both negatetwo ofthe last resorts of human intellect, those of knowledge and imagination.
effects; through the use of commerdal applicatíons and the dependency on their design In fact, as Novak points out, whHe the clause 'if-then' is a syllogístic structure that leads on
possibilíties, the designer'5 work is at risk of being dictated by the language-tools they to new knowiedge, the ciause 'if-then-else' involves the alternative /else/ that may point to
use. By unknowlngly convertíng to the constraínts of a particular computer application/s theoppositeofknowledge,thatis,to/thatwhíchdoesnotfollowfrom its roots, or, indeed,
style, one runs the risk of being associated not with the cutting-edge research, but wíth that whose roots can no longer be traced, or have become irrelevant, or are unknown, or
a mannerism of /high-tech/ style. follow from principies outside prevíous understanding'.17

98 Computational Design Thintíng 99 Algorithmic Form

because it is his creation. Humans, on the other hand, can understand civilisalion, because
A paradigm shift is defíned as a gradual change m the collective way of thinkir
made it. 'The worid of civil society hás certainly been made by men, and its prmciples are
It is the change of basic assumptions, values, goals, beliefs, expectations, theories and to be found within the modification of our own human mind.

knowledge. It is about transfomnation, transcendence/ advancement, evolution and colloquial sense, computerisationrefersto the process offumishingwith a computer 01 a
computei-system.
transitíon. While paradigm shift Is closely related to sclentific advancements, its true
Julia, 'Memoire surl:ltérationdes fonctions rationneïles', Journ.al de Math. Pure etAppl, vol
effect is in the collective realisatíon that a new theory or model requires understandim 8,J918,pp47-245.
traditíonal concepts In new ways, rejects old assumptions and replaces them wlth new. For Benoïl B Mandelbrot, TAeFrac^Geomefryo/'Na;L're,\VHFreeman(NewYork), 1983.
See Dana Cuff, 'One Educator's Thoughts on Design Sofiware's Profound Effecfs on Design
T5 Kuhn,18 sdentifíc revolutions occur during those periods when at least two paradigms 'lïunking and Teaching', Arcïutectural Record vol9,2001, pp 200-06. Inthis article Cuffconsiders
coexist, one traditional and at least one new. The paradígms are Íncommensurable, as that computing is 'one of the most important transformations of the contemporary profession' and

are the concepts used to understand and explaln basíc facts and beliefs. The two live in ttiat today 'computing hás become a populist Bkiïl'.
Theworks ofHerbert Simon and Allen Newell in lhe 1960s and I970s are undoubtedly some of
different worlds. The movement from the old to a new paradigm is called a paradigm shift. the best examples ofüie study of artificial mteUigence. ANeweU, Uniííed. Theoríes ofCogniííon,
Traditionally, the dominant paradigm for díscussing and producing architecture hás Han/ard University Press (Cambridge, MA), 1990 and I-IA Simon, The Sciences oí 'the Artificial MIT
press (Cambndge, MA), 1981.
been that of human intuition and Íngenuíty. For the first time perhaps, a paradigm shift
H GLymii inAmmateFomi op cit,p 19, describes machine inteUigence 'asüiatofmmcüess connecfcLons'.
is being formulated that outweighs previous ones.19 Algorithmic design employs methods 12 Novak continues:
and devices that have no precedent. If architecture is to embark into the alien world of A cube is not less topological than a blob. However, when worlang algorithmicaHy, what remains
invariant is tíie algoriLhm, só that a new notion of topology, variable topology' is introduced. Wïiile
algorithmic form, íts design methods should also Íncorporate computational processes.
the variations in the space ofthe parameters and control strLictures that implement the algorithm
If there is a fonn beyond comprehension it will lie within the algorithmic domain. maybe continuous, the product ofthe algorithm may be to showtears and discontinuities and ever
Whíle human Íntuition and íngenuity may be the starting point, the computational and fracture into a cloud ofparticles or an explosion ofshards,
See Marcos Novak, Transarchitectures andHypersurfaces', m Arclntecture andScience, G Di
combinatória! capabilities of computers must also be integrated.
Cristma (ed), Wiley-Academy (London), 2001, pp 153-7.
13 See Lionel March, Architectonics ofHumanism: Essays on Number in Arclntecture, Academy
Notes
Editions (Chichester), 1998, p ix,
l The work in Üiis área is as old as computer-aided design. An early atfempt was MIT's BUILD
14 See Peler Eisemnan, Diagram Diaríes. Universe Publishing (NewYork), 1999, p 27,
system, whích would take a building program, indicalmg dimensions and connections for each
15 Theprefix meta-indicates one levei ofdescriptionhigher. If X is some concept. then meta-Xis data
space, The computer then arranged the spaces, thus solving the problem. This approach hás
about, or processes operating on, X. For example, meta-syntax is a synlax for specilymg syntax, meta-
since been used extensively for solving complex design problems thal are related to arranging
language is a language used to discuss language, meta-data is data about data, and meta-reasonmg is
parameters in optimum ways. These approacheB focus on the functionality of the design scheme
reasoning about reasoning (defíniüon taken from the Free On-lme Dicïionary of Compuíing).
and do not take into account aesthetic or artistic parameters. In áreas such as the design of
16 KarlPopper arguedthat the world as awhole consists ofthree interconnectedworlds. World One
compuler chips, nuclear plants or hospitais, automatic spatial aüocation plays a ver/ important role,
is the world ofphysical objects and their properties - willi their energies, forces anel motions; World
eventoday. SeeADietz, Dwelling Hoiise Constmction, MITPress (Cambridge, MA), 1974, and C
Twoisthe subjective world of sUites ofconsciousness, orofmenialslates-vi/ithintentions, feelings,
Yessios, BParent,WBrownand C Terzidis, 'CKAAD Expert; A Computer and Knowledge Aided
thoughts. dreams, memories and só on, in individual minds, World Three is the world of objective
ArchitecturalDesignExpert', mDesign TlieorySQ, NSF Grmtee Workshop on Design Theoiy and
contents ofthought, especiaüyofscientifíc and poetic thoughts andofworks ofart. WorldThree is d
Melhodology, Rensselaer Polytechnic Instilute (Troy, NY), 1988, pp 1-8.
hmnan product, a human creaüon, which creates in t\im theoretical systems with thelr ovm domains
2 GregLynnrevealsthat; 'because ofthe stigma and fear ofreleasingcontrol ofthe designprocess
ofautonomy. See KarlRPopper The Logic ol 'Scientific Díscoveï'y, Harper&Row [NewYork), 1968.
to software. few architects have altempted to use the computer as a schematic, organizmg and
17 See Marcos Novak, 'Alien Space: ïhe ShockoftheView'. article reproduced from Ari +
generative médium for design'. See G Lynn, Animate Form, Princeton Architectural Press (New
recAnotó97,supplementofCTRCA,vol90,pp 12-13.
York), 1999, p 19.
18 See Thomas SKnhn, The Structure of Scienïific Revoluïions, Lhird edition, University of Chicago
3 In response lo this discrepancy the ancient Greeks de^sed a 'fair' method of acknowledgement
Press (Chicago, IL), 1996.
of authorship. The Pythagorem theorem, the spiral ofArchimedes and Euclidean geometry ai-e
19 PeterEisenmanreferredtoUieideaofcinelectronicparadigmshiftinarchitecturem 1992. Hewrote:
attempts to give proper credit to the authors regardless ofthe status oftheir subjects as inventions
During the fifty years since the Second World War, a paradigm shift hás taken place Ihat should
or discoveries.
have profoundly affected architecture: this was the shift from the mechanical paradigm lo the
4 The Eerm algorithmic is often connected with complexity. Whiie the objective ar result of an aïgo"
electronic one. This change canbe simplyunderstoodby comparingthe impact ofthe role of
rití-rm may be complex, the strategy itself does not necessarily fotlow that complexity. For mstance,
the human SLibject on such primary modes of reproduction as the photograph and the fax; the
chãos is the study ofhow simple systems can generate complicafed behaviour.
photograph wilhm Üie mechanical paradigm, the fax within the electronic one.
5 Claude Perrault, the architect ofthe peristyle oflheLouvre, argued that architecture is afanlastic
PEisenman, '''Visions" Unfolding: Architecture in the Age ofElectronic Media', Domus, vol 734,
art of pure invention. He asserted that architecture reaUy exists in the mind of lhe designer ana that
1992, pp 17-24.
there is no connection to the natural world. In addition, architecture as an imagmative art obeys
its own rules which are internaï and personal to each designei-, and that is why most creators are
vaguelyawareoftherules ofnature anciyetproduce exceüent pieces of art. A similar point was From Kostas Terzidis, 'Algorithmic Form', Expressíve Form, Spon Press
alsoarguedbyGiovanniBattistaVico. Inhiswork The New Science (1744), Viço argued that one (Abingdon) - Taylor & Francis Group, 2003, pp 65-73. Reproduced by
can know only by imagining, The Iwisting of language and meaning can lead one to discover new permission of Kostas Terzidis and Taylor & Francis. © Taylor & Francis.
worlds offantasy. He argued that one can know only what one makes, Only God can understand

100 ComputationalDesignThinking 101 Atgorithmic Form

Architecture and Practical Design Computation much in their slim numbers as their being truly exceptional as deeply creative thlnkers. My
third premise asserts that with the extraordinary opportunities offered by the computer

Mark Burry to assist the architect in computation, we have entered a new woHd of opportunlty. Só,
in considering the topic 'computation and design thinking' who is our constituency?
MarkBurry/sworkindecipheringthecomplexgeometriesfortheconstruction Should future architects become more familiar with mathematics, more skilled In the
of GaudFs Sagrada Família hás established an ongoing legacy for the world of numbers? l am going to argue here that there is a place for the architect at
relevance of computation in the realm of design. In this text, he tracks the least within the 'computationally learnecf, and that this is a positíon of design strength
development of specifíc geometric and procedural methods driven by the jf not actual professional necessity. f shall do this by looking at íwo case studies. Both
criticai nature of connecting computational form with material form. Burry examples have practical exigencies as the prime motivator for using computation to assist
first describes this challenge in identifying how particular mathematical the resolution of desígn thinking and design makíng; that tricky issue of locating an
means for generating geometry may not align with rules which relate to effective representaticm of ideas on the path towards achieving successful built artefacts.
scalar issues of structure, materiality and assembly. In his second example, The fírst case study considers the problem of the flattened arch and the second looks at
Burry describes the lineage of processes developed for extrapolating forms GaudFs dedicatlon to second order geometry during hls last 12 years.
from Gaudfs underlying geometric principies. Through these deductions,
Burry projects the concept of agreed principies - computation being the The /rT Centred Arch
realm in which they are captured. While computation in Ítself implies Arches are at theír most effident when they accommodate the forces to which they are
operation, it is really the selection, alignment and coordination of the subjected towards a vertical gravitatíonal alignment. The catenary curve describes the
operators which allows for knowledge to be inherited into the process and force path within the uniformly loaded arch, essential knowledge emergíng during the
specifícity relevant to material and context to be realised. Burry identifíes Renaissance. The catenary curve echoes the parábola most closely visually Calthough
this as a capacity of the architect, while belng the casual observer, who is not mathematically), and is derived from a hangíng chain spanning between two poínts
still integral in the coordination of the 'observatíons'. whose self-weight acted cm by gravity is an analogue calculator providing criticai
information on potential performance. Wren/s triumph with the dome of St Paulrs with
Introduced to a sophisticated role for geometry withln the design process a decade or the scientifíc contribution of Robert Hooke follows on from Brunelleschirs largefy Íntuitive
morebeforecomputersbecamean optionwithinthestudio, l have had threedecades now discoveries when building the Florence Duomo dome. Both offer conspicuous evidence
to refíect cm the role of computation in design thinking. During thfs time computational of a shift to greater architectural sophistication through the use of design computation.
design hás transposed from analogue to digital practice. Taking /compute' as to estímate In considering appropriate geometries for the arch - espedally the bridge - the ellipse
or determine by arithmeticai or mathematica! reckoníng; to caículate, reckon, count^ l is a more favourable altemative to the Romana semicircular are, for Ít can be far wider
shall avoid the temptation to critique the philosophical limitations of such a defínition than ít is tall, and Ít more or less visuaíly if not functionally directs forces on to a vertical
and stick to this literal interpretation, because the issues raised for the education and trajectory at its two horizontal extremes (Figure 1). With the right proportions an ellipse
practice of an architect remaín complex enough without having to step towards the can approximatea catenaryand parábola far more closeíythan a semídrcular Roman arch.
esoteric. Before l leap in l should clarify what l mean by mathematics too, as l am relying How likely was Ít that the Renaissance masters were familiar with the mathematical
on the older defínltion of mathematics bemg the colíective nome forgeometry, arithmetic, equation for an ellípse and its practícal application?
ana certain physica! sciences (as astronomy ana optscs) involving geométrica! reasoníng2
which, l believe, is closer to the lay view of the work of the mathematician than the stricter x2 v2
modern usage definition ofitbeing an abstract sdence which investigates deductively the ^+z=l
a2 b2
condusíons impiicit in the elementary conceptions ofspatiaí ana numérica! relations3
Of course other premises further constraln my refíection. The first is while the
majority of architects are not mathematidans, they nevertheless compute within their With a > b > 0.
practice to a certain ievel. The second premise is that very few qualified mathematicians
have subsequently moved to professional roles as architectural practitioners. There have Most líkely the 'string loop with two naíls' description would have suffíciently fascinated
been a few exceptional individuais who were blessed with mastery of both disciplines their enquiring minds and would have been more accessíble to most architects of that
(Brunelleschi, Wren, da Vinci, Aiexander, products of a time), buttheir exceptíonality is as time. The two nails are set apart by a dístance /. A string whose length L is greater in

102 ComputationatOesignThinking 103 ArcMtecture and Practícal Design Computation

2, is made into a loop and fitted around the two nails positioned at p7 and p2
2). If an inscribing device is inserted into the loop at d, and pulled away from both
-il the string is taut, it can move around the two naiis along a path that ensures
theloop is kepttaut.Thetrajectoryforthe resulting inscription is elliptical. Logically
condltions apply; pi to p2 is fixed as distance /; pia + p2d is a constant.
^ingiooplengthL=pró+p2tí+/.
-p^jç method for constructing an ellípse is effective, haptic and correspondingly
wíthin the designers' instinct than its defíning maths, one might argue, but Ít is
accurate. Nor is it convenient for the stonemason working at full scale for an arch
metres wide. A different construction of the ellipse and alternative to an equation
is that of the two ardes method with radii equivalent to the ellipse^s major
and minor o and ó axes from which the geometry may be approximated. Even if an
ellipse couid be accurately deríved and laid out at full size using this method, it implies
unique templates for every piece of masonry from arch slde to arch centre - the left-
hand templates obviously being mirrored for the right side or vice versa. But what sheet
material existed then to template at this scale, using an approach that we can envisage
só easily for our time? This was presumably a root cause for the need to simplify the
problem of describing an ellipse using connected cotangent ares. In this way a 'three
centred arch' hás a fíatter are in the top of the ellipse brought round more steeply at
either side by an are with a tighter radíus, ÍntersectÍng at the lateral extreme by an

Figure l. Ferdmand Meldahl,

are with tighter radius still. For a more accurate result a /five centred arch/ could be
eUipticaI arch, Bygningen blev composedwherebyfive connected cotangentarcsform the haifellipse CFigureB). Where
tegnet og opfórt som grovsmedie
to position the ares is the challenge here, and there is a choice of orthodoxies. Some
(former naval forge), Copenhagen,
Denmark, 1861. ©MarkBurry, are better than others, and the test is making an exaggeratedly flat arch whereby the
least successful method reveals its relatíve defídendes.4 For the stonemason, working
with a reduced set of templates through relatively easíly inscribed arc curves is easier to

Figure 2. Mark Burry, constmcting

countenance than the gradated curve of the ellipse.
an eüipse using the 'string loop and It may seem that l am wandering more into an account of descriptive geometry and
two nails' method. © Mark Burry.
stereotomy than deslgn computation buí that is not my intentíon at ali - there is a thin
line between computing for a designed material outcome and the appiication of geometry
asa setoftransferable skills between masterand apprentice. Myargument isthat despite

Figure 3. MarkBuny, geometry

such orthodoxies for constructing the /five centred arch/ that emerge over time, even
ofthe fíve-centred arch (based within a relatively reduced set of accomplishments such as flattening an arch there is still
on diagram by Paul L Rosin and
opportunity for choice; and a rich one too. A good eye is guíding the best combination
MichaelLVPitleway), ©MarkBuny.
of parameters such that the optimum might emerge as much from human judgement as
<x2,y2)
from a set of agreed principies. But it is clear from the account of others that considerable
computational effortwasalso appiiedtoseeka mathematicallyperfected outcomeforthe
best arrangement of /n/ centred arches, as Rosin an.d, Pittewayrs account só painstakingly
reveals (referred to in previous endnote). Here the computation is generative in terms of
-k
a practícal effícacy rather than being restricted to a purely computational device.
1 As an architect l havethecapadtyto understandthe dilemmasposed bythefíattened
are jom s
arch and to Ínvolve myself in the computation leading to the best possible outcome. If l

104 Computationat Design Thinking

105 Architectureand PracticalDesign Computation
were to busy myself wíth someone else/s software to compute the most effícient archw;
using the least material, extraordinary leveis of computation would have been
but l could not c!aim the terrítory from a mathematical perspective, as \ am men
operative. The computer hás done the computing - how can l, as architect, make any
daim? l therefore wonder at which levei architects work knowledgeably as part of
computationally leamed, or falsely as the dilettante mathematician, for regardless oflevf
they are not working mathematically at ali. Even in the era of digital computing we
not have a failsafe that allows us to be blissfully ígnorant of the maths hidden in the black
box; vigílance is still requíred on the part of the designer, The stained-glass rose window
for the Passion Facade at Caudfs Sagrada Família Church/ for example/ is a giant ellipsç
measuring 4 by 2 metre5 and was completed in 2001. The masonry that accommodates
Figure 4,MarkEun-y,
the window was described using high-end aeronáutica) design sofíware. The metal frame hyperboloid ofrevolution
for the glass was templated using software dominatíng the architectural sector at that of one sheet. © Mark Burry,

time. )t was not produdng a 'true ellipse' we díscovered, when we carne to attach the
trame to the masonry only to fínd that despite being the correct height it was too wide
Figuie 5. Aïriom Gaudí,
forthe opening by several centímetres. Sagrada Famüia Church,
Barcelona, circa 1920.
Original photograpïi_oE
Computational Gaudï 1:10 scale model of Gaudi s
Gaudï moved away from free form towards a strict geometrical schema followíng clerestory window of the
Sagrada Família Church,
completion of the Casa Milà apartments (approximately 1905 to ^ 912). For some these
Barcelona, taken prior to the
apartments symbolise his creative zeníth but arguably they were also his professional models destrucüon dunng
nadir; a rapid descent from his status as conspicuous and successful architect fêted by tlieSpanishCivilWar
C1936-9'J.©MarkBuïry.
the city/s well-to-do and revered by the citizens of Barcelona where most of his work
was built.5 From '1914 until bis death in 1926 there are three broad characteristics that
distinguish him from ali that had gone before in a long and illustrious career. First/ he took
on no further secular commissions. Second, he devoted himself entirely to the completíon
of the Sagrada Família Church design/ which up to that point he had worked on for
31 years. The third characteristic of this period is his surprising migration from a free
form predilectíon to one of committed applícation of three second arder Cdoubly ruled)
surfaces/ ali formulaically described using relatively simple equations: the hyperbolic
paraboloidr the helicoid ana the hyperboloid of revolution of one sheet CFigure 4).G
At the beginning of this final period, the building was only a few percentage points
built, while the design for what remained to be completed was hardly much further
advanced. He needed to push the design way forward as the implications of his advancing
age became more apparent to him.
There are highly original computationalfy generative aspects to his decision
to apply second order geometry, and the aesthetic, philosophical and practical
consequences of choosing them have only become better known in recent years/ many
decades following his death. In the context of computational design thinking l am
describing the revelations that carne from unravelling his approach very succinctly, a
kind of reverse engineering motivated by the expressed wísh of the Sagrada Familiar
commissionmg body to continue the building faithfutly following the surviving

Architecture and Practical Design Computation

106 Computationat Design Thinking 10Ï
,ents of GaudFs "1:10 and 1:25 scale plaster models, his principal design too]
El rfí 55. To compute the best possible outcome for completing this project
í!S^,rt GaucTTs direct authorial vision we had to reveal his computationalIstrategy
the secrets contained within the models. In using these geometries -
l -+'hflmaticalsurfaces - was Gaudï acting both as geometer and architect?

, Design Computation^
medescribehisapproachtocomputational design.The 1:10funicular Changing)
Gaudfs ínteractive analogue parametric desígn/computer'for hís uncompleted
at Colónia Güell on Barcelona/s outskirts. The hanging model itself was a network
catenary systems. He adopted a similarly haptic approach to modelling the Sagrada
lília Church in the cíty itself, albeit usíng a different médium and approach: plasterof
r;<; Gaudi was working in a hands-on manner with his model-makers composing walls,
columns, domes, towers and vautted ceilings predominantly by intersecting hyperboloids
of revolutíon. These intersected conglomerates were decoratívely embeliished by
articulating the surface unions with triangular planes or hyperbolic paraboloids, the
directrices of the hyperbolic paraboloids being designed to coincide with selected
generatríces from the hyperboloids (Figure 6).
We know exactly how Caudï worked as there hás been an almost unintermpted
apprenticeship from the generations who assisted him then to the current model-
makers on síte today. Hís approach is subiímely convíncíng especially when ít is seen
in action, but in itself seems not to have been based on calculation beyond what was
at least an ineffable intuítion informed by actual mathematical insight.7 The approach
that he took for major wall and ceiling vault elements, for example, is a rich one
as there are nine parameters that govern the relationship between any two adjacent
surfaces and the character of the ÍntersectÍon: in other words there is infínite choice
with nine dimensions of possibility. In summary, the first three parameters are the x, y
and z coordinates for the hyperboloid centroid. A second cluster of three parameters
ïïçMre G^Antom Gaudi, Sagrada
comprises the x, y and z axes of rotation. Parameters 7 and 8 are respectively a and
Família Churcli mudei, Barcelom,
arca 1920. RestorecU: l Q scale model b, the major and minar axes to the hyperboloid collar ellipse (or circle when o = b),
srestory wmdow fonhe and the ninth is c, the asymptote determining the slope of the surface. Because GaudÍ
Sagrada Família Church, Barcelona,
<£J went to the trouble of first devising then applying this geometrical codex, we presume
to aid his successors in capturing his vision for the building in totality, the challenge
for us, his post mortem successors, hás always been to extractthe corresponding data
accurately, interpret it as second order geometry, and apply it such that the constructed
outcome responds unequivocally to his intentions, albeit só subtly expressed in his
models but precisely redacted for the purposes of building.
GaudÍ died at the age of 74 in a street acddent and he had probably held off fully
describing how his design approach would be taken up by hís successors, só this hás
had to be undertaken forensically. The fírst efforts to engage seriously with untangling
GaudFs geometries and applying them to the construction did not begin in earnest until
half a century after his death with the Spanish Civil War of 1936-9 further complicating

108 Computational Design Ttínking

109 Architecture and PracticalDesignComputation
the issue. The digital revolution was taking place at the same time as our research
into the surviving models, whích hás therefore taken place in a series of shifts in levei
of sophjstication, always u5Íng the best computational assistance avaílable, and noi-
necËSsarily coming from the architecture sector.

Design ComputationThrough Drawing

[f GaudFs composition made through his computational plaster modelling process was
levei 1 then Íts later transfer without hís direct Ínput can be seen as levei 2: there was no
possibilíty of simply extrapolating from where he left off because not only had he not
wrítten down hístheoryatall, he had neverapplied ruledsurfacesto masonryconstmction
in the way proposed by his scale models from which hís followers might have learned. This
is also to say that Caudí himself never saw any of this second order geometry applied
to the built fabric as his models for the Sagrada Família Church determined, despite his
earlier experiments with hyperbolic paraboloids at Colónia Güell chapei. Furthermore, if
(and it is quite a bíg Íf) his sketched geometries and other graphical musíngs contaíned
the secrets to his computational deslgn approach, they were destroyed when vandals
AKC 'iitïHïs ~H.a<

sacked his workshop based on site during the Civil War.

My first effort to undertake the necessary analysís of the survivíng model fragments
was through projective drawíng. l had not been suffíciently schooled in descriptive
geometry at architecture school to fínd this an easy challenge, nor was there anyone on
site who knew how to approach the problem, My two senior collËagues were both nearing
90 years in age, Senyors Bonet Garí and Puig Í Boada, and while they had followed GaudT/
coínciding theír youthful enthusiastic archítectural apprentíceship with frequent visíts
to the atelier cm site during his last years, if CaudÏ had hímself known how he would
approach the translatíon of ruled surfaces from scale model to a construction method, he
had not shared thís with his younger acolytes. It was therefore a matter of working with
fírst principies.
Essentially the problem was intersecting adjacent hyperbolíc surfaces and plotting
the resulting fourth arder curves of intersection. More crucially, it was ascertaining the
Jntersection between anythreeormore ofthese curves-thetnp/epo/nts-thatdrovethe
investigation, because only when ali the tríple points matched the data measured from
GaudFs model woufd we be able to claim a faíthful rendítion of the models to built work.
As CaudFs models were effectively 3D sketches, it was not unreasonable that a certain
amount of approximation had taken place, but for thís system to work at full scale Ít was ^ISPF.fibOLO.aS f»'-! f-iNíià: ÍALÏ HAUlt s ATFRALS
MFIT Ul;w R

essential that the various ímperfectíons were ironed out. í-ACWk NAiXEMENT i P^.510 de 2D 3t>' a 40 ?C m.

The skills of the geographer and surveyor were called upon; unsurprisingly, a late bisixï.t 3®' B^» ?3 : líSO F',^Si T ;0

Modem Movement education had provided me with few of these for mapping complex
surfaces. Once l had made the necessary conceptual leap and started reading the triple
poínts as the equívalent of geographical mountain peaks and the surface intersectíons Figure 7. MarkBurry, sections ofthe
as ridges, contours could be computed by referring simultaneously to three separate hyperboloíds projecLed graphically,
Sagrada Família Church, Barcelona,
orthographic projections each the slze of tablecloths. It seemed an odd departure to be
1980. ©MarkBurry.
using concurrently a 5cientifíc calculator [recently invented at that time) to fínd points in

Ill ArchitectureandPracdcatDesignComputation
110 CompLrtationalDesignThinking
space Jn order to draw them with my Rotring pen usíng sets of railway curves. This in the digital era was the leap in functionality provided by such high-end
1979, and the projective drawing approach, while less labour intensive than experiment? desígn software available then oniy on UNIX-based workstations. Amazingiy,
modelling with setting plastec was intensive ali the same. Once a combinatíon ofsurf^ _., nnint of view/ there is not a great deal more that can be done on our software
had been graphically tested painstakingly for fít a plaster model would then be made for l'. -,innn<;t two decades later. It seems that the cost of such sophistication, and its
verification purposes, eachtime suggestingfurtherdesign refínement.Thiswasa process ibjiity through being só different in stmcture and use for the design studio,
in which each Íteration took many months CFigure 7). rchitects the opportunity to work to this levei of digital design computation untíl
of cheap software for use cm everyday machines só many years later. This
Digital Incursion account for the unembarrassed heralding and sudden rapid awareness and take-up
With the arrival of accessible architectural CAD in the late 1980s, the cavalry had seemed later than the period In the early 1990s that l describe, although the actual
to arrive, but early CAD at best offered only latent computatíonal design potential of parametric design hás not become any easier in the intervening years. Indeed
Intersecting two surfaces in space could only be undertaken usíng a variation of the l/7-.ramptricism'9 could be argued as a phenomenon that hás burgeoned despite its innate
graphical method described above. Surely two surfaces passing through each other familiarity to a majority of designers. Is Ít a case of a tail wagging a dog? (Figure 8).
would be trimmed with a single clíck of the mouse one assumed - it was not só. It took Levei 5 of the quest was a reaiisation that we needed to move from 2D anaiysís
three years to get access to extraordinarily expensive aeronautical desígn software that of curves approximating a hyperbola to 3D analysis. From matching a mathematically
boasted this trimming facilíty - one that only became core to affordable design software derived hyperbola with those we had carefully traced from GaudFs surviving model
a decade later, or more. In this interim period a number of approaches were applied as we moved to matchíng points on the surface wíth a best-fit optimisation
a third levei to our investigation. The one pertinent to this essay was an effort to trace which brought home to me the difference between architects dabblíng in
representative hyperbolas extracted from the model and match them to curves calculated computation at this levei and the refinement that applied mathematicians could bring
by mathematical equation. Copies of plaster models of GaudFs surfaces were Ínterrogated to solution-seeking of this nature.
by cutting them normal to the hyperbolold centres along the x and y axes and tracing the
hyperbolas as 2D profíles on paper, and digitising them using a ruler and set square. The Best-fít Optimisation
Íntention then was to build 3D hyperboloids digitally derived from the traced hyperbolas, [n 1993 my architectural colleagues at Universitat Politécnica de Catalunya wrote a
and this was especially ambitious since nearly ali the hyperboloids of revolutlon that make program in BASIC to home in on best-fít surfaces. We measured 3D points from the
up the derestory window elements, for instance, are elliptical. surfaces for which we were seeking geometricaliy defined hyperboloids. Working with a
l discovered at that time that there was no native hyperbola curve function in any set of minima and máxima values, the program would iteratively run through potential
of the most widely used CAD packages, thus the only way to draw the mathematically solutions such that al! of the points in our sei were tested for confomnation to each
con-ect curve was to calculate the position of suffícient points and interpolate them hypothetical solution. We gradually reduced the máxima and minima until fínally there
with a fitted pofyline. This was a trial and error approach that seemed at odds with would beonlyoneresult-the/bestfit/. In some cases the programwould run forseveral
the efficiency espoused by the emerging software proselytisers whose software was days before ylelding a candidate. We experienced the frisson that comes from not knowing
designed to tackle the inefficiencíes of manual draftíng. Looking under the hood when to give up/ as, for instance, when the máxima and/or the mínima were set too tight
of various architectural software packages l dlscovered the world of the macro and to the extent that there might not be a result at ali and we would be waíting for days in
scripting (Levei 3). i was surprised by its potential to enrich design decision-making and vain - lest a result suddeniy emerge, as we experienced. l described this process to an
became engaged immediately. For the decade that followed l marvelled at the lack of engineer and programmer colleague, Peter Wood, and he introduced our team to not
general take-up ofsuch powerful design adjuncts, but l conceded thatwe had emerged only genetic algorithms, but bespoke UNIX-based tools with which he wrote a program
as a design profession displaced by several generations from the time when geometrical that he christened Xhyper, taking advantage of UNIX's user-friendly GUI windows - a
engagement wíth design h.ad far greater signifícance.8 novelty at that time. This was Levei 6 in our development.10 This program could home in
on the best-fit result from a gíven dataset comprised of measured points in a matter of
Parametric Design seconds. This meant that we could entirely avoid the informed trial and error methods
The measuring by hand method preceded easy access to computer-linked measuring that characterísed earlier efforts of leveis 1-5 described above.
devices such as surface scanners or point dígitisers, só it was slow and quite awkward. Agamthistaskwasfarhardertodo in theearly 1990sthan itwould be now. In 1992
In 1993 parametric design made its appearance in our digital toolbox via software called we had to measure ali 550 relevant points in space on the interior and exterior of the
CADD55 from Computervision™. Levei 4 of the quest for the ruled surfaces now fully surviving partsofthe 1:10 plastermodel ofthe clerestory. We built an apparatusto help

112 ComputationalDesignThinkmg 113 Architectureand Practicat Design Computation

gs accurate as possible, but within a few years medicai measuring technology
LÍ^ through fírst to the animation Índustry and then on to projects such as our own.
in LA were leading the vanguard in this regard. Our measured points
for cruciai and less crucial relevance. Triple points and points on the
loíd collar were the most important. Points on the curve of intersection carne
and points on the surface that were neither intersections nor on the collar were
less important. Gaudí's method of making the models was very accurate in
but various rasping marks on the surface suggested that when the outcome was
the geometries were honed into shape meetlng the needs of the eye but
yjsibly iosing their geometrical credentials. The interior and exterior points from
surfacewerecombined as a single.cfatfíle, and read intoXhyperfor processing.We
seeking concrete results for up to nine parameters, but in some cases the values
yf several parameters were known such as the x, y and z coordinates of the hyperboloid
that GaudÏ had governed using a proportional system.11
Depending on the number of unknown parameters and richness of the data we could
choose Ínteractively between the Simptex and Hitl OimbÍng optimisatlon algorithms.
General!y a was known through GaudFs proportional system só we were down to seeking
values for tive unknowns: rotation aboutx.yandz, and constants b and c.Xhyperallowed
us to constrain and relax parameters interactively, and finally generate a script that would
automatically build the hyperboloid in our chosen parametric solid modeller (Figures 9,10).
l am going to emphasise an aspect of this story to make my point about the nexus
between the 'architect as architect' and 'mathematician as mathematidan/ as one
condition/ and the architect as being computationally leamed as the other.
GaudFs approach to this surface modulation is extraordinarily subtle. The general
condition for his approach to composition at the Sagrada Família Church in his final
12 years was for each hyperbolic surface to intersect with two others. One component
surface hás only to tilt by a fraction of a degree for the character of the composition
to change markedly (Figure 11). Therefore despite Xhyper zooming in on a best-fit
solution confonning to our captured points, and a script that would recompose an entire
parametric assembly such as the nave derestory window in less than an hour, we had
to ovemde the constraints automatically generated by Xhyper 96 times until we had a
solution that visually matched GaudFs, even if apparently the fínal tweak had to conflict
somewhat with the data. But thís is the essentíaf point; GaudT did not have access to
FÍgure81MarkBurry'parametncallydnven digital these tools. In our quest to unravel hís geometries, which were slightly approximate due
^culptmgthrough Boolean subtraclion, Sagrada
Família Church, Barcelona, 1994. © MarkBm-ry.
to the modelling process itself, changes to the plaster in time and humidity effects, we
had inadvertently built him a design tool. For armed with our digital tools he could have
modelled his surfaces roughiy using clay, for instance, digitised this approximate model
with a scanner, and sought appropriate ruled surfaces uslng Xhyper. The fact that he did
not have this advantage makes his method no less presdent in many ways, and neatly
trames the tension between what designers might do for themselves üsing computation/
and whatthey might request of others more expertthan the necessarily Júnior status of
the architect as mathematical dilettante.

114 Computationat Design Thinking

115 Architectureand Pratíicat Design Computation
^«ïL.US^i

l^l.sCAlt^l

üfi<u9 |o*!f.i»^t»<eïï,i

SJfífipleK

Híll Oifolïi

Figures 9 and 10. Mark Burry Figure 11, Mark Buny,

(programming Peter Wood), overlay of digital and
Xliyper, Sagrada Família Church, physical datei, Sagrada
FHe View Barcelona, 1994. Xïiyperis a Família Church, Barcelona,
standalone software that takes 1994. Changingthe
^.b^iiï surface point measuremenls as parameters even by srnaH
input and calculates the besf- degrees hás profoimd
fít hyperboloid of revolution effects on the outcome: here
to suit. The software offers a the darker shaded surfaces
(0.4,5.0,0.051 choice between Simplex and HiR are seeking a match with
Climbing optimisation algorithms
72(1.385,4^0^: © Mark Burry.
Gaudí's original model (pale
grey). © Mark Burry.

t Simples

1506 z 10.00000 z

Error; 3.3@ne-i

116 Computaíional DesignThinking llï Architecture and Practical Design Computation

Beyond Design Computation
Notes , ^ ^ ^g Ox&rd Eag-feA DjcüOTary, second edition, 1989, CD-ROM.
My fírst example looked at a situation where the designer uses computation l _onÏ,^ hr. Üie collective name for geometry, arithmetic, and certain physical sciences (as
her role as a solution seeker, in this case for building elÍÍptical fíattenedlarches."fln .Y 2 !^nmv aíid optics) mvolving geometrical reasoning. In modern use applied, (a) in a stncl
as 'Jthe abstracl science which invesügates dedLictively the conclusions implicit in the
second example, computation is engaged as a design generator as in the case nf+hln e
sen ^^rnnceGtíons of spatial and numerical relations, and which inclLides as its mam divisions
Gaudí. One is a deceptively símple 20 profíle challenge; the other is a 3Dcomnn'^.'ater ele^ofLw.-'arittoielic, anel álgebra; and (b) in -d wider sense, só as to include those branches of

s^ategywith communícation being the key driver. The architect was hardlyrem'otï^ ?^li^Í or other researcti which consist in the application of this ribstract science to concrete
"When the word is used m its wider sense, the abstract science is distinguished as pure
the computation Ínvolved in either example. What is clear is that evena"great'detsi^
d l^a+n-.R and its concrete applications (for example, in astronomy, varions branches of
wíll not know what to ask for when collaborating with a mathematicían~if3thevuh^'gner ?L rci-TRthetheory of probabiliües) as applied ormixedmathematics.' 'Mathematics.: Def, l, Tlie

working knowledge of computatíon, and designing with the computer is notnecpZ^0 ^OxfordEnglishDiction-dry, second ediüon, 1989, CD-ROM.

computational design in thís regard.

^,1 T, Rnsin and Michael LV Pitteway, The Eüipse and the Five-Centred Arch', The Mathemaïical
Up until the end ofthe 19th century. descriptive geometry was an essential c Gazeíte voÍ85,no502 (Mar 2001),pp 13-24 PuJolishedbyThe MathematicalAssociation.
cegada Nonell, Elgran Gaudi. Ausa (Barcelona), 1989, pp 511-28.
ofan archítecturel educationfor most schools ofarchitecture.The arríval ofnewm^
MCBurry, Tt-ieExpiatoryChurcliofthe Sagrada Família, Pha.idonPress (London), 1993,Figs
andtëchniques followíng the Industrial Revolution, a century or more earlierJecTto
46-50,
gradual erosion of the ínevitability of masonry constmction for architects. 'Untiï Gaudi's education included descriptive geometry -- approximately one-sixth ofhis university
studies. His textbook was the French classic by CFA Leroy, which mcludes projects of
mtroduction,_of.tensl'estructu''es and poured comPressive materiais, computationTf
hyperboloids of revolution: CFA Leroy, TYaAé de géomélne descnptive: suivi de laméthociedes
statics and the intersection of volumes and surfaces through projection presented"a'tir és et de la théoiie dês engrenages cyïmdríques et coniques, avec une collecüon depures
delimítatíon of the architect-s computational demands: geometry, arithmet'icanduba^ 'composée de 69planches. editions du Maüet-Bachelier, Gendre et Successeur de Bachelier,

álgebra were sufficient. Once the age of steel and concrete carne to domïnatelmateZ editeurs (Paris), 1855.
g 'In architecture's nexl phase of development the ground-plan must disappear completely.
selections for major constmction efforts, the different demands for expert ca[culation7oaf The two-dimensional spatial composilion fixed m a ground-plan wiü be replaced by an exact

these new stmctures resulted in greater computational sophistication, beyond the skiÏls constructional calculation - a calculation by means of which the supportmg capacity is restncted
to the simples! but strongest supporting points. For this purpose Euclidean mathemritics wiH be
of most conventionafly trained architects.12 In many countries this signalledtíie arrivaToÍ
of no further use - but wilh the aid of calculalion that is non-Euclidean and takes into account the
the englneerwho would assert a new role as the designei problem 'solver. Perhaps"th^ four dimensions eveïythmg wíll be very easy.: Proposiüon 9: '1924 Theo van Doesburg: Towards
division of labour is as responsible for the separationof the designer from' matihem'atí^ a Plastic Archilecture'. Extracted ü-om: Ulnch Conrads, Programs and Mamíéstoes on SOïh-century
Arcliitectui-e, MIT Press (Cambndge, MA), 1970, p 79.
as the arrival of non-Eudídean geometry and the great mathematical theorems and
9 http://wiAw.patril<schumacher.Gom/Texts/Paramelricism%20-%20A%20New%20Global%20
abstractions, espedally during the 1 9th century? Style%20for%20Architecture%20and%20Urban%20Design.html
10 The programming and interface design described here was undertaken by PeterWood at Victoria
Asthe computer assists architects through making design computation só much more
Universily ofWeUington, New Zealand, where we were colleagues at the time.
accessible, this fadlity also acts as a major stimulus to look beyond the richness of creative 11 The current Architect Director for the Sagrada Familia Church, Drjordi Bonet. hás provided a full
and practical problem solving. This facilíty can be argued as potentially freeingdesig'ners accoLint ofGauclí's use ofproportion appliedto his design for the church: mJBonet, The Essential
Gaudí, Editorial Pòrtic (Barcelona), May 2001 (MC Burry, translated from the Catalan origmal
^p to r^engagewithmathematics at a more philosophical levei than we have been doing
L-Vltím Gaudj.)
fo^much ofthe last 100 years. Indeed, as my earller quote from Theo va n "Does b u7g"^ 12 Some countnes, mcludmg Spam, stíH have stmctural calculation as part ofthe architecfs core
1924 posits (see note Q\ this is not exactly a new call, but it is nevertheless a remjnder competencies,

that spatial designers are not uniquely posítioned Ín this regard, and that mathematidans
© 2011 John Wíley & Sons Ltd.
have been engaging in n-dimensíonal thinking for considerably longer. The educational
implicationsofwhat seems like a growlng distance between design and mathematics-have
always struck me^as beíng profound, especially when só few schools seem-game~enough
to rise to the challenge, yet the opportunities for computation enriching (digital) design
have never seemed greater than they were at the end of the 19th century.'Ïopïng~wrth
the practícalitíes of design resolution through computation might well act as the essential
precursor to thinkíng more deeply about mathematics and the opportunitíes Ít offers for
rethínking some fundamental aspects of design.

118 ComputationatDesignThinkmg
119 Architecture and Practical Design Computation
An Introduction to^reatiye^yolutionarySy<teni; tinely evolved by computers. This automatic generation of Ínnovation by
Ear^v/p Rvolutionary systems allows our designers to consider more solutions faster.
Peter] Bentleyand David W Come lc ^c+pms allow us to sidestep limitations of 'conventional wisdom' and 'design
Ihef:^n'' Creative evolutionary systems can even suggest entirely new methods and
In this text^from the book CreatÍve Evolutionary Systems (2002), .s that we can then exploit in our own designs.
BenÍ.ley-andD'""cl come def1'.nethe ""'P°"e"ts ofadesign proc'es's whicch p ThR listofapplícations ofthese approaches is long and increasíng at a ferocious rate.
1 confígured to function in a manner akin to evolution and natural 'gre still only taking the first steps along this road. We are still leaming how to
selection. Applying this in the context of creativity in des,gn7evol"utioan volutionary computation to provide us with this kind of assistance. There are as
serveLas an exploratory en9ine m Producing variety withm genotypic l.nw nroblems yet to overcome as there are debates about the topic. The chapters in this
parameters, and natural selection acts as a search for effective functionmc
book are our best answers só far.
ofphenotypic fesults'ln relaying the steps from genotype to-embryoge Thankfullywe have a very good mentor to inspire us and help us when we run into
and phenotype to evaluation, Bentley and Corne describe an~,tera3ti^ ries. Natural evolution is a very creative problem solver, and the solutions of nature
process for realising specifíc solutions from generalised information."As ever present to remind us Just what evolution is capable of. Can we make a machine
a necessary_capacíty, specifícity is relevant to the means and context of 3millimetre5 long that is capable offíying under its own power? What about giving Ít
evaluation. There exists an indirectness between the initial paVamete^s siaht, or the ability to keep Ítself functioning by converting chemicals into energy, or
and a specifíc outcome, but the expression of the phenotype "demanda even the ability to make copies of itself? We simply do not have the know-howto achieve
^ direct relevance to the measures by which it is being analysed~.'Wh^e these marvels. But this is what nature hás achieved in a creature as simple as a fly. Other
Bentley and Corne call these -transformations-, they are'alterations in"the examples abound. For example, would we have thought of tuming fíippers designed
phenotype_for means of evaluation which lead inevitably to thegenotype to propel físh through the water Ínto arms and highly dextrous hands? Would we have
becoming less effective in moving the evolution forward. Às a specifíc setof thought of using hair to make the horn of a rhinoceros? Despite the constraints nature Ís
procedures: evolutlonary algorithms (EAs) operate through recursÏvesteps faced with - only a small number of materiais available, the fact that life must grow from
of constant reconfíguration of initial generative parameters through means a single cell and maintain itseif - the a-eativity of nature is overwhelmlng. We do not have
ofevaluation.^ In a conceptual manner, this describes a functioning' design to wait for some science fiction scenario where aliens from other worlds give us superior
space, a combinatorial matrix of potential solutions developed technology.Thetechnologyofthe natural world is already ali around and within us-and
examinatíon and selection.
it is vastly superior to our technology.

Introduction
A Framework for Creative Evolution
Ln,m.any.ways'_uslng the word;creativíty/ in a fíeld of computer science opens up a From this investigation of creative evolutionary systems, we can construct a framework
3's box of controversy. Convincing people that our aeativity can'berenhancred that contains fíve elements Csee Figure 1):
using computersoftware is sometimes diffícult. Arguíng with peopfe that the technia
themselves might^be capable of creativity is often futile. ButfeearïanTncrea^n • An evoiutionary algorithm
of researchers prepared to try.
• A genetic representation
^Theseare the reseaf'chers who pioneer- creati^ evolutionary systems. Their programs • An embryogeny using components
enable artists to evolve stunníng pieces of art, or allow musiciansto'createinewl^uun'd's
• A phenotype representation
and new compositions. By guíding evolution as ít happens, the use~r7 are"ab7e'To • Fitness functionCs) and/or processing of user input
lxplore:new ideas as theyemer9e through the mechanisms of-evolutÍon."Many'fínd
thislMgh'yrewarding act'vlty - explor'ng the 5Pace of P°"íbleimagesiorlsculptureüs To summarise, a a-eative evolutionary system requires some kind of evolutionary algorithm
orcomposítíons, and being constantly surprised as possibílities are revealed to t hem. to generate new solutions. The algorithm modifles genotypes defined by the genetic
Othercreativ^evolutíonary systems take this approach one step further"'Guídance"is representatíon, which should be designed to minimise disruption caused by the genetic
proyided by automatic software routines thatjudge evolving solutíon"swÍthoutThe' need operators.An embryogeny Cor mapping process) should decodethe genotype and/ using some
l input.Designs are evolved from random blobs to fully functional forms. Novel kind of components, should construct the phenotype. The phenotype representation should
círcuits/ boat hulls, architectural forms/ aircraft manoeuvres, even chemicaTi5tiructurues be desígnecf such that it permits quick and effídent evaluation bythe fítness functionCs). It

120 Computatíonal Design Thinking

121 An Introduction to Creative Evotutionary Systems
evolutionary algorithm, the genetic representation, and, to some extent, the
will be generic and suitable for reuse for most problems without modification.
representai i o n and fítness functions will usually be specific to the current
of the system. The fitness functions may also be designed to process human
rather than actually evaluate the solutions directly. The following sections explore the
jg^ents of the framework for exploratlve evolutionary systems in more detail.

EvolutionaryAlgorithm
The evolutionary algorithm forms the core of any evolutionary system. ... [TJhere are
foLir main E^s in use today: the genetic algorithm/ genetic programming, evolutionary
strategies and evolutionary programming. Sadly, only the GA and CP are commonly used
for the exploration of creative solutions. The reason for thís can perhaps be found in the
way these algorithms work. The genetic algorithm malntains genotypes and phenotypes,
with a mapping between the two. ... [T]his distinction hás helped to encourage some
GA researchers to use component-based genotype representations that map on to the
representatíons, thus allowlng explorative evolution to begin. In the same way,
genetic programming also makes use of genotypes Cthis time with tree structures) that
are mapped onto phenotypes such as programs, images or drcuits. GP hás the advantage
thatítsgenetic representationrequ/restheuseofsmallercomponentsCinthefunction and
terminal sets), só ali applications of GP demonstrate the exploratlve power of evolution.
This may explain why the first notion of 'invention machine/ carne from John Koza, the
inventor of GP - his algorithm ensures that explorative evolution will always take place. In
contrast, algorithms such as evolutionary strategies and evolutionary programming make
no distinction between genotype and phenotype. By directly modifying the solution and
with no provision for mapping to new representations, these approaches make the use
of components to construct solutions diffícult to implement and hence almost unheard
of in this fíeld. Nevertheless, there is nothing inherent in any evolutionary algorithm that
prevents Íts use in a creative evolutionary system.
WithinanyevolutionaryalgorithmthereareotherÍssuesthatmustbetackled. Handling
muttiple objectives, multimodality, noise, premature convergence and fuzzy or changing
fítness functions must ali be considered. Solutions to ali of these problems, using ideas
such as Pareto optimality, region ídentifícation, speciation, variable or directed mutation
rates and steady-state GAs are now emergíng in evolutionary computation CBentley
1999; Parmee 1999; Vavak and Fogarty 1996). These issues, although important, are
not the most signífícant consideration for creative evolution. Indeed, even the choice
of evolutionary algorithm (or indeed any other search algorithm) is secondary to the
Figure l. The five elements ofthe framework
for creative evolutionary systems. Peter representations, for it is the representatíons that permit evolution to explore.
Bentley and David Come, Image from original
pablicatíon, Reprinted by permission of
Elsevier Books. © Elsevier Boolcs. Cenotype Representation
The genotype representation defines the search spaçe of the algorithm. A poor
representatlon may enumerate the space such that very dissimilar solutions are dose
to each other, making search for better solutions harder. For creative evolutionary

122 Computatíonal Design Thinking 123 An Introduction to Creative Evolutionary Systems

computation, where genes will represent (directly or indirectly^ a variable are widely used for exploratíve evolutionary systems because they
compm?;the 5earch space is tyPi""Y'°f vanableldmien^^^lZ ^,ber "t ^echanismforconstructingwholesolutionsfromcomponents.Threetypesof
design even harder (Gero and Kazakov 1996). .-..-—„.„ u,u. ^^ ^
lrDV /^nu areused today. The first and most common is externai, where a programmer
,ne[e ale ,a's°,other problems'Because of the use °f components to hb^Ï^softwaTe^that performs the mapping, and the process cannot be evolved - it
solutbns, the likelihood of epistas.s dramattaHy^nc^s^: Not'^"^o^^pre5ent ?l^ltó'the genotype (Bentley 1999). More recently, exp//c/t embryogenies have
TesCTtatí^"have this effert (e9-a one-to-one maPPÍ"9 "? ^axel'^^ed e nnnular. with every step of the growth process expiicitiy held as part of the
^^^plmand.zem ep'tas^ H°"^'^tZ;p^ ZT^:n 3eco^npandevolved (Bentley 1999). Examples include cellularencoding, used by Koza
^nkedto tteir companions for the solution to work as a whole.Aic^lre^Ton'^rently ge^mfortheevo!utionof^ circuits (Koza et al 1999); Lindenmayer systems,
between tecomponents, a melody relies on links betweenTOtes7a"hous7reli<lslnnne ;L"ks F ^ Coates (1997) for the evolution of architetíural forms; and shape grammars,
betweenwa"s'"program relies onllnks between comm^ds."Thes'eHnlk3gÏslacirmeo,'1n';n.ks ïus A hy/ Gero for the evolutïonoffíoor plans (Schnier and Gero 1996). Despíte the
rorrcspo?nggenes become ePistati""y ""ked, resultm^ pototial^^^^,^ !useuderable success of these embryogenies, they often requíre connplex additíons to
fo^volution: Wth polygeny só p.vaientlin'thes:pS^:^e^^ ^^ cons^representatíons and operators to allow evolution to work. The third type of
^^de^of.the_9enotype rePresentati0"' "^ ^^P^^g^et,^^^^ i"process, the impiídt embryogeny, hás shown the most exciting resultsand 9reatest
mmimisethedisruptionofinherítance. -,--.-..3 3-— ^idLors to
^+ËntialÍn ïecent work. Instead of evolving the mapping as a set of explicit steps in the
.practitlon_ers ofGphave lon9 been aware ofthese Problems- and many solutions , an implídt embryogeny uses a set of rules, typically encoded as binary stríngs
^^en^ Mod!fícatlons can be made to th'g^fcirepresenta^n^^^ in a GA genotype. For each solutíon, a /seed/ component is created, and then the rules
functionality and decrease dismption, for example, encapsulatiion"AD^,"ADI's.'1 are iteratively applíed. Over many Íterations, with mies activating and SUPPressin9 each
and.slon.(KI:>za et al 1999)' New genetic OPerators that enforce typingMp en"su'rc other, the growth, position and type of new components are built up, fínally resulting n
^ret'Ltrees^remtJhuffled to° drastica"y durin9 the Pr°duct,on"o? off:pringu( the development of a complete solutlon. This emergent growth process shows remarkable
Poli/ and Langdon 1999). ' ^ ' ~~-"""' "• u"^""y ^dyy,
fies of scaiability/ with the genotype describing solutions of Íncreasing complexity
..GAS-do,not/ec'u'rethe use oftree-str"rt"^d genotypes, só genetic representation- without any increase in the number of mies needed; the rule-directed growth process
based problems are often less prevalent. GAs ca'n be'used to 3evolve vïriaNe^ is simply aliowed to mn longer. This is in contrast to ali other approaches, which require
9!notypes.and structured 9enotypes- typ'""y wiih operators designedto" signifícantly larger genotypes to define the increased growth of more complex solutions
CTSS,Dv.eL°'1!y,at..p,o,'ntiof slmilanty between two Pare"t genotypes-(for~e;amolel (Berrtleyand Kumar1999).
Ber^ley andWakefíeld 1996, Harvey 1992^^^^ G^^i^Zl;^^^ The process of mapping from genotypes to phenotypes is clearly of importance to the
,°.ule^by±m,pt;on°fep'statlc links between 9enes have also beein"idemon"s"trZ investigation of explorative evolution. Issues ofscalability, evolvability and biasesÍnduced in
(Coldberg 199%. In addition, GAs do not suffer from the classicCP'probÍeml"orbl°oaT search have yet to be considered by researchers in any great depth. Increased understandmg
w^re9e."otypestend to increase'"slze' with redundant g<-"eticmaterl,a^clom1ng"e"vaer in this área would benefit both computer science and developmental biology.
9rc.T'n"'utlons']t 'sclear that the crcation of suitable genetic ^esenWo^ná
corresponding operators Is a considerable problem in its own right^ FutTthe^orZ'recenï Phenotype Representatlon
research mdicates that signifícant benefits may be gained from u^ing lesscompiexq'en'eti'c Once constructed bythe embryogeny, the resulting solution is defíned bythe phenotype
reGpr,esenta"CT.S.an_doperators'instead makln9 use of^bryogenies'ofgreate;complex~ity representation. Typically this representation is specifíc to the cun-ent application: if we are
and Kumar 1999^1. evolving circuits, the representation might define networks of connected components, if
we are evolving buildings, the representation might define exterior shapes and/or interior
Embryogeny walls, fíoors and staírs. An important criterion is evaluation. TypÍcaliy the phenotype
,A^embwny's a.speclal kindof maPPin9 Process from 9^°tyPe to phenotype. representation will be designed to allow direct evaluation by fítness functions without
Withintfw process^the genotype is now regaróed as a set of 'growing^stmction7t"a intermedíate transformations or calculations. A poor choice will detrimentaHy affect
re^plt,h.rtde_f''neshow the phenotype w'" develop' poly9eny : Phe.wtypïc "trate being processing times and solution accurades.
produced by nwltiple genes acting in combination - is common. Research ^thfe'7re7haas The distinction between genotype, embryogeny and phenotype representations is
rewa'ldsome,ofthe potent'a' ofthese advanced maPPin9 Pr°cesses;Advantage7in'c"lude often blurred in this field. Some CP practitioners regard ali three to be the same. Others,
L^durtlonofth^searchspace^etterenumerationofth"earchspace7te such as JakobFs work on evolving neutral networks (Jakobi 199G) or Taura's work on
more complex solutions and adaptabilíty (Bentley 1999). evolutíonary contiguration design (Taura and Nagasaka 1999), use dífferent and distinct

124 Computational Design Thinking 125 An Introduction to Creative Evolutionary Systems

representations for each stage. It is still unclear whether the component growing
of the embryogeny should use the same represcntatíon as the phenotype. Should
phenotype be represented by blocks, or should the blocks be merged into a single,
description of the solution? For example, should an evolved musical composition ^
represented by a series of notes or by a single, complex waveform in the phenotype? T^
answer is likely to depend on the fítness functions.

Fitness Functions and/or Processing of User Input

The fítness functions must provide an evaluation score for every solution. For creative
evolutíon computation, this is often a líttle harder. Typically the use of components
rather than a parameterised solutíon means that early results can be chaotic, to
the least. A desígn of a car may resemble a shoe; a melody may sounct like a burglgr ^04) (496 246) » •

alarm. Before evolution hás had time to improve these initially random solutions, &TH8K226132) * » Deriw: couipoïieats
^Sê:Í38)^i%)
can be nothlng at ali like the desired result. And yet the fitness functions must always Y58'330)(U02SO)
be able to provide a titness score that makes sense. The task is made even harder
when unknowns or approximations must be incorporated into the evaluation, or when
constraints ancf objectives are varied to aid exploration. Potential solutions include the
use of custom-desígned modular functions (Bentley 1996) and fuzzy logic (Parmee
1999) to cope with such problems.
Asdiscussedearlier, manycreativesystems usehuman Ínputtohelpguideevolution. Embryogeny

Artists can completely take over the role of a fítness functlon ÈSims 1991; Todd and
Latham 1999), and more recent work hás ínvestigated the use of these techniques
for evolving photorealistíc images of faces - potentially useful for the identification
of criminais (Hancock and Frowd 1999). These applicatíons raise numerous human-
computer interfadng issues. Will a creative system detrimentally affect the style of an
artist or the memory of a crime victim? These software tools have been shown to aid
Pmt and cutout
imaglnation and creativity, but what is the best method to let the user inform evolution
theshape
of hís/her preferences, and what is the best method for the computer to report the
structure and contents ofthespace being explored? Clearly, further research Ís required
to address these issues.

Example Creative Application

To illustrate the potentíal of creative evolutionary systems, we will now explore a símple
applicatíon - the generation of shapes that, when constmcted out of paper, fali as slowly
as possible to the ground. In thís case we wíll not use human mteraction; this will be the
second type of creative evolutionary system described earlier.
We can construct a simple creative evolutionary system, followlng the framework Figure 3. The paper fali application uses eight vértices as
provided above. A /simple GA/ can be used as the evolutionary algorithm (Coldberg 1989). its Gomponents. These are extracted'from the genotype
and transformed mto real papei- shapea by the use of a
The genotype representatíon can be a fíat chromosome of 1 6 bínary coded genes. A very
'connect the dots and ÍÏ11 the Rhape: embryogeny (and
basic mapping process or embryogeny can be used to derive eight [x, y) parameter values someone to prmt and cut out the shape). Peter Bentley and
David Come. Image from original publication. Reprmted
from each chromosomejoin each vector to Ítssuccessor byan edgejointhe last vector to
bypermission ofHsevier Books. © Elsevier Books.
the fírst with an edge, and fíll the resulting shape. An easy way to perfomn this process is

126 Computationat Design Thinking 12ï AnIntroductiontoCreativeEvoLutionarySystems

fit of this solution appears to be the way rotation stabilises the motion of the
Ia ^Lpn+ina Ít from slipping sideways in the air and plummeting to the ground at
11ap j/.,,c çnppd. Even with the very Íimited resources evolution was given, a 'creative'

|^tiontôtheProb!emwasfound"

Creative Evolutionary Systems to Creativity

of the introduction hás introduced the ideas behind creative evolutionary
Such systems are used to aid the creativity of users by helping them to explore
Í-ïaac, during evolution, or help show users new Ídeas and concepts by generating
new solutions to problems previously thought to be only solvable by creative

!5è^^=^^
^cam^eJeed.(Notsh™avsc^^^
from on^publ,cation"Raprinteïb7^^^^
A brief background ofthe field was given, showing how difficult applications such
programming, drcuit deslgn, and art and music composition triggered
development of a new approach. By using human Ínteraction we can enable the
of aesthetically pleasing and other diffícult-to-evaluate solutions. By employing
idge-lean component-based representations, we remove constraints to search in
representations and allow evolution to assembie new solutions. Using such methods
ito ex,eaj,te ^Ïlpl'nstructlon"utput by the system' Print the shaPes. ^d "se scissors we have turned evolution into an explorer of what is possible, instead of an optimiser of

^^í^^9=^^e^^s
tocut-°utthe shapes (see Figure 2)- The resultin9 Paper phenotypes"('-r'eP'reseTte'ribl;
whatisalreadythere. [...]

^essio^Eachphenotype can be a"°cated theirfitness ^°re bycalcula^gl7(ti^ References

Bentiey, PJi 'An Introduction to Evolutionary Design by Computers', m PJ Bentley (ed), Evoïntionai-y
^^^ng that evolution would attempt to 3—^eno^s^ DesigiibyComputers.MoïganKau.ím-dimÇSanVTan.cisco), 1999,pp 1-73.
Bentley, PJ, Genenc Evolutionai-y Design ofSoM Objecls Usinga Genetic Algoríthm, PhDThesis,
Division of Computing Contrai Systems, Department ofEngineering. University of Huddersfield,
.™Lapp'lcation'"ustrate5.the use of an extremeiy basic component-based
1996.
^weZrepresTtimAS,Fi9ure2 "lustrates- the ^Ponent^^ly^ Bentley, PJ and S Kumar, Three Ways to Grow Designs: A Comparison of Embjyogenies for an
Zti^(^pote De5Pitet^^enL^'^,^Z Ï Evolutionary Design Problem', in Genetic and.Evohition.arf Computation Conference (GECCO

t.hlulc°nstrained freedom of position of each vertex re^ ^ °thTveurfes"nZs '99), 14-17 July 1999, Orlando, Ronda, 1999, pp 35-43.
Bentley, PJ and JP Waicefaeld, 'Hierarchical Crossover m Genetic Algorithms' , in Proceedings oftíie Ist
^t.this ramPonent-based representation allows the deflnitíon"of"a^asïn^be"^ Onlme Workshop on Soft Compnting (WSC1), Nagoya Umversity, Japan, pp 37-42,1996.
d?rer;^hapes; aearlythis.is ^knowledge-lean representation'(no informa^n^o^t Coates, P 'Usmg Genetic Programmmg and L-Systems to Explore 3D Design Worlds', m RJunge (ed),

ïhïapesarebest is provided)-lt sho^ also beclearthatth^rep^n^1 aliZ CAADFutures '97. Kluwer Academic (Munich), 1997,
Gero, JS anel V Kazakov, 'An Exploration-Based Evolutionary Model of Generative Design Process'
evoiut'°",toexplore the solut]on space in an u"c°"strainedmanne'r a"n7th"attuheure ^ Microcomputers in Civil Engineermg, 11: 209-16, 1996.
many míllions of good and bad solutions for thls problem. Goldberg, DE, Genetic Algonílims in Search, Opíimizaíion and M-dchme Leausn.g, Addison-Wesley
(Readmg,MA), 1989.
,f <//pretoably: the_.use ofreal;l'fe tÊstin9 is time-consuming and laborious Cespecially
Goldberg, DE, 'The Race, the Hurdle, and the Sweet Spol: Lessons from Genetic Algonthms for the
l^makethemtíake °f performln9 the exPeriment-y°"r^lfinaeado"f"en^^n:g1^ Automaüon of Design Innovation and Creativity', m PJ Bentley (ed), Evolutionary Design by

^es^süjdent)' consequent^ fo^ the purposes cfthis illustration~population^zes Computers, MorganKaufmann (San Francisco), 1999,
Hancock, P and C Frowd, 'Evolutionai-y Generation of Faces', m PJ Bentley and DW Come (eds),
: -10 individuais were used in an evolutionary run of 10 Qewratío^
Proceeàmgs ofthe AÏSB'99 Syniposium on Creative Evolutionary Sysíems (CÊS), Society for the
Jhe results were "1terestin9- DesPte these excessively lowvalues, evolution was able Studyof Artificial IntelligenceandSimulationofBehaviour (AISB), 1999.
L°.mtel'?lfí-T mprovements on th^ Harvey. I. [The SAGA Cross: The Mechanics ofRecombinationfor Species vràhVariable-Length
Genotypes', in R Manner and B Manderick (eds). ParaUeS Problem Solvmg from Nature H, North-
ínlhenexperlments-tlmes takenforinit'a"y ^ndom shapes to falHSO IJHimetresav"a"rie"d
HoUand (Amsterdam), 1992, pp 269-78.
^mo.Lsecond5to^:8.5eco'1d5' Bythe 10th 9eneratïon^ ^s-hapes"took"ocnl^r°agleu Jakobr N, 'Hamessmg Morphogenesis', mProceedíiigs of tïie Iníematíonal Conference oflnformation

,morlth.an,two-seconds to fa"the same distance- Fi9ure 3 showsrtwo"o7the's"olutioyns Processingm CeR cind Tíssne, 1996.
Koza, JE, F Bennett, D André and MA Keane, Genefic Progrãimnmg III: Dai-winian ïnvention and Problem
m.^e^a',p°pul:ati°n\convergence has begun to o^"^withmost"shaperusínuglïhle Solvíng, MorganKaufmann (San Francisco), 1999.
same technique of having a smaller fíap that causes the shapes to"rotatea"sfcy"falí! The Page, J, R Poli and W Langdon, 'Smooth Uniform Crossover with Smooth Point Mutation in Genetic

128 ComputationalDesignThinfdng
129 An Introduction to Creative Evotutionary Systems
Programming: APrelimmary Study', in R Poli, PNordin, \'TO Langdon and T Fogarty (eds),
ProceedmgsofíheSecondEm-opean Workshop on Geneíic ProgT-ammmg-EuroGP'99, Constrained Generating Procedures
Gothenburg, 26-27 May, 1999, Sprmger Verlag (Berim), 1999.
Parrnee, I, 'Explormg the Design Potential of Evolutionary Sedrch, Exploration and Optimization'. ir
Bentley (eci), Evoluíionaij Design by Compaters. Morgan Kaufmann (San Francisco), 1999, john H Holland
Schnier, T, andJS Gero, 'Leammg Genetic Representations as anAltemative to Hand-Coded Shape
Grammars', in J Gero andFSudweeks (eds), Ai-ïifscjal InteRigence sn Design '96, lüuwer
is a process by which complex phenomena arise from the
(Dordrecht), 1996, pp 39-57.
Sims,K, 'AtíifícialEvolLitionforComputerGraphics', Computei'Graphics 25(4): 319-28, 1991. tion of simple conditions, where the possibilities of such complex
Taura. T and I Nagasaka^ 'Adaptive Growth Type Representaüon for 3D Configuration Design', m PJ cannot be identifíed within the individual initial assertions.
Bentley (ed) Firsí Special Issue on Evolutionary Design, Aitificial Intelligence for Engmeei-mg
This is inherent to any procedural process when compared with a discrete,
Design, Analysis, and. Manufactunng (AIEDAM) 13(3): 171-84, 1999.
Todd S, andWLatham, 'The Mutaüon and Growth ofArtby Computers', in PJ Bentley (ed). Evolutioin dírected one. John Holland, a scientist who established in the 1970s
Design by Computers, Morgan Kaufmann (San Francisco), 1999. the concepts and techniques for genetic algorithms, expands upon the
Vavak, Fand T Fogarty'Companson of Steady State and Generational Gás for Use m Nonstationary
computational mechanisms underlying emergent systems in this text. In
ETiwonments\mProceedmgsoftlieIEEE3rdLjteiTiationaíConferenceonEvolutíonary
ComputationICEC'96, Nagoyajapan, 1996. the book Emergence: From Chãos to Order (1998), Holland explains in this
chapter the properties of a /model/ in which systems containing emergent
From Peter Bentley and David Corne, 'An Introduction to Creative characteristics can be produced. The use of the term "modeF is unique as
Evolutionary Systems', in PJ Bentley and DW Corne, Creative Evolutionary
it describes the colfection of procedures and operators which interact to
Systems, Academia Press (San Diego), 2002. Reproduced by permission of
producea5efofpossibilities;the model is notdefined asthefinite product
Elsevier Books. © Elsevier Books.
of a select set of individual operations. Holland describes a key feature
of these procedures as the 'transition function^ which is a mapping of
the possíble states of a system that can arise from this function. Such an
operational framework is expressed further by Holland in the use of cellular
automata. Fundamentally, this asserts the capacity of simple agents (or
functions) to develop complexemergent behaviours but onlywhen engaged
into a combinatória! network - where the conditions of interaction are as
pivotal as the properties of the agents themselves.

The Greeks already knew the benefits of describing diverse objects (machines) in terms of
elementary mechanisms (the lever, the screw, and só on). This chapter develops a similar
setting for desaibing the diverse systems that exhibit emergence. The Greek approach
to machines also provides a helpful, intuitive guideline for developlng this setting. It
suggests looking at emergence in terms of mechanisms and procedures for combining
them. To make this work, we have to extend the ídea of 'mechanism'. The generalisation
used here comes dose to the physidsfs notion of an elementary particle (say a photon)
as a mechanism for mediatíng interactions (causing an electron to change its orbit around
an atom). We can use this extended notion of mechanism to provlde precise descriptions
of the elements, rules and Ínteractions (for example, agents) that generate emergent
phenomena. And we can compare quite different systems, increasíng our chances of
fínding common mies or laws that describe systems exhibiting emergence.
[...] We'11 be mimicking the path that scientists often folfow in movíng from
intuition to precision. The result is an explícit definition of a broad class of modeis 1/11 call
Constrained Cenerating Procedures (CCPs). The models that result are dynamic/ hence
procedures; the mechanisms that underpin the model generate\\\e. dynamic behaviour;
and the allowed interactions between the mechanisms constraín the possibilities, in

130 Computational Design Thinking 131 Constrained Generating Procedures

the way that the mies of a game constraín the possible board confígurations, /\\\
systems we've looked at só far can be described as CGPs. Potentially, any constrflinl!
generating procedure can exhibít emergent properties.
In outlíne, the path to constrained generatíng procedures involves the following

1 First, we'll translate the notion of mie (for instance, the mie forjumping in
checkers, or Hebb'5 rule) Ínto the notion of mechanssm. As with rules for
games, or laws for a physlcal system, mechanism5 will be used as the defíning
elements of the system.
Roughly, a mechanísm responds to actions Cor informatíon), processing that
Ínput to produce resultant actíons (or information) as output. The simple lever,
one of humankincTs earliest discoveries, provides an easy example (see Figure
l): when you pull down on one end of the lever Çthe input), you get a force at
the other end (the output) that is multiplied by the ratio of the two lever arms
Cthe processing). More complicated mechanisms may have several inputs and
may produce several different outputs - think of a mechanísm that sorts coins.
The word /mechanism/ hás variant meanings in common usage/ but 1/11 use a
defínÍtion and notation that restricts the word to a single, unambiguous meaning.

2 Next, we'11 defíne ways of tinking mechanisms to form networks; these networks
are the constrained generatlng procedures, the CCPs. Because most models
involve more than one kind of mechanism (as the dífferent elementary particles
in physícs), the setting must make evident how the actions of one mechanism
can affect others.
Withín this settíng, it is the ínteractíon of the mechanisms that generates
complex, organised behavíour. Typically, the mechanísms allowed wíll be few
in kind and simple to descríbe, Ënforcing the deletion of many details. Still,
the ínteractíons provide possibílities not easily anticipated by inspectíon of
the individual mechanisms. The complexity íncreases when multíple copies of
the basic mechanism5 are allowed, as ... seen in the case of ant colonies and
neural networks.
We'11 fírst look at CGPs in whích linkage captures the constraínts imposed
by the fíxecf landscape of a board game, or the geometry of a physical system.
Later we')l took Ínto a generalisation whereby links can be made and broken from
within the CGP, changmg the undedying geometry. This fadlítates models based Figure l. A simpie mechamsm - the leven Reprmted
on mobile agents. b7permissionsofBasicBooks,amemberof^
thelPerseus Books Group, andJoïmHoÜmd. ©
BrockmarL/HoÜancl.
3 Once the mechanisms are línked, we come to the counterpart of the game
tree - the set of possibílíties generated by the constrained Ínteractions of the
mechanísms. The examples already examined show that the tree of states is
a handy way of modelling the possíble courses of action - the strategles and
dynamics - of games. Now we want to extend that notion to systems in general.

132 Computational Design Thmking 133 Constrained Generating Procedures

To accomplish this objective, we have to define the state of a whole CCp
intef'msofthestatesofthecomponentmechanisms.We h ave to beabletodo have not yet saíd what outputs a mechanism produces in response to an Ínput
vyhatwe d Íd for the checkers player and neural nets: condense everythíng about It simplitiesthe notation, without changing the range of rule-govÊmed systems
.CG^that is relevant to future P°ssibilities into a single entity called Íts globa !hat can be modelled, ifwe simplytake the state ofthe mechanism to be its output, as if
state Then we go on to describe the legal ways of making transitions from^staTe observe everything or anything that takes place in the mechanism. Of course,
to state ... we'11 see that the possibilities for state changes are described Drecí^ ii<ïtic cases we may only be able to observe a part of the state, as when we can see
by a transítion function. ^hp handsof a watch but not the internai works. That complication is easily handled when
ivide for the Ínteraction of mechanisms as we/ll see in the next section. WÍth thls
4 Fmally, we want to provide a speclfíc procedure in CCPs for definlng hierarchies tíon the transition function f completely spedfies the mechanism's behaviour.
of subassemblies, using more complex mechanísms built up from the bask what is called an abuse of notation, the symbol f will be used as the mechanism's
mechanisms Advantages jn explanation will accme/ along the Unes noted by in addition to deslgnating its transition function.
Simon in 1969, and the resultlng organisatíon parallels the hierarchical nature of
most systems exhibiting emergence,
Interactions and Linkage
The major advantage here is again the one the Creeks perceived. It should Our basic concern In defining constrained generating procedures is to provide a
be possible to treat a whole CCP as a mechanism that can be used to build still setting in which we can study the complexities, and examples of emergence, that arise
more complex CGPs. This treatment parallels the use of subroutines/ themselves when rule-governed entities interact. l/ve adopted mechanisms, with their detining
composed of elementary ínstructíons, as elements of a computer program. transiíion functions, as the formal counterparts of rules. Emergence, in this setting/ is
By making such definition a part of the CGP apparatus, we can capture the closely connected to the generated complexlty that arises when multíple copies of like
hierarchies that are major features of systems exhíbiting emergence, like the mechanisms interact. Our concern now is to provide for the interaction of copies of a set
molecule/ organelle, cell, organ, organism, ... hiererchy in bíology. of elementary mechanisms.
A little more carefully, the definition of a constrained generating procedure starts with
Mechanisms
the selection of a set of mechanisms F, called primitives, from which everything else will
The term -mechanism-, as used here, is defíned by means of a transition function. ... For be constructed. The mechanisms in F are the counterparts of the elementary mechanisms
a mechanism, we must first define the state of the mechanism before we can define the the Greeks used to construct ali machines. When the CGP Is used to model a game, neurai
transítion function.
net, cellular automaton, or other system exhibiting emergence, the primitives are the
As a simple example of the state of a mechanísm, consider the state of a watch. The elements that are connected to form the model.
watch hás (a) a wound spríng that transmits power (b) through a ratchet that advances (c) to In this precise setting, one mechanism is connected to another when the state
produce the rotar/ motion of an Índicator ChancT). The state of the mechanism includes the sequence of the fírst mechanism determines a sequence of values for one of the second
tightness of the spring and the position ofthe ratchet, as well as the position of the Índicator. mechanisn-Ts inputs. Once F is selected, we connect copies of mechanisms in F to form
When we look at the watch as a mechanism, we see a succession of states as a network of interacting mechanisms, as when we connect neurons to form a neural
the hands move around the face of the watch. Intemally, this movement of hands is network. In short, we obtain a particular CCP when we select a set of primitives F and
caused by movement of the ratchets and gears, unwinding of the mainspring, and só then interconnectthem Csee Figure 2).
on - ali components of the state. This state trajectory, the sequence indicated by the * * *

movement of the hands, Is the dynamic of this mechanism. The watch is somewhat Once each mechanism in the CCP is assigned an índex, we can develop a kind
atypícal of mechanísms because Íts ínput is Íntermíttent. The trajectory ofstates proceeds of lattice to illustrate the interconnections and neighbourhoods in the CCP, a lattice
autonomously between wíndings ofthe mainspring/ whereas mechanisms moretypically reminiscent of the tree that showed how moves are ínterconnected in a game.
are under the contrai of a steady stream of inputs Csimilar to the succession of decisions Node / in the lattice corresponds to the mechanism with index ;. If mechanism / is
that determine the unfolding of a game). Still, the watch provides an example ofthe kmd connected to mechanism7/ then an arrow runs from ; to j in the lattice. For example,
of dynamic we can expect from mechanisms, and Íllustrates the way in which slmpler the mechanisms might be connected in a regular square array/ like a checkerboard.
mechanisms (the mainspring, ratchets, and só forth) can be combined to yield a more Then each node would have four neighbours to whjch it is connected, and the whole
complex mechanism. IVIuch as the combination of mies defines a game, combínations of array could be laid out as a square tiling pattern. However, any kind of connection
complex mechanisms buílt from simple ones have a major role in CCPs. pattern is allowable; ít need not be regular in any way, We can even allow the number

134 Computational Design Thinking 135 Constrained Generating Procedures

5 (copies of mechanisms) to become infinite, yielding an intinite array such as
kerboard extending indefinitely in two directions.

Automata as CCPs
Cellular
make an early test of the ability of CCPs to model complex systems by looking at
Wecan
automata. Cetlular automata are the brainchildren of two of the most renowned
imatical physidsts of the 20th century, Stanislaw Ulam and John von Neumann.
i/s Idea Csee the 1974 collection of hís papers) was to construct a mathematically
tined model of the physical universe within which one could build a wide range of
/«iai-hines/. This model physics would retain the essence of a real physics - Ít would have a
and a set of locally deflned laws (a transitíon functíon) that held at every point
Ín that geometry. Ulam saw that you could use a checkerboard geometry with an identical
sei of states assigned to each square Csee Figure 3 for a simple example, Conway/s
automaton). Ulam completed the construct by specifying a transition function that
determines thewaythestateschange overtime.Von Neumannthen usedthis constructto
design the self-reproducing machine ... Cellular automata have proved to be useful tools
for investigating complex systems, só they constitute an appropriate test for CGPs.
Each square, or ce//, in the cellular automaton, with Íts assoclated states and mies,
becomes a mechanism in the CCP framework. That is, the states of the cell become the
mechanism/s states, and the mies of the cellular automaton define the mechanism/s
transition function. In these terms, the cellular automaton is characterised as a CGP
employing copies of a ssngie mechanism, connected to form the regular square array
described in the previous section. By picking an appropriate pattern of states in this array,
we can provide a kind of active blueprint for any conceivable machine.

A Simple Cellular Automaton

One of the simplest cellular automata was designed and named by John Conway. It is
called ''L[fe/. Life, despite its símplicity, offers some fascínating examples of emergence. The
f ^\ f
automaton will be fully defined here, and we will discuss one of the examples of emergence
therein in some detail/ but the reader should see Gardner's 1983 book to gain an idea of the
extensive research that hás been devoted to this automaton. In CGP terms, Life is formed
from copies of a single mechanism that hasjust two states/ represented by 1 and 0. For
descriptive purposes, we can think of a cell as occupied by a partícle if the corresponding
mechanism is in state 1; otherwise Ít is empty. The lattice of connections is a square array,
as described earlier, but now ali eight squares that surround a given square are connected to
that square.That is/ each node hás eight immediate neighbours (see Figure 3].
The transition function is also easy to describe. If a cell is empty (state 0), and exactly
three of its eight immediate neighbours are occupied (state 1), then that cell becomes
Figure 2. Constructing CGPsïiom a set of primitive occupied on the next time-step (its state changes to state 1); otherwise, it stays empty. If
mechamsms, Reprmted by pemiissions ofBasic
Books, a memüer of fhe Perseus Books Group, and
a cell is occupied, and eithertwo orthree of its ímmediate eight neighboursareoccupled,
Joïm Holland. © Brockman/HoHand. the rest being empty, then on the next time-step the cell remains occupied; otherwise, it
becomes empty. That encompasses the full range of possibilities.

136 CompLrtationat Design Tfnnking

137 Constraíned Generating Procedures
a simple CCP would seem to offer little that is instructive conceming emergence,
thatis not thle case' T^*e st surPrise is t^iat you can embed a general-purpose Cfully
rammable) computer in this simple space, Specifícally/ you can design a pattem
cells in this space that interact só as to form a general-purpose computer.
nsequence, any process that can be modelled on a computer can be imitated in
>ics/ of Conway's cefluíar automaton. [ wí!f not attempt a description of the
ilicated 'general-purpose computer' pattern here, but ínstead describe a much
example of emergence in Conwa/s automaton.

Gliders
Conwa/s automaton can contain a simple, mobile, self-perpetuating pattern called a
íiríer. It consists of a pattern offíve occupied cells, surrounded byemptycells(see Figure
The transition function just described produces a sequence of changes in the pattern
over successive time-steps. Though fíve occupied cells are always involved, the pattern
changes shape in a regular way and Ít moves Cglides/) diagonally across the space. At
intervalsoffourtime-5teps,the pattern recurs, nowshifted one cell diagonally, saydown
and to the right, as In Figure 4. If you continue on in time, the pattern continues to go
\ J_ ^ JL x -L / through the same sequence of changes, gliding downward across the lattice, as long as it
Sf «t *-
does not encounter another pattern of occupled cells. The transition is só easy that you
-i

>i T^ T ti T K can try it on a piece of ruled paper.

? t>
Simple though the glider is, Ít tells us much about emergence. First of ali, the
x ï ^ l: ^ ï < glider is not a fixed set of particles bound together and moving on a trajectory through
Ml ?

/ T < T < -Y
\
the space. Rather, particles are contlnually being created and deleted to produce the
glider. This persistence of pattern, despite a continuai turnover of constituents, recalls
the persistence of the standing wave that forms in front of a rock in a rapidly fíowing
stream. The glider is more complicated than the standing wave because it changes form
as it moves, and it does not stay in one location. Still, the glider and its glide are a

! i— consequence of the simple laws of this universe.

The possibility of such a spatially coherent moving pattern is not something easily
• determined by direct Ínspection of the laws of Conwa/s universe. The possibility only
exists because ofthe strongly nonlínear Ínteracíions ofthe particles Cstates) in adjacent
celis. Even Íf we restrict attention to the five-by-tive array of cells sufticient to contain
a glider, along with the immediate surround of empty neighbours, no extant analytical
technique will predict the existence of a glider pattern. We can only discover the glider
by observatlon, watching the laws play themselves out in different contigurations. A five-
by-fíve array in LÍfe hás 225 ~ 32,000,000 distinct confígurations, só the task is not easy.
If we undertake an empirical exploration, we soon fínd out that many conflgurations

Figure 3, Conway's automaton. Reprinted by

are ephemeral, lasting only a few time-steps before dissolving to a set of empty cells. Other
permissions ofBasic Books, a member ofthe Perseus configurations remain fíxed in place, either retaining a fíxed form or 'bjinking/ through a
Books Group, andJohnHoüand. © Brockman/Holland.
recurrent sequence of changes. Still other confígurations mígrate over long periods, assuming
a multiplidty offorms, only to dissolve or wind up as a tixed confíguration. The glider fíts none
ofthese categories, because it persists índefínitelywhile moving through emptyspace.

138 Com putati anal Design Thinking 139 Constrained Generating Procedures
Ëven though we may infer the persistence of the glider from extended observation,
jnferences can be treacherous. For instance, there are patterns in Lifethat go through
long sequences of changes before ultimateiy dissolving. We can only be sure of
r/s persistence ífwe can prove thís property within a theory founded on the laws
'define Life. In this, Conway's artificial universe is not só different from the universe
•m vvhich we líve. Both require a combination of experiment and theory to discover and
explain regularities
Once we establish the glider's persistence, we can thmk about it as a component of
more complicated patterns. In fact, this step was taken in showing that a general-
computer could be embedded in Life. In an embedded computer, the gliders
serve as signalling devices, conveying informationfrom onepartofthe patterntoanother.
Additional research showed that there are confígurations that emit gliders/ and others
that respond to a collision with an incoming glider. These are the basíc elements for the
creation, transmission and reception of information. Once information can be transferred
from point to point/ it takes only some simple 'bit flipping' and storage schemes to
computation. Thus/ the existence ofthe glider encouraged a search for a way of
constructing a generaf-purpose computer within Conway's universe. The ensuing search,
and proofs, were relatively straightforward once conceived, but it is unlikely that they
would have been pursued with the necessary intensity without this encouragement.

Lessons about Emergence

What lessons about emergence can we draw from this example?
We see that:

1 Rules (transition functions) that are almost absurdly simple can generate
coherent, emergent phenomena.
2 Emergence centres on interactions that are more than a summing of independent
activities Cimposed by the nonlinear mies in this case).
3 Persistent emergent phenomena can serve as components of more complex
emergent phenomena.

Allthesepointshavebeen madeearlier, butheretheyappearinacontextsostarkthatnothing

lies hidden In the complexity, por is there any room for mysterious, unexplained activity.

From John Holland, 'Constrained Generating Procedures', Emergence:

From Chãos to Order, Perseus Books (Reading, MA), 1998, pp 125-42.
Copyright © 1999 John H Holland. Reprinted by permission of Basic Books,
a member of the Perseus Books Group.

Figure 4. Successive transitions ofthe güder slate

pattern in Conway's automaton. Reprinted by
permissions ofBasic Books, a memJoer ofthe Perseus
Books Group, andJohnHolland. © Brodanan^oüand.

140 Computational Design Thinking

141 Constrained Generating Procedures
ReaTVirtuality . concerned with the use of these procedural entities to model physical
^^Drecïselybecause these formal entities are defíned by what they do that is,
Manuel DeLanda
Íe5se!hPülsexhibitacertainbehaviour/ thatthey can be used to simulate the behaviour
ause
.UÏ^lJntitiesTheterm -virtual reality' nicely captures the idea that both functions
IL^ltext:..Manuel DeLanàa details the r^l^i°"ship of Rhv.. Í^i+hmscan be used to create simplified worlds in which the workings ofthe matéria
^m;fejnihe,cont?t.ofcomputationaldesig"^ l,j^v be studied. The term, of course, is already in use to ref e r to a narrow class of
estabHshed defjnition of -virtual realit/^D^Za 3fl^'l?nhntnhelon9-
^^rsimuïations/ those in which only the optical behaviour of physical objects is
anc";tera"y its syntax- distinguishi^^e ^J1^°^on- 1PL] e ^g behaviour of physical surfaces as they interact with light and the behaviour
:^^a';resultwrsus.computin9the''—l^^l^^:n^^ itinteracts with an observer. In other words, these simulations mimic not only
possible yet specific results. In addressmg'p^sïs"."th'e '^,Lfie'.dof

^^^d^^^^D^^^tÏS
: llg^ang of objects orthe casting of shadows but also the perspective Pr°Jection that
?eL. Ihïpns and shadows create in the plane of vision of a moving subject. Yet, despite
^means which define the subsequent:sïate^:^'^ocSt;^ 'that the tem hás already acquired currency in relation to these simulations, it
behwlour:. actions whichentitieT^°l"^v7hLaeLprteeru^^^^^^ E -i^nnroDriate to designate any kind of simulated world created by mathematicians
:^he[than propertíes- the ^c"material"de:c^,r^ ^^m
^p^aï^ ^^^^^^^^^l
l /^mniiter scientists, even if these worlds do not have a visual representation. In other
l?Lrrk the term is also appropriate to refer, for instance, to the worlds created by classical
^°fÏ.TateÏ1 entity ' rather th^ ^cesses need ^teZ^^'^^,^ ; when they modelled the behaviour of gravity almost four centuries ago.
tíï^lanü.ma9'vtuúe af^^^^ ^W^^M Ïis in this extended sense that the term 'virtual realityr will be used in this essay
^^^^n'^i°n°fmaterialpl:operfc-'D^n^'^^S^ Whatabout'real virtuality'? This term will be used to refer to the Ímmanent pattems of
?e^^n^ysicsï1 mathematics^""-t^^n^^Z<^ becoming that physicists and other scientists have discovered in the world, partly through
^utaÏ'!T>.aw'vach:He broac^^—^;Z^^,h thelr use of formal models and computer simulations, partly through causal inten/entions
l^lÏ^"tíLprocess's defined '" the ""^"ofinte^^'^^^3; performed in laboratories. We normally use the tem law' to refe rto these pattem5' but
^^^"!T:r^ert^be^li^iduz^^^Ïn it may be argued that this term was coined when scientists were still deeply religious and
specifíc mechanisms that represent specifíc physicaTlaw^ thought of themselves as revealing the secret mies that a legislating god had issued to
govern hls creatlon. Today/s scíentists are not nearly as devout as they used to be but the
word W hás survived their change of allegiances. For purely practical purposes, of course,
there is nothing wrong with keeping the word, but from a philosophical point of view its

=SSïiÍ^S=ïS
theological connotations are problematic. On the other hand, nothing would be achieved if
we merely switched labels, from lav/ to something else, as if merely changing the way we

Eï^^^^EEEEïS
s^^^^^
taik about immanent patterns made any difference. Talk, as everyone knows, is cheap, and
só is relabelling. Só we must start by analysing a concrete case of a natural law, the law of
gravity, and discover what in this case could properly be called 'real virtuality'.
In his analysis of the character of physical law, the late physicist Richard Feynman argues
that the law of gravity hás three completely different versions. There is the familiar one in terms
of forces and accelerations, the more recent one of using fields, and the least well-known

EEïS^^i^SES!
version in terms of singuSarítíes, such as the minimum or maximum values of some parameter.
As Feynman argues, the three different versions allow physiclsts to make the exact same
predictions só they cannot be told apart by their consequences. Feynman, like many other

HSë=VÏJii^SÏ
^s,tn^^u::d.on^,mathematlcalfL^tic;ns:^e:rai:^^^^ famous physicists before him, subscribes to a philosophy of physics referred to as 'positívism'.
A positivist believes in the mind-independent reality of objects and processes that are directly

s^sïsE^^^^^^:^
obsen/able by us, ali other entities being only usefui theoretical constructs.To put it differently,
Feynman believes that the task of physics is not to explain the Ínner workings of the world but
only to produce compact descriptions that are useful to make predictions and that íncrease

142 Computational Design Thinking

143 ReatVirtuality
the degree of contrai we have of processes in the laboratory. But since the three vprc,. .• the tendency to minimise the difference between kinetic and potential
the law of graviíy make the same predictions, it is useless to speculate which one nf7hÏ^S Df I ^ n othervvords, Ít was discovered that a simple singularity Ca minimum) structured ali
-reall/ explains gravitational processes. Are there really forces which act as causes to' '^^ :>ner?^cc hilitv spaces of classical processes, But this does not answer the question of what
thevelodtyof celestial bodies? Or does reality really contaín gravitational fíelds? Qr, eveln'^! !:he Ïri+ip^ areMore technically, what is the ontological status of these singular points?
strangely, is it ali a matterofsingularities? For Feynmanthereis no answertothesenn^;.110^ ;sical process is actually taking place one can discern spedfíc causal mechanisms
Moreover, the very question /what is a law?/ is not stríctly speaking answerable, since th. lw^icinciispecifíc effects, and these mechanisms vary from one type of process to another:
ímmanent pattems of becoming that laws refer to are not directly observable. This is \ l^m^anisms through which a ray of light finds the shortest path have nothing in
positívísts use the word /law/ they typically use it to refer to the functions or equatíons' ln.nn withthose through which fíooding water seeks the lowest point of altitude. But
express a law, equations being directly observable when written on a piece of paper. lco"u^nbritv Ítself is mechanism-índependent. Ifthe ontological status of causes may be
Realist philosophers - those who believe in the mind-independent reality of hydr ; ,^^arhiaLthat is, if causes must actually exist in arder to produce their effects, the
and oxy9en. Yiruses and bactéria or even electrons and protons, nane of which is directk ;"ofthe singu!arity is not that of an actual entity but of a virtual one. It is
observable - do not have to abíde by posítivist proscriptions. 5o when it comes to laws thev ri! "every bit as~real as are the causes and the effects: singularities deflne a real virtuality.
take the realíty of immanent patterns seriously, even if Ít means confronting the embarra '"The next milestone in this history carne in the iast decades of the 19th century with the
r ? off!red by themultiP'icity of versions of one and the same law. The first two versions work ofanother great mathematidan Henri Poincaré. As appropriate to the study of^Possibility
of the law of gravity offer no problem if we thínk that many physícal entities behave both like the breakthrough was linked to new resources that had become available for the
discrete partides (the kinds of entíties to which forces can be applied) as well as continuous 3rnetrical exploration ofspatial questions. A hundred years earlier performing this exploration
fíelds. In other words, the divergence in our models tracks an objective divergence in realit ïieant studying metric spaces, that is, spaces in which the notions of length/ área and volume
But what to make of the third version, what would singularitíes be? The simple answer is that are fundamental, like those of Euclidean geometry. A mathematical space is baslcally a set of
sjngularities form the structure ofa space of possíbilíties, and it is this structure that constitutes só basic concems for mathematicians are the means avaílable to con-ectly identify a
a real virtuality. To see what this dÊfínition Ímplies we need to explore, however briefly, the given point in a space. The metric solution to this is to give each point an /address', that is to
histor/ ofthis version of classícal mechanlcs, the so-called 'variationaF verslon. locate it relative to a set of fíxed coordinates or axes. One determines the distance that each
In one of its forms the varlational versíon is, Índeed, well known. In 1662, Pierre de point hás from these axes and the set of such rigid distances becomes its address But in the
Fermat proposed that light propagates between two poínts só as to minimise travei time. His 19th centur/, mathematicians invented other ways of achievlng this identifícation. One could,
basic ínsight can be explained this way: \f we knew the start and end points of a light ray, and for example, determine the rate at which the cuivature of a space changes at every polní, and
ifwe could form the set of ali posslble pathsjoining these two points Cstraight paths, crooked use this instantaneous rate of change to identify a point. Such a space ceases to be a set of
paths, wavy paths), we could find out which ofthese possíbilitíes is the onethat light actuallses X, Y, Z addresses and becomes a fieid of rapidities ana slownesses, the rapidlty or slowness
by selecting the one that takes the least amount of time. In the centuries that followed. with which curvature varies at each point4 Dífferential geometr/ uses this approach, as does
other least principies' were added to Fermafs Cleast actíon, least effort, least resistance, least the even less metric geometry called /topology/. Poincaré's genius was to bring topological
potentialenergy). But the real breakthrough wasthe development in the 18th centuryofa way resourcesto bear on the study ofthe possibility spaces defíned by differential functions.
to extend this Ínsight into the world of dífferential functions, the basic technology underlying Tapping Ínto these resources implies that the mathematical model of a physical process
the virtual realities Cthe formal models) created by clássica! physics. ThÍs was the catculus of can be glven a spatial form. How is this achieved? To create the original model one needs
variatíons created bythe mathematicían Leonard Euler. Before Euler, the main problem was to enumerate ali the retevant ways in which the process in question is free to change. For
to fínd a way to specífy the sei of possíble paths só that it was maxímally inclusive, that is, example, a given physícal process may be characterised bytwo changlng properties, such as
só that Ít contained ali possibilities. This was done by 'parameterising' the paths, that Is, by temperature and pressure. These are called its 'degrees of freedom'. Of course, the process
generating the paths through the variatíon of a single parameter.2 But there are many physical may also change in an infinlte number of irrelevant ways, the art of modelling being based
problems in which the possibilities cannot be parameterlsed by a discrete set of variables. in large part on the ability to tell the signifícant from the insignlfícant. Once a processas
Euler's methocf solves this problem by tapping .ínto the resources of the dlfferential calculus. degrees offreedom are discovered, the model can be given a spatíal form by assigning each
Wíthout goíng into technical details, these resources allow one to rigorously specify the space ofthem to a dimension of a topological space. In the physical process being modelled, any
of possibilities and to locate the minimum, maximum and inflection points (that is, ali the combination of values for temperature and pressure fully specifíes the state in which the
singularities) ofthe functíons thatjoin the start and end points.3 process finds itself at any moment of time. When these degrees of freedom become the
Bythe mid-19th century, ali the different processes studied by classical physics (óptica), dímensíons of a space, each combination of values becomes a point in that space. That is,
gravjtational/ mechanical) had been given a variational fonn and were unified under a single each polnt represents an instantaneous state for the process, the set of points as a whole

144 Computatíonat Design Thinkmg 145 ReatVirtuality

blc^m'^g! _ïa-ce_c?f posslblc states- This Possib'lity space is called -state space' or /, it may be objected that this argument applies only to mathematical
space/. As a physical process changes in time following a sequence of states.thP^tphase l ,Ai. h it hás little relevance to material reality. What would correspond, for example,
sequ,^sappea,r ,'n state 5pace as a serles_of Points'-that is' as a curve~o7a t'raTecctloe^poral 10 s.il,rÍ+ ps in the real world? To answer this question we must distinguish between
Ifthiswas ali that topologists could offer, thelr contribution would amounttoh,> ÏnP+ es that define a real entity - an atom, a molecule, a living cell - from the
discovered a way to visualise the behavíour of differentlal functions, or morep^^'n9 ^^tendendes ana capadties. The distinction between properties and capacitíes can
the solutions to those functions. This would be a valuable too) but not a"fund^Lylof lntity^gtódwithasimple example; a kitchen knife.The knife is characterised by certain
advance fui) ofontological consequences. But their real achlevement was tomo^Zn?' ^rtiessuchasitsweight, Íts length and whether it is sharp or blunt. Sharpness as a
mechanism-independent levei. In other words, topologists did not study thesina^0;^6 (pró 'is always actual: at any given instant, the knife either possesses this property or
structuring the possibility space of a process free to~change in its tempe^tuT^ {prí^c""not But the knife also hás capacities/ such as Íts capacity to cut things, that may
pressure, but of ali processes wíth two degrees of freedom, whatever these \iZ . flupr be actual if they do not happen to be exerdsed, that is, if the knife is never used.
changing were. And similarly for processes with any number of degrees'offree7om7hys.of 1^1 when the capacity does become actual ít is not as a state, like the state of being
!^ïlceÍwíth any number of dimensions Poincaré/ for example/was able to7howthactL 'isn m"but as an event: to cut/to be cut. The event itself is always /double/ because the
the two-dímensional case, there was a wefl defíned repertoire of singularities,"a^D^^n shÏps capadty to affect is contingent on the existence of things that have the capacity
shared by any process that can change in only two ways, whatever the causal'mech7nLU "beaffectedby it. Thus, while properties can be spedfíed without reference to anything
mvolved m thosechanges. He found four different types of point singularit-ydÍ5itÍnQüu'i^'^ p(l)se capacities to affect must always be thought in retation to capacities to be affected.
l°.rT^OT b,yth^form ofthe.fíow of nearby trafertories: nodes, saddlepoinTs°f^ 'Something that is real but not necessarily actual, like an unexerdsed capacity, is an
lnlc^TlHe.Ï0 discoyered a line 5íngularity dosed on ítself forming a loop"(a"li^t mple of real virtuality in the material world. Another example is. tendendes. A knife
cycleQ5 These exhaustedthe ways of structuring two-dímensional possibilityspaces. hás the property of solidity within a wíde range of temperatures, but in environments
^-tero,n one of his followers showed that/ m the three-dimensional case, th e four ^ exceed that range the knife may be forced to manifest Íts tendency to liquefy. Like
point síngularíties were still there but now líne singularities carne m three d(fferentformus': capacitíes, tendencíes are real even when they are not actua lly manifested, and when
stable^unstableand saddle-shaped loops.6 In fact, the state space for processes^ they do become actual it is as events: to melt orto solidify The main difference between
three degrees offreedom also contains another line singularity, one that canbe"píctured tendencies and capacities is that the former are typically finite but the latter need not be
as a loop thathas been repeatedly stretched and folded' Poincaré got a glimpse o^the^ We can enumerate, for example, the possible states in which a material entity will tend
strange singularities but they did not get a name until much later: deterministic chao^ to be (solid, liquid, gás, plasma), but capacities need not be tinite because they depend
Finally, other topologists discovered that not only singularities come in weH-defíned on the capacities to be affected of Ínnumerable other entities: a knife hás the capaclty
re.pertolre5but al50that sudden (or catastrophic^ swítches from one to another singular to cut when it interacts with something that hás the capacity to be cut; but it also hás
have an equally limited way o^happening. In catastrophe theory, for example, anyproces's the capadty to kill in interaction with large organisms that have the capacity to be killed.
Ïvelbya_potentlal thathas four degrces of freedom can have on'y seven dÍfferentways To return to the main argument, the singularities structuring mathematical possibility
of^swítching, ali of which have been classifíed. These seven catastrophes/ as'they'are spaces allow us to study tendendes, such as the tendency of a process to minimíse some
callecLdo not depend on the nature of the degrees of freedom, only their number,~"oron property, or its tendency to be attracted to an oscillatory state. But capacities cannot
the nature ofthe potential (gravitationat mechanical, chemical}, only on its existence7 be modelled the same way because they are only exercised in interactíons, interactions
From a philosophícalpointofviewit is this conceptual movementtowardsthemechanism- between an entity that can affect and another that can be affected. On the other hand,
independem levei, or ratherthe direction ofthe movement, that is híghly signifícant7because other forms of virtual reality can be used for this purpose, since in computer simulations
it poínts towards a real virtuality. Unlike the trajectories which alwayspomt Ín the direction of we can stage Ínteractions between simulated entities, vary the parameters under which
material reality, since they are possible sequences of states for the process being modelled, the Ínteractionstake place, and usethe results as a guide to map the structure ofthe space
the number of dimensíons of thç space, and the repertoíre of síngularities availableto~it,'do of possibilities. An interesting illustration is provlded by cellular automata, simulations
not refer to any aspect of the process^ They are topologíca! invaríants, properties of the space in which a population of fínite state automata Cmachines that can carry computations
that remaln unchanged regardless of how it is folded, stretched orrothewiseVransformed' without using memory) interacts and generates collective patterns of activity.
T^these mvariants do not point in the direction of anythíng actual is shown by'thefact In the well-known Game of Life, a cellular automaton in which the automata can be in
that those singularities on which trajectories converge are done só asymptotically, that is, the only one of two states, the interactions between neighbours generate collective patterns
traj-ectories get for ever dose to them without ever reaching them. In other words. the states of states that can be steady-state (blocks), periodic (blinkers) or mobile (gliders, shuttles).
the síngularities represent are never actualised, remaining for ever virtual. Using these emergent pattems as elementary buílding blocks, we can construct more

146 Computational Design Thinking

147 ReaLVirtuality
complex pattems (glider guns) and usíng these, in tum, we can build into the space NaturalJVIodel for Architecture
Game of Life an entire computer (Turing Machine).8The interactions between the fínn.
stateautomataaredefíned byrules, soto explainthe capacityoftheautomaton populatir prazer
to give rise to such a complex set of emergent patterns, we need to study the ]ohn
o// possibie rules. This space is divlded into four áreas Cdefining four dífferent classes john Frazer/s seminal book An Evolutionary Architecture C1995), from
mies) with different capadties: the fírst class of rules leads to fixed/ homogenous this essay is extracted, a fundamental approach is established for
the second class gives ríse to periodíc pattems; the third dass leads to relatively randnm natural systems can unfold mechanisms for negotiating the complex
configurations;andfínally,thefourthclassCor/classlV')producestmecomplexity, includir design space inherent in architectural systems. In this essay, which forms
locaiised structures capable of motion, like the glíders and other mobile patterns of Life s a criticai part of the book, Frazer draws both correlations and distinctions
The results of this analysís give us a sense of what a singularity would be in the possibilii from natural processes as emulated in design processes and forrn as active
spaces assodated with capacitíes: dass IV rules are singular or special not only because thf manifestations within natural systems. Form is seen as an evolving agent
lead to a maximum capacity for pattern generatíon but also because they are relatívely rgrg inerated via the rules of descriptive genetic coding, functioning as a part
constituting a small área ofthepossibilityspacesandwiched between class II and class III 1o of a metabolic environment. Frazer's process-model establishes the realm
Other simulations give us the means to explore other aspects of real virtuality: genetic in which computation must manoeuvre to produce a valid solution space,
algorithms, for example, are ver/ good at searching a space of possible solutions to a problem. including the operations of self-organisation, complexity and emergent
allowing us to conceive of biológica! evolution as a search in the space of possible animais behaviour. Addressing design as an authored practice/ he extends the
or plants; neural nets can map símilarities in the pattems used to train them into reiations of transference of 'creativity' from the explicit impression into form, to the
proximiïy in the space of possible strength patterns of their interconnections, allowing us to investment of thought, organisation and strategy in the computational
picture the mind as a kind of infoldíng of real virtuality; and multiagent systems let us stage processes which produce form. Frazer'5 text concentrates astutely on the
ínteractions between agents of different degrees of complexity to discover combinations of practising ofthe evolutionary paradigm, the output of which postulates an
mutually stabilising behavloural strategies that give us a giímpse into social possíbility spaces.11 architecture bom of the relationships to dynamic environmental and socio-
As memory and computing power decrease In price and our desktop computers acquire the economic contexts, and realised through morphogenetic materialisation.
computational capacity that supercomputers had a decade before, they will become not only
machines in whlch virtual realities are deployed as tools, such as 3D modelling and rendering
tools, but also a privileged place forthe sdentific and philosophical exploration of real virtuality, The Nature of the Analogy
Archítecture hás frequently drawn insplration from nature - from its forms and structures,
Notes
l Richard Feynman, TíieCharacterofPhysicalLaw,MlTPress{Cambndge,MA], 1997,pp50-3. and, most recently, from the inner logic of its morphological processes. It is therefore
2 DonSLemons, Perfect Form: Vanational Principies, Metïiods, and AppHcations m Elementai-y necessary to be clear as to where architecture is literally considered as part of nature,
Physics. Princeton Umversity Press (Princeton, NJ), 1997, p 7.
3 Ibid, pp 17-27.
where there are analogies or metaphors, and where nature is a source of insplration.1
4 A more detailed presentation of the relation between metric and nonmetric spaces in the spirit We can say that architecture is literaily pari of nature in the sense that the man-
ofFelixKlem'sErlangenProgramisdonem: Manuel DeLanda, Intensive Science aiid Virtual
made environment is now a major part of the global ecosystem, and man and nature
Philosophy, Continuum Press (London), 2002, pp 23-6.
5 June Barrow-Green, Pomcaré ãnd. the Three Body Problem, American Mathematical Society/ share the same resources for building. In turn, our description of an architectural
London Mathematical Society, 1997, pp 32-3. concept in coded form is analogous to the genetic codescript of nature. Analogies,
6 lan.StewaTt,DoesGodPlayDice:TheMatheniaficsofCíiaos,'ãasï[B].acï^veïí(OxïoYd~), 1989, p 107,
7 Alexander Woodcock and Monte Davis, Catastrophe Theoi-y, EP Dutton (New York), 1978, p 42.
particuiarly biologícal/ bedevil architectural writing. Sullivan, Wright and Lê Corbusier
8 The implemenlation of a Turmg Machine in the Game ofLifeY/asperformed in theyear 2000 by ali employed biological analogies, and the concept of the organic is central to the 20th
Paul Rendell, and can be seen in operation at his website,
century. In our case, the primary inspiration comes from the fundamental formative
9 SlephenWolíram, 'Universallly and Complexity in CellularAutomata', m CelIuSar Auíomaía and.
Complexity, Addison-Wesley (Readmg, MA), 1994, pp 140-55. processes and information systems of nature.2
10 Christopher G Langton, 'Life at the Edge of Chãos', m Artificial Li 'i "e II, edited by Christopher G
Langton, Charles Taylor, J Doyne Farmer, Steen Rasmussen, Addison-Wesley (Eedwood City, CA),
1992,p 44, The Problem ofthe Blueprint
11 These simulations are discussed in delail in: Manuel DeLanda, Pïiilosophy and Simulaüon: The Architecture/s use of biological analogies hás a counterpart in the use of architectural
Emergence ofSynthetíc Reason, Continuum (London), 2011.
analogies in science. There is an abundance of books with titles like The Cosmic
© 2011 John Wiley & Sons Ltd. Bluepnnt (by Paul Davies) or Blueprint for a Cell (by the Nobel Laureate Christian de

148 Computatíonat Design Thinking 149 ANaturaLIVlodelforArchitecture

SÏES^=^=&" icrtant to differentiate between the nature of different kinds of theories.

Éi^SS^ïïS^^
,'!lrÏÏi hás wntten: /Logic hás interests in abstract forms. Science investigates
'^onme^ context for their^erpr.ta^^v 1el. f^nnJ Design ímtiates novel forms. A sdentific hypothesis is not the same thíng as
,,,b'ueprint?arch'tectureandt°^^ï^

^^^^^^as^r^^^
Ia L L,nnthes!S. A lógica! proposition is not to be mistaken for a deslgn proposal/6 In
:^ ï "". a:genetic-ïan^^^f^^^^, F ^IJr nu rs is not aitheory of explanation/ but a theory of generation.
hor,.lt_mayprovide at iea"st ^^Mof^'2 gjs co^w^'
^3C^tnTOrk' even if it ls not a direct P-^^fZ ?,: 3
a fo'mil
tionofNature ^ _ ^ ^ ^ _^
"erfection and variety of natural forms is the result of the relentless experimentation
Efpvoiution Bymeans of profli9ate prototyping and the ruthless rejection of fíawed
? ev;lpn^. nature hás evolved a rich biodiversity of Ínterdependent species of plants
Ï animais that are in metabolic balance with their environment While vemacular

^onic that ^e^lth^^^^^^^n^^

these reverse analogjeï^ply.1^'' aLLUd"y lac'<s the methodo'°9í"l 'ncísivenesï^
might occasionally share some of these characteristics, the vast majority of
,;iriinas in our contemporary environment most certalnly do not.
Our- analogy of evolutionary architecture should not be taken just to imply a form of
Intentionality l development through natural selection. Other aspects of evolution, such as the tendency
eif-organisation, are equally or even more significant. Natural processes such as

S£ss???ï=^;ï:;.:^h-
^^=5SS=ï5^
metabolism and the operation of the laws of thermodynamics are central to our enquiry,
asare the general principies of morphology, morphogenetics and symmetry-breaking
Charles Darwin established a new world whích broke away from the Newtonian
a watch^ Ïh:1^^:::^^^^ :;,:tt;-lmpMthe^^^ of stability - a world in a contínuous process of evolution and change. Modern

;^^:=C^"^^^^B^^:Ë^:
contending fhat the blind forcïs'ofD^.U/,In^ncu.^llrnl,ü'lndwatchm3kerar9u^
physics now describes a world of instability. llya Prigogine hás discovered new propertíes
ofmatter in conditions that are farfrom equilibrium, revealing the prevalence of instability

exp^ ^^Z^l^^3^^^ ^^S

anïomïh;n:=:^^^^skins in tumh—^^
which is expressed by the phenomenon of small changes in imtial conditions leading to
large amplifications of the effects of the changes.7

P°eJ;el^^^rr:S:^rc:r[^-.^ ..nce „
Eïl^::^:^s^'S^r^^ïte?
Natural and Artificial Models
The modelling ofthese complex natural processes requires computers, and it is no coincidence
Purposeful intent.To usth7conn^Tn^''JlwtÏ°,a_pply it to our new mocfel with that the development of computíng hás been significantly shaped by the building of computer

^^:'S^^^^^^^z^
no^whenwefc^ ^m^ns^hlt^ 'deslgn',are veo' d?rentfrom';he models for simulating natural processes. Alan Turing, who played a key role in the development
of the concept of the computer (the Turing Machine), was ínterested in morphology and the

^^^s^^^^^m^^^^
W;thh.d,t^ S^gTvZLWlha^^t'tnOgJhinsn?dnes51can cause concem to^ho; simulation of morphological processes by computer-based mathematical modeis. The Church-
Turing hypothesis stated that the Turing Machlne could duplicate not only the functions of
conv,nced'thatTh7ha?ne^'nnanef\Z»o^o."tr0'."ke a computer^
^Z^Ï^^Ï^SSth--^^^o=;
mathematical machines but also the functions of nature. Von Neumann, the other key figure in
the development of computing, set out explicitly to create a theory which would encompass both
natural and artificial bíologies, starting from the premise that the basis of life was information.
Sourcesoflnspiration A signifícant example of this dual approach in terms of our genetic model is John
Ho!land's Adaptation in Natura! ana Artificia} Systems. Holland starts by looking for
^^^:e^^S=S^--ceso.e,,..,on. commonality between different problems of optimisation involving complexity and
^^'S^^^";™0"-itis "sent^^ha^^e
'^^^^s^^^w=a^^
uncertainty. His examples of natural and artifícíal systems range from /How does evolution
produce increasingly fít organisms in highly unstable envirpnments?' to 'What kínd of
^'^^^•^5 less' and mi-"de.tood ori;ven'h^^ economíc plan can upgrade an econom/s performance in splte of the fact that relevant
economic data and utilíty measures must be obtained as the economy develops?'8

150 Computational Desïgn Thinking

151 A Natural (VlodetforArchitecture
Although Holland suggests that such problems have no collective name_ ries. Construction remains labour intensive: Ít hás never made the transition
to share a common concem with questions of adaptation. They occur at rrii-i.-,i se611 lul ? l^^e^àus^ with adequate research and development capabilities. lt hás
ln .elds .as divef'se as evolution, ecology, psychology, economic plannnn Lp.olr1^ capJtÍ^k'to individual architects to take the risk of performing experimental and
artifícial intelligence, computationaf mathematics, sampfing and ínferencp Tn .ontroll l left ÏL/ninn in an uncoordinated and romantic or heroic manner. The ensuing
we must now add architecture. lisf Bvatlve.p[Zrpïha^ been catastrophic for both society and the architectural profession.11
|^table^d^';7inHinatÍons of some to return to a previous era are an equally inadequate
Artificial Life I^LnüatLoe^ylevel to the current predicament. The archetypes of the past do.not
In nature Ít is onlythe geneticallycoded ínformatíon ofform which evolves, butsF KstehfidcLhcanging'demands of society, the realities ofthe construction industry/ orthe
is based on the expression ofthis coded information in the outward form ofanorn^on! E^need for environmentally responsible buildings.
The codes are manufacturing instructions, but their precise expression is envïrnnml^';
dependent. Our architectural model, considered as a fomi of artifícialïife,also^ :lon . . . . , ..

coded manufacturing instructíons which are environmentally dependent, but as in lstlnut^e"is"'notthe only fíeld to be concemed with these problems. In industrial
real wodd model Ít is only the codescript which evolves, ^hle'all^m-bracing concept of mass production for a homogeneous intemation^
Kpt hás givenway to a search for a new fíexibility in design and manufacture-The
Generative Systems l characterístic of this approach is that Ít focuses on the dynamic processes

An essential Par"t of thís evolutionary model is some form of generative techniqug. Es^eyruexper1ence ratherthan on physical form.12 Or, in JohnThackara-swords, design

AgJ"rLth!s is an area charged..wlth problems and controver'sy' The history of generaiïve ,'now/beyondthe objecta13 ....,.,.,..
systems is summarísed by William Mitchell, who maps out a line from Aristotle to- LuÍI "Industnal production used to be associated with high tooling costs and very large
through the parodies of Swift and Borges. After tracíng back the use of generati^e Lodu'c>tÏonr'un's.This is now changing because the computer hás Pavedthewayforwhau
systems in architectural design to Leonardors study of centrady planned churches and rcalled 'the electronic craftsman7. The direct relationship between the designer at the
DurancTs Précis dês Leçons d-Architecture, he outlines the concept of -shape grammars". l^mputer console and the computer-controlled means of production potentiall^^ means
or dementai combinatorial systems.9 |nuot'Iust~a dramatic reduction in the production costs of the tools for mass ProdLiction;
From our point of view, there are several problems with this approach. AH of these l^dto^horter economic runs, but a one-to-one control of production and assembly
generatíve systems are essentially combinatorial or confígurational a problem which Rquipment/This is effectively a return to one-off craft technology, but with ali the
seems to stem from Aristotle/s description of nature in terms of a kit of parts that can lcapabilityoftheprecision machinetool.
be combined to furnísh as many varieties of animais as there are combinations of parts. As Charles Jencks desenhes it The new technologies stemming from the computer
Fortunately, nature is not actually constrained by the límítations ímplied by Aristotle. lhavemade possible a new facility for production. This emergent type is much more
Mitchell regarás architectural design as a special kind of problem-solving process. geared to change and individuality than the relatively stereotyped product-ive Proce5ses
This approach hás limitatíons which he recognises in principie. First, it assumes we can ofthe First Industrial Revolution. And mass production, mass repetition, was of course
construct some kind of a representation whích can achieve different states that can be one of the unshakable foundations of Modern Architecture. This hás cracked apart, if
searched through for permutations corresponding to some spedfied criterion Cthe criterion not cmmbled. For computer modelling, automated production, and the sophisticated
of the problema Unfortunately for this goal-directed approach, ít is notoriously difficult techniques of market research and prediction now allow us to mass produce a variety
to describe architecture in these terms/ except in the very limited sense of an architectural of styles and almost personalized products. The results are closer to nineteenth-century
^4 15
brief to whlch there are endless potential solutions. The other problem is that any serious handicraft than the regimented superblocks of 1965/14
system will generate an almost unmanageable quantity of pËrmutatÍons.10
The Environmental Case
Problems with Industrialisation Natural ecosystems have complex biological structures: they recyde their materiais,
A problem arose in the 1960s when architecture started toying with new desígn processes, pemiit change and adaptation, and make effídent use of ambient energy. By contrast,
takjng up a particular form of component-based rationalisation and methodology which most man-made and built environments have incomplete and simple structures: they
embraced a generic approach, modular coordination and a perception of construction as a kit do not recycle their materiais, are not adaptable and they waste energy. An ecologjcal
ofparts. For despite rhetoric to the contra ry, the architectural profession - and the construction approach to architecture does not necessarily imply replicating natural ecosystems, but
índustry as a whole - have falled to leam from developments in the aircraft automotíve and the general principies of Ínteraction with the environment are directly applicable.

152 Computational Design Thinking 153 A Natural ModetforArchitecture

The constmction industry is a signifícant consumer of raw materiais and
energy required to process, transport and assemble them. Buildings in use are pi0 th6
^ec^cM^
;rs not as an army of tedious clerks who will thwart ali creatívity with their
si9n'ficant consumers of energy for heating and cooling. An ecoíoglcal approach \^e i^lndsïor^ precise information, but as slaves of infínite power and patience, However
drastically reduce construction energy and materiais costs and allow most buildinnc .ould ^n +ers are not without their dangers. If used unimaginatively, they have a tendency
toexportenergyratherthan consume it ""'"y^inuse
ïjlcritical faculties, Índuce a false sense of having optimised a design which may
While there is a growing awareness of the Ímportance of envir ,„ ^mdamentally ill conceived, produce an atmosphere where any utterance from the
ecologically sound design among both archítects and enlightened clients, therei^^ii^ !^n ter is regarded as having divine significance, distort the design process to fít the
comprehensíve design theory and few built examples of an ecológica) architectureL"T^ ïl^-innç ofthe most easily available program, distort criticism to the end product rather
solutíon to our envíronmental problems may lie in relatíng architecture to the newïniil to an examination of process, and concentrate criticism and feedback on aspects of
understanding ofthe stmcture of nature. - - " — "-"v nuiistic
g problem which can be easily quantifíed.
/|magmative use' in our case means using the computer - like the genie in the
Responsive Environments... - to compress evolutionary space and time só that complexity and emergent
Another ímportant Íssue for our model of an evolutionary architecture is that it ntectural forrn are able to develop. The computers of our imagination are also a
n!.re^ponrs'y_e_t^evolv'!1g ln notjusta virtual but a real environment. In his article, The source of insplration - an electronic muse.19
Des(gn.of lntelligent Environments/, Warren Brodey proposed an evolutíonar^' seff!
or9an[5íng' complex, predictive, purposeful, active environment. He asked: can weteach The Role of Human Creativity
our environments fírst complex, then self-organisíng intelligence which we can ultir One of the world's leading exponents of neural networks, Igor Aleksander/ is also
refine Ínto beíng evofutionary? These issues preoccupy us too. very conscious of the unique capabilitles of the human brain. He reminds us that
jt is extraordinarily good at making guesses based on experience/ at retrieving
... and Soft Architecture knowledge from memory without the need for exhaustive searches, at perceiving
Brodeywent on to describe in enthusiastic terms some ofthe hypothetical impllcations of analogies and forming associations between seemingly unrelated items.20 These
intelligent environments, and to introduce the concept of 'soft architecture'16 aspects of intuition, perception and imagínation are the traditional creative engines
, ld-!a of a soft' r"esP°nsive architecture afso preoccupied NÍcholas Negroponte, for architectural ideas. While the model of architectural creativity proposed by this
author of TheArchÍtecture Machine. He suggested that the design process, considered text departs in many ways from the traditional model, it still relies on human skills
as evolutionary, could be presented to a machlne, also consídered as evolutionary, to for the essentiai first step of forming the concept.21 The prototyping, modelling,
gíve a mutual training resilience and growth.17 Negroponte placed high expectations testing, evaluation and evolution ali use the formldable power ofthe computer, but
fírst on computer hardware, then on software through artifídal intelligence. Neither the initial spark comes from human creativity.22
delívered any answers.

Problems of Complexity
The Role of the Computer -The 'sheer imponderable complexity of organisms' overwhelms us now as surely as it did
Chrístopher Alexander dísmíssed the use of computers as a design aid; /A digital Darwin in his time. The developmental processes of nature inevitably lead to complexity,
computer is, essentíally, the same as a huge army of clerks, equipped with rule books, buttheyworkwith simple building blocks and an economy of means to achieve complexity
pencil and paper, ali stupid and entirely without ínítiative, but able to follow exactly in a híerarchical manner.The coding ofall natural forms in DNA is achieved wíthjustfour
millíons of precisely defined operations ... In asking how the computer might be nucleotídes, which in turn usejust 20triplets to speclfy the amino-acidsthat manufacture
applied to architectural design, we must, therefore, ask ourselves what problems we protein. The hierarchical structure of living systems is analysed by James Miller on seven
know of in design that could be solved by such an army of derks ... At the moment, leveis: celi, organ, organism, group, organisation, societyand supranational system.These
there are very few such problems/18 are broken down into 1 9 criticai subsystems whlch appear at ali leveis. This hierarchical
Our evolutionary approach is exactly the sort of problem that could be given to and self-similar descripíion applies also to organisms and social systems, which are seen
an army of clerks - the diffículty lies in handíng over the mie book. Much of thís text as part of the same continuum.23
concerns the nature ofthese mies and the possibílities of devefoping them in such a way ... Even very simple local rules can generate emergent properties and behaviour in
that they do not prescríbe the result orthe process by whích they evolve. a way apparently unpredícated by the mies. CollectÍons of small actions ripple upwards,
combining with other small actions, until a recognisable pattern of global behaviour

154 Computationat Design Thinking 155 A Natural Model for Architecture

and as a consequence design hás come to be misunderstood ;as a problem-
emerges. When a certain criticai mass of complexity is reached, objects can approac
. activity'
ï acti^yl ^eory and Design in tïie SeccmdMacAmeAcre,BlackweU (Oxford) 1990
and self-reproduce in an open-ended fashion, not only creating their eaifalc ^'.adrlísÍ 5UJ-VJ^ ^ -j rJïllQ^

j^gïtin t'aw^^ Redefming Designmg: From Fonn to £xperience, Van Nostrand Remïiold (New
parenting more complicated objects than themselves. Von Neumann remnnic^., a^d C Tïio:rnas •l

depends upon reaching thls criticai levei of complexity. Life indeed exists on the pTil ifÍ Yorl<)pJ.9£ra, Design after Modemism, Thames and Hudson (London) 1988
^w^^~T^Language ofPost^Modem Architectnre, AcademyEditions i
chãos, andthis isthe pointofdeparture for our new model ofarchitecture. yeoi
"cílaries ii^^andïmade a presentation to the Royal Insütute of Bntish Architects Armu^
In' 197^A1T ^icmmafoTsurvival, which emphasised the need for a responsibleapproach fo ^
The New Model ofArchitecture co^rT.?lrmïSaAl&ough'therewas some genume mterest m resource problems it was dear
ener^a^l^u^alo?e~profession,notofthe planei, which
There is só far no general developed science of morphology, although the ülat it T. ufi^litecte hadse^edthe imtiative at this time and formulated a comprehensive energy
of form is fundamenta] to the creation of ali natural and ali designed artefactsL s ni ^on ironlcm^ ^ahthaveensured a future role for the profession. Imteadthe chance was wasted
PÏCY'.TU^ÏaSoosefit7low energy' campaign motivated more by poUtics than byasenou^
is stíll searching for a theory of explanation, architecture for a theory of qenerflti^
wtLa^unuSessttieissue.' See John H Prazer and Alex Pike, -Simple Societie^and Complex
and Ít Ísjust possible that the latter will be advanced before the former. In other ^StoÏeT^Ü-^^^^ Gonference (Lancaster 1972) and mBA
fomn-generating models developed for architectural purposes Cor based on unnr+h^
or incon-ect sdentífíc views) may be valuable if they model a phenomenon that y\p^\, ^Ï ^To^yeTheD^ oHntelligent Environments; Soft ArchitectureS Land^^^^

are seeking to explain. 1£J? NsOTODonte, raeA-ctítetíure Mac-Ame, MIT Press (Cambridge, MA), 1970,^
In a modest way some contribution hás already been made. Many papers relatir in ^^phe^Alexander, -The Question of Computers m Designa Landscape, Autumn 1967, pp

this work Cparticularly on computing and computer graphics) have been enthusiastic;
leeToto H Prazer, 'The Use of Computer-Aided Design to ExtendcrlatiyltJ^mH^
received at international conferences and cited in further publications. Many of ^^(ed^atíonaFconference Arï and Design m Education, electronic proceedings, NSEAD
resulting graphics techniques are now an integral part of the computer programs used
IÍ^Sander and Helen Morton, An Introducíion io Neural Computmg, Chapman and HaU
in architectural offíces. If taken much further, thís could contribute to the under^tanHir 20
of more fundamental form-generating processes, thus repaying some of th e debt that StssÍouisexp^dedmJHFrazerandJMFr^er/Design^mkmgjCrea^^^^^
21
architecture owes the sdentific field. Perhaps before the turn of the century there will be D^e^Lsionï\:n^reatíyerami<mg:AMuí^ceteáApproacA^^fe
^rïyPress~Ï9Ï47andJohn^HFrazer,TheArctetecturalRelevance^^
a new branch of science concerned with creative morphology and intentionalil
* * * l, vol 10, no 3, pp 43-4. . ^ ... . ^ , ,.
^uTrtalA±Kauffman"ÏelOrigzns ofOrder. Self-Organïzation and Selectíon m EvoMion, Oxford
22
Notes University Press (NewYork), 1993.
James GnerMiUer, Livmg Systems, McGraw-Hill (NewYork), 1978.
l As earlyas 1969 Charles Jencks was predictingthatbiologywouldbecome the major metaphor 23
forthe 1990s andthesource ofthe most significant architectural movement this century-the
From John Prazer, •A Natural Model for Architecture-, An Evolutionary
Biomorphic School. See Charles Jencks, Arcïutectm-e 2000: Predicíions and Methods, Studio Vista
(London), 1971, Architecture, Architectural Association Publications (London), 1995, pp
2 A criticai ovemew of biológica] analogies is given m PhHip Steadman, Tíje Evolution ofDesigns: 9-21. Reproduced by permission ofjohn Prazer and the Architectural
Biológica! Analogy in Architecture and tlie Applied Arts, Cambridge University Press (Cambridge; Association Publications. © Architectural Association Publications.
NewYork), 1979.
3 D Barthélémy et al, 'Architectural Concepts for Tropical Trees' in LB Holm-Nielsen et al (eds),
Tropical Forests, Academia Press (London), 1989.
4 WiHiam Paley, Natural Theology, or Evidences of the Existence ana Attríbutes of tïie Deity
Collecïeà from íhe Appearance ofNature (Oxford), 1802.
5 RichardDawldns, TheBlindWatclimaker.LongmanÇLondoTï), 1986, and Tiie Selfish Gene, Oxford
University Press (London), 1976,
6 LionelMarch (ed), The Architectuï-e ofForm, Cambridge University Press (Cambridge), 1976.
7 Grégoire Nicolis andIlyaPrigogine, Explonng Complexity, Freeman (NewYork), 1993.
8 John H Hoüand, Adaptation in Natural and Artificial Systems, Universify ofMichigan Press (Ann
Arbor), 1975.
9 WiUiam J MitcheU, Ttie Logic ofArcMlecture Design, Computation, and Cogmïion, MIT Press
(Cambndge, MA; London), 1990.
l O These difficulties stem largely Êrom the views of Herbert A Simon as expressed in Tíie Scíences
oí 'tïie Artificial, MITPress (Cambridge, MA;London), 1969, an exploration of the problem of
complexity with particular reference to artificial (man-made) systems. While Simon's views on
the inevitability of a hierarchical strategy m nature have been influential in the formulation of our
own theories, ïïis scientifïc method does not recognise the need for a generating concept when

156 ComputationaL Design Thinkmg 152 A Natural Modet for Architerture

'v>^lP^logenesis^ndthe Mathematics of Emer tructed systems. We shouid start by asking: what is Ít that emerges,
1cla"y ^emerge from, and how is emergence produced?
MichaeLWeinstock lt e!Ltirthas always played a criticai role in architecture, but the character and
^!noTmathematics in relation to the theories and the material objects of architecture
Michael Weinstock hás contributed significantly to establishir K^pHsói that a detinitive account remains elusive. It is evident that there is a
theoretical and practical foundations for employmgdesÍgn~comPut'Z.the {!e^nn need for a more developed mathematical approach in current architecture.
?!r.ira.t!?1g.^ch.ite^.L"'al system5 based on the 109ics and me<=hani's^s'lÏ ?_! +hp iberation of tectonics from the economic straitjacket of orthogonalgeometry
natural models. In this important text, from the Emergence: MorphoQe^l Ínrk more precision in the interface between architectural definitions of form and
Design Strategies issue (2004) of Architectural DesSgn. he culFscZrT^ Íma^nnter-driven fabrication processes of manufacturing constructors. Second, the
and techniques from the fíelds of mathematics and biology"to7r'e^ le.^prïna design for the complex geometries of contemporary tectonícs must begin
^\ap^,r.0!?-wh^e emer9ence/ within a "mputational'environine'ntl lm a deflnitive mathematical base. And third there is a lacuna m the theoretical body
is a^valuable and necessary component for the generation of~'rohiïcl. E)architecture'an absence that is marked by the proliferation of design Proces5es_that
^^hitecturalsystems: weinstock lays out the prinaples oforgan,sautZÏ !nrarowtheappearanceofscientifícmethodsyetlacktheirdarítyofpurpose,mathem
w and interaction of natural systems, along with theïr nstruments and theoretical Íntegrity.
correlating mechanisms in mathematics, cybernetics and systems There is an intricacy in the interchange of ideas and techniques between the
to^ropose a computationally based paradigm for design.'Across"the^ jisciplines of biology, physical chemistry and mathematics.1 They are seParate' almo5t
domains, the characterisation of pattern becomes essentÏal. As Weinstock Eiscrete but overlapping, and the boundaries between them are indeterminate-The
exp!ains'.pattern is.always an outcome, bom of local relationships/and not loriginating concepts and the subsequent development of emergence are founded in
rlea'Í16d throu?h.910bal_ImPosit'o" or projection. Pattern is an'Ímage"of Ithese interchanges.
growth, not of fínality. The repercussion of this is that if architectjreï
modelled on natural systems then it must be modelled as itself aidynamic
process and Form
^l,n:.and,_WJthin,a.lar?er.evolving /ecolo9y/- Weinstock succinctly ÍLiving organisms can be regarded as systems, and these systems acquiretheir complex
establishes the foundation for this approach and"a metabolistic envisïomr :forms and pattems of behavíour through the Ínteractions, in space and over time, of
of architectural and cultural systems. ; their components. The dynamics of the development of biological forms, the accounts
ofgrowth and form, of morphogenesis, have become much more central to evolutionary
Emergence is a concept that appears in the líterature of many disciplines, and is • theory than in Darwinrs thesis. Darwin argued thatjust as humans breed living organisms
COLrdateclto eTOIUtionary biol°gy/ artifídal intelligence, complexity theory,~cybemuet^ by unnatural selection, organising systematic changes in them, só wild organisms
and general ^systems theory. [t is a word that islncreasingly common~in"arc'hitectu^ themselves are changed by natural selection. Natural selection is a process involving
e, where too often it is used to conjure complexity but without the attendant predators and parasites, and environmental pressures such as food supply, temperature
^clpts.and mathematical instruments of science. In the simplest commonly used and water. Successful organisms will survive the fíerce competition and have greater
n, emergence is said to be the properties of a system that cannot be deduced breeding success and, in turn, their offspring wilt have greater reproductive success, and
^components, something more than the sum of íts parts. This description is só on. DarwÍrTs arguments had a remarkable alignment with the then current theory of
^true in a very general sense, but rather too vague to be useful for the purpose competitive struggle in capitalism2 and the concepts of mechanisms in Índustry. Theories
iresearch In architecture. One can tmthfully say, for example, that everyhigher- of morphogenesís, the creation of forms that evolve in space and over time, are now
property can be described as a consequence of lower-level properties. In inextricably entwined with the mathematics of Ínformation theory, with physics and
Ïlsaences/ the term,refers to the Production of forms and behaviourby natural systems chemistry/ and with organisation and geometry. The pattern of alignment with concepts
^have an irreducible complexity, and also to the mathematical approachn'eces^ryito and technologies of economics and industry remains consistent
model such processes in computational environments. The convergent lines of thought between biology and mathematics were initiated
^Thetask for architecture is to delineate a working concept of emergence and to early in the 20th century, particularly in the work of Whitehead and DrArcy Thompson.
outlme the mathematics and processes that can make Ït usefuÍ to us-asde7igner5.ThÍs D/Arcy Thompson, zoologist and mathematician, regarded the material forms of living
means we must search for the principies and dynamics of organisation and interaction, íhings as a diagram of the forces that have acted on them.3 His observations of the
mathematical laws that natural systems obey and that can be utifised by homologíes between skulls, pelvises and the body plans of different spedes suggested

158 Computational Design Thinking

159 Morphogenesis and the Mathematics of Emergence
a new mode of analysis, a mathematisation of biology. Morphological measurements arç Cybemetics organises the mathematics of responsive behaviour into a general
specifíc to species and at times to individuais within a species/ and só are various, but theory of how machines/ organisms and phenomena maintain themseives over time.
there are underlying relatlons that do not vary - the 'homologies'. |t uses digital and numerical processes in which pieces of information interact and the
Homology hás two distinct but related meanings: to biologists it means organs or transmission of information is optimlsed.5 Feedback is understood as a kind of /steering/
body parts that have the same evolutionary origin but quite different functions; and to device that regulates behaviour, using information from the environment to measure the
mathematidans it is a classifícation of geometric figures according to their properties. actual performance against a desired or optimal performance.
Form can be described by mathematical data, mapping points in 3D coordinate space. Work in thermodynamics by Prigogine7 extended this (and the second !aw of
dimensions, angles and curvature radii. D'ArcyThompson's comparison of related forms thermodynamics) bysetting upa rigorousandwell-grounded studyofpatternformation
within a genus proceeds by recognising in one form a deformation of another. Forms are and self-organisation that is still of use in the experimental study and theoretical
related if one can be deformed into another by Cartesian transformation of coordinates. analysis of biological and nonbiologica! systems. He argued that ali biological organisms
Comparative analysis reveals what is missing in any singular description of a form, no and many natural nonliving systems are maíntained by the flow of energy through the
matter how precise, and that is the morphogenetic tendency between forms. systÊm.Thepatternofenergyflowissubjecttomanysmallvariations,whichareadJusted
At around the same time the mathematician and philosopher Whitehead4 argued by /feedback/ frorn the environment to maintain equilibrium, but occasionally there is
that process rather than substance was the fundamental constituem: of the world, and such an amplifícation that the system must reorganise or collapse. A new arder emerges
that nature consists of patterns of activity interacting with each other. Organisms are from the chãos of the system at the point of collapse. The reorganisation creates a
bundles of relationships that maintain themselves by adjusting their own behaviour in more complex structure, with a higher flow of energy through it, and is in turn more
anticipation of changes to the patterns of activity ali around them. Anticipation and susceptible to fluctuations and subsequent collapse or reorganísation. The tendency of
response make up the dynamic of life. 'self-organisecT systems to ever-increasing complexity/ and of each reorganisation to be
The union of these two groups of ideas is very interesting - form and behaviour produced at the moment of the collapse in the equilibrium of systems extends beyond
emerge from process. It is process that produces, elaborates and maintains the form or the energy relations of an organism and its environment. Evolutionary development in
structure of biológica) organisms Çand nonbiological things), and that process consists of general emerges from dynamic systems.8
a complex series of exchanges between the organism and Íts environment. Furthermore.
the organism hás a capacity for maintaining its continuity and integrity by changing Geometry and IVlorphogenesis
aspects of its behaviour Forms are related by morphogenetic tendencies, and there is Theoreticians fíercely contest the precise relatíonship of morphogenesis to genetic
also the suggestion that some, if not ali, of these characteristícs are amenable to being coding/ butthere is an argumentthatit is nottheform ofthe organísmthat is genetically
modelled mathematícally. The ideas are particularly relevant to us, as in recent years encoded but rather the process of self-generation of the form within an environment.
both architecture and engineering have been preoccupied with processes for generating Geometry hás a subtle role in morphogenesis. It is necessary to think of the geometry of a
designs of forms in physical and computational environments. bíological or computational fornn not only as the description of the fully developed form,
but also as the set of boundary constraints that act as a loca! organísing principie for self-
Pattern, Behaviour and Self-Organisation organisation during morphogenesis. Pattern and feedback are as signífícant in the models
Form and behaviour have an intricate relationshíp. The form of an organism affects Íts of morphogenesis as they are in the models of cybernetics and dynamic systems. Alan
behavíour in the environment, and a particular behaviour will produce different results Turing put forward a hypothesis of geometrical phyllotaxis, the development of form in
in dífferent envlronments, or Íf performed by different forms in the same environment. plants, which offered a general theory of the morphogenesis of cylindrical lattices. These
Behaviour is nonlinear and context specific. Mathematical descriptions of behaviour are areformed locallyratherthanglobally/ node bynode/ and arefurthermodifíed bygrowth.
found in the elaboratíon of Whitehead's 'anticipation and response' by Norbert Wiener, To mathematically model this process/ it is necessary to have a global informing geometry,
who developed the first systematic description of responsive behaviour in machines and the cylinder, and a sei of local ruies for lattice nodes.
animais.5 Wiener argued that the only significant difference between controlling anti- Turíng had a life-long interest in the morphogenesis9 of daísies and fir cones, of
aircraft fíre and biological systems was the degree of complexity. He had developed new polygonallysymmetrical structures such asstarfish, ofthefrequencyoftheappearance
programs for ballistics guidance, in which information about the speed and trajectory of a of the Fibonacci series in the arrangements of leaves on the stem of a plant/ and of the
target is input to a contrai system só that anti-aircraft guns could be aimed at the point formation of patterns such as spots or stripes. HÍs simple earfy model of morphogenesis10
where a target would be. The control system could record and analyse the data from a demonstrates breakdown of symmetry and homogeneity, or the emergence of a
series of such experiences and subsequently modify its movements. pattem, in an originally homogeneous míxture of two substances. Equations describe

160 Compuíational Design Thinking 161 Morphogenesis and the Mathematics of Emergence
the nonlinear changes of concentrations of two chemicals (morphogens) over time, as Cupimjngs argues that the interaction and diffusion of morphogens in cellular
the chemicals react and diffuse into each other. lt offers a hypothesis of the generatíon ;^isaisoaffected bythe Gaussian and mean curvatureofthe particular membrane
of pattern from a smooth sheet of cells during development in the formation of buds. ,13 só that changes to any particular curve in a curved membrane will produce
skin markings and límbs. Chemicals accumulate until sufficient density is reached,then changes to cun/ature elsewhere. Successful computational models have
act as morphogens to generate organs. extended this approach by incorporating the mathematics of curvilinear
The reaction-diffusion model is still of interest in mathematical biology, in which Qi-dÍnate meshes used in fluid dynamícs to the simulation of the morphogenesis of
a great deal of research is concentrated on the Ínterrelated dynamics of pattem and imetrical organs and limbs.14 Folding and buckling offíat and curved sheets ofcells
form. Turing's model operates on a slngle plane, or a flat sheet of cells. Some current gre the basis of morphogenesis in asexual reproduction.
research11 in the computational modelling of morphogenesis extends the process that The lineages of organisms that reproduce asexually exhibit convergent evolution,
Turing outlined on fíatsheets to processes in curved sheets. Ceometry is inherent inthese similar forms and properties emerging without common ancestors. There are generic
models of process, incorporating morphological 'units' that have a dynamic relationship •rns in natural morphogenesis/ and this adds to the set of geometncal operations
to each other, and to an infonning global geometry. in these mathematical models. An intricate choreography of geometrical constraínts
and geometrical processes is fundamental to self-organísation in biological
morphogenesis. Computational models of morphogenetic processes can be adapted
Spiral organiâation
ofbroccoliflorets, for archítectural research, and self-organisatlon of material systems is evidenced in
displaying a Fibonacci physícal form-finding processes.
mathematical series
in which each number
is the sum ofüie two The Dynamics of Differentiation and Integration
previous numbers,
Feedback is not only important for the maíntenance of form in an environment; it
evident in many
natural patterns, is also a useful concept in modelling the relationship of geometrical pattern and
George Post/Science form during biológica! morphogenesis. In pattern-form models, feedback is organised
Pïioto Lïbrary ©
Science Photo Library. in two loops: from form to pattern and frorn pattem to form. In these models the
unstructured formation of biochemical pattem causes morphogenetic 'movements'
and a consequent transformation in geometry. The change in geometry dísmpts the
pattem and a new pattem emerges, which initíates new morphogenetic movements.
The process continues untíl the distribution of morphogens is in equilibrium with the
geometry of the evolving form in the model. The feedback loops, from pattem to
form and from form to pattern, constmct a mathematical model of morphogenesis15
as a dynamic process from which form emerges.
Cybernetics, system theory and complexity have a common conceptual basís, as is
evidencedbythefrequencyoftheterms'scíencesofcomplexity/and/complexadaptíve
systems' in the extensive literature of thermodynamics, artificial intelligence, neural
networks and dynamical systems. Mathematically, too, there are commonalities in the
approach to computational modelling and simulations. It is axiomatic in contemporary
cybemetics that systems increase in complexity/ and that in natural evolution systems
emerge in increasing complexity, from cells to multicellular organisms, from humans
to society and culture.
System theory argues that the concepts and principies of organisatíon in natural
systems are índependent of the domain of any one particular system, and contemporary
research tends to concentrate on 'complex adaptive systems' that are self-regulating.
What is common to both is the study of organisation, íts structure and function.
Complexitytheory16formalisesthe mathematical structure ofthe process of systems from

162 ComputationatDesignThinking 163 (Vlorphogenesis and the Mathematics of Emergence

which complexity emerges. It focuses on the effects produced by the collective behaviour problem of how to introduce the dynamics of gene processes into Cummings' model of
of many simple units that interact with each other, such as atoms, molecules or cells. The morphogenesis is suggested by Kauffmanrs work on periodic attractors in genetic networks.
complex is heterogeneous, with manyvaried partsthat have multipfe connections between The concepts and mathematical techniques to produce collective behaviour from
them, and the different paris behave differently, although they are not independent. simple local responses have the potential to radically change architectural environmental
Complexity increases when the variety (distinction) and dependency (connection) of systems. It is evident that the current methods of producing 'smarf buíldings with
parts increases.The process of increasing variety is called differentiation, and íhe process hybrid mechanical systems that are controlled by a remate central computer are inferior
of increasing the number or the strength of connections is called integration. Evolution conceptually and prone to failure in operation. The self-organising capabilities of
produces differentiation and integration in many 'scales' that interact with each other, dlstributed dynamic systems have produced intelligent behaviour in natural organisms
from the fomnation and stmcture of an individual organism to species and ecosystems. and in computational simulations, and await architectural applications.
Models of self-organisation based on distributed variation and selection have been
The Genetics of Collective Behaviour put forward by Heylighen.20 In this argument is the view common to many approaches, in
The coliective behaviour of semi-autonomous individual organisms is exhibited in the which complexsystems, such as organisms and ecologies, are evolved fromthe interactions
social or group dynamics of many natural species. Flocks of birds and schools of físh of elements that combine into a variety of 'assemblies'. Some 'assemblies' survive to go on
produce what appears to be an overall coherent form or array, without any leader or to fonn naturally selected wholes, while others collapse to undergo further evolution. The
central directing intellígence. Insects such as bees and termites produce complex built process repeats at higher leveis, an emergent whole at one levei becoming a component
artefacts and highly organised functional spedalisations without central planning or of a system emerging at a higher levei.31 Further, natural evolution is in general not a
instructions. Structured behaviour emerges from the repetítion and Ínteraction of simple single system but distributed or parallel, with multiple systems co-evolvíng with partial
rules. Mathematical models have been derived from natural phenomena, massively autonomy and some interaction. Self-organisation of the ecology as a whole is as criticai
parallel arrays of individual /agents/, or 'cell units' that have very simple processes in each as the morphogenetic self-organisation of an individual wíthin Ít.
unit, with simple interactions between them. Complex pattems and effects emerge from A very recent series of models have been developed based on phenotypes rather
distributed dynamical models. Wolfram/s extensive study of cellular automata17 offers a than genotypes or genetíc actlvity.22 Stewart argues that ali self-organising systems are
comprehensive account of their characteristics and potential. The study and simulation of organised notjust bythemselves, butalso bytheir contexts.The focus in this approach
co-evolution and co-adaptation is particularly effective in distributed models. to modelling evolutionary self-organisation is on speciation, fítness landscapes and
The current work at the Santa Fe Institute uses heterogeneous networks of interacting species selection. The models exhibit nonlinear and collective effects and represent
agents to pursue studies in self-organisation. The objective is to develop computational specíation as a symmetry-breaking bifurcation. They are collections of ordinary
simulations of life, Ínvolving gene activity, molecular dynamics, morphogenesis and life differential equations arranged in network structures, and the network topology is
cycles - virtual organisms with realistic morphologies that have artificial life. refíected in the form of the equations. Recombination is preferred to mutation, as it
Adaptive processes in natural systems provided the initiating concepts for the produces substantial diversity in each generation, and selection operates on individual
fírst genetic algorithms, developed by Holland to design artifícial systems based on organisms in the context of other organísms and the envíronment.
natural ones.18 Genetic algorithms inítiate and maintain a population of computational
individuais, each of which hás a genotype and a phenotype. Sexual reproduction is Architecture and Emergence
simulated by random selectlon of two individuais to provide 'parents' from which In answer to the question: what is it that emerges/ what does it emerge from, and how is
/offspringr are produced. By using crossover (random allocation of genes from the emergence produced? we can say the following.
parents' genotype) and mutation, varied offspring are generated until they till the Form and behaviour emerge from the processes of complex systems. Processes
population. Ali parents are discarded, and the process is iterated for as many generatíons produce, elaborate and maintain the form of natural systems, and those processes Ínclude
as are required to produce a populationthathasamong it a range ofsuitable individuais dynamic exchanges with the environment. There are generic patterns in the process of
to satisfy the fitness criteria. They are widely used today in control and optimisation self-generation of forms, and in formsthemselves.Ceometry hás both a local anda global
applications and the modelling of ecológica! systems. role in the interrelated dynamics of partem and form inself-organised morphogenesis.
fVlathematical simulations of genes acting in Boolean networks with varying strengths Formsmaintaintheircontinuityandintegritybychangingaspectsoftheirbehaviour
of connection can produce differentiation of tissues and organs in models, and Kauffman' and by their iteration over many generatíons. Forms..exist in varied populations, and
argues that the self-organisation produced by such networks is complementary, ratherthan where communícation between forms is effective, collective structured behaviüur and
opposed, to Darwin's selection by fitness to the environment The solution to the central intelligence emerges.

164 Computational Design Thinking 165 Morphogenesis and the Mathematics of Emergence
3 D'ArcyThompson, On Growth andForm, CambridgeUniversityPress, 1961 (fírst published 1917).
The systems from which form emerges, ancf the systems wíthin individual 4 Mred North Whitehead, The ConceptofNature, Cambridge UniversityPress, 1920,
forms themselves, are maintaíned by the flow of energy and information through 5 Norbert Wiener, Cybemeíics, or Coníroland Commuiiication m tïie Animal ana tfieMachine, MIT
Press (Cambndge, MA), 1961,
system. The pattern of flow hás constant variations/ adjusted to maintain equilibrium
g CE ShannonandWWeaver, The Maïïiernatical Tïieory of Communicaïion. fífth edition, University
'feedback' from the environment. Natural evolution is not a single system but distributed of Illmois Press (Chicago), 1963.
with multiple systems co-evolving in partial autonomy and with some interaction. /\^ 7 Uya Prigogine, Introduction io Thermodynamics o f IiTeversible Processes, ]oïmWï[ey & Sons
(Chichester), 1967,
emergent whole form can be a component of a system emerging at a higher leve! - and
g Any physical system tïiat can be described by mathematical tools or heuristic piles is regarded as
what is /system/ for eme process can be 'environmenf for another. a dynamic system. Dynamic system theory classifies systems by the mathematical tool rather than
Emergence is of momentous importance to architecture, demanding substantial the visible form of a syslem.
g PT Saunders (ed), The Collected Works ofAM Turíng, Volume 3, Morphogenesis, includes 'The
revisions to the way in which we produce designs. We can use the mathematical models
Chemical Basis of Morphogenesis {Pïulosophical Transacïions, 1952), 'A Diffusion Reaction Theory
outlined above for generatíng deslgns, evolving forms and structures in morphogenetic of Morphogenesis in Plants', 'Morphogen Theory ofPhyllotaxis: Geometrical and Descriptive
processes within computational environments. Criteria for selection of the 'fittest can Phyllotaxis' and 'Chemical Theory ofMorphogenesis'.
IO AlanTuring, 'The Chemical Basis ofMorphogenesis', PMosophical Transactíons, 1952.
be developed that correspond to architecturai requirements of performance, índuding
11 ChristopherJ Marzec, 'Mathematical Morphogenesis', Institute of Biomolecular Stereodynamics;
structural integrity and 'buildability'. Strategies for deslgn are not truly evolutionai •A Pragmatic Approach to Modéü.ïngMorpïioge'nesï.s'',J'oumalofBiologicalSystems, vol 7, 2,

unless they include iterations of physical Cphenotypic) modelling, ÍncorporatÍng the self- 1999; The Morphogenesis ofHigh Symmetry: The Symmetrisation Theorem'. Joumãl of Biological
Systems, vol 7, l, 1999; The Morphogenesis ofHigh Symmetiy: TheWarpingTheorem',/ourj33jo/
organising material effects of form finding and the industrial logic of production available Biological Sysïems, vol 7, 2, 1999.
in CNC and laser-cutting modelling machines. 12 LG Harrison and M Kolar, ' Coupling between Reaction-Diffusion Prepattem and Expressed

The logic of emergence demands that we recognise that buildings have a lifespan, Morpïioge-n.es.i.s^fournalofTheoreticãlEiology, 130, 1988, and A Hunding, SAKauf&nan andBC
Goodwin, 'Drosophila Segmentation; SupercomputerSimulation ofPrepattem Hierarchy',7ouma^
sometimesofmanydecades, and thatthroughoutthat lifethey haveto maintain complex
ofTheoreticalBiology, 145, 1990.
energy and material systems. At the end of thelr lifespan they must be disassembled 13 FW Cummmgs, 'APattem Surface Interactive ModelofMorphogenesis^/ouraalo^Ïïieorefaca^
Biology, 116, 1985; 'On Surface Geometry Coupledto Morphogen',/oumaJo^TAeoreüc3^-BjbIogy,
and the physical materiais recyded. The environmental performance of buildings must
137, 1989, 1990; 'AModel ofMorphogeneticMovemenfJouraaJo/TAeoreüca^B^ogy, 144, 1990,
also be rethought. The cun-ent hybrid mechanical systems with remote central processors 14 CH LeungandMBerzins, A Computaíional Model for Organism Growtïi Based on SurfaceMesh
limit the potential achíevement of 'smart' buildings. Intelligent environmental behaviour Generation, University ofLeeds, 2002.
15 Alexander V Sprirov, 'The Change in Initial Symmetry m the Pattem-Form Interaction Model of Sea
of individual buildings and other artefacts can be much more effectively produced and
UrchmGa.s^ruiation.Joi.imalofTheoreticalBiology, 161, 1993.
maíntained by the collective behaviour of distributed systems. 16 Warren Weaver, 'Science and Complexity', Amerícan Scientisí, 36, 536, 1948.
We must extend this thinking beyond the response of any single individual building 17 StephenWol&-am,AIVetvfiadof5'ae^c'e,WolframMedia(IL), 2002, and Ceïlular Auíomaïa ana
Complexíty: CoBected Papers, Addison-Wesley (Reading, I\'IA). 1994.
to íts envíronment Each building is pari of the environment of its neighbours, and it
18 JohnHHoüand, Adapïaïion in Natural and Artificial Sysïems: An Introductory Analysis wiíh
follows that 'urbarT environmental intelligence can be achieved by the extension of data Applicatioiïs to Bíology, Contrai and Aiïifícial InteJligence, MITPress (Cambridge, MA), 1992
communication between the environmental systems of neighbouring builcfings. Urban (fírst published 1975), and Hídden Order: How Ad.aptation Bíiüds Complexiíy, Addison-Wesley
(Eeadmg.MA), 1996.
transport Ínfrastructure must be organised to have similar environmental responsive
19 SAKauffman, 'Antichaos andAdaptation', Scientifíc Amencaii, August 1991; The Orígins ofOrder:
systems, not only to control internai envíronments of stations and subways, but also Self- Organization and Selection in Evolution, OxfordUniversityPress, 1993; At Homem ïhe
to manage the response to the fíuctuating díscharge of people on to streets and into Universo: The Search for Laws of Self-Organization and Complexity, OxfordUniversityPress, 1995.
20 Francis Heylighen, 'Self-Organization, Emergence and the Architecture of Complexity',
buildíngs. Linking the response of Ínfrastructure systems to groups of environmentally
Proceedings cf l st European Conference on System Scíence, 1989 ,
Íntellígent buildings will allow hígher-level behaviour to emerge. 21 HASimon, 'The Architecture ofComplexity', Proceedings of ïhe American Phiïosopnical Society

We are within the horizon of a systemíc change, from the deslgn and production of 106,reprintedinTAe5cjej-3ceso/ttieArOÏ3aaíthirdediüon,MITPress(Cambndge,MA), 1996.
22 IStewart, 'Self-Organization in Evolution: A Hathemaücal Perspective', Pïiilosophical Tsransactíons.
individual /signature/ buildings to an ecology in which evolutionary designs have suffícient
The Royal Society ofLondon, 361, 2003.
Íntelligence to adapt and to communicate, and from which intelligent cities will emerge.
Michael Weinstock, 'Morphogenesis and the Mathematics ofEmergence',
Notes
AD Emergence: Morpïiogenetic Design Strategies, vol 74, no 3, John
l I include computation as part of mathematics for the sufficient reason that ali computation proceeds
from low-level operations on l and 0. Wiley & Sons Ltd (London), 2004, pp 10-17. Reproduced by permission of
2 In an acquisitive hereditary society he stated acquisition and inheritance as the primary means of Michael Weinstock and John Wíley & Sons Ltd. © John Wiley &• Sons Ltd.
smvival'; Geoffrey West, Cïiaries Darwin; A Porïraïf, Yale University Press (NewHaven, CT), 1938,
p 334, and 'the application ofeconomics tobiology': Oswald Spengler, The Decline offhe West,
Knopf(NewYork), 1939, p 373.

166 Computational Design Thinking 16; Morphogenesis and the Mathematics of Emergence
Philosophy of IVlathematícs for Computatíonal Design gbstract, linguistic, mathematical, diagrammatíc and perhaps logícal means. There may be

Spatíal Intuition Versus Logic cio transcendent, objective view of the whole. In relinquishing the primacy of the object,
thevalueoftheimage alsofades.Whetherwespeakofthe imageasthe projection ofthe
]ane Burry object into two dimensions in the sdentific tradition of Desargues or of Bachelard/s poetíc
jmage that collects and creates resonance in imagination,1 the system model realised
In this text/ Jane Burry explains a specifíc contemporary application through design computation on a programmable machine confounds the ocular-centridty
of computation to design - one in which the foci of design stems of the deslgner. The image that hás been central and all-powerful in design thinking,
from serial defínitions of dynamic spatial constructs. Burry proposes jn the tield of both the outer and the inner eye, gives place to less immediate ways to
that the conception of such a design space lays criticai bearing on the know the model space in computatíonal design. Donald Schòn gives us three types of
understanding of geometry and the mathematical means by which it is seeing for designing: literal visual apprehension, appreciative judgements of quality and
presented. While it is geometry which provides a particular depiction, apprehension of spatia! gestalts.2The fírst and the last, at least, are compromised as we
it is the mathematical relationships which define the "state space' - the move into model spaces that can be experienced as having as many spatíal dimensíons
range of morphological potentials. As Burry delineates, it is in the history as they have variables or degrees of freedom. The second, appreciative judgements
by which the theoretical relations of space, geometry and mathematics of quaíity are adaptable to aesthetíc frameworks predicated more on /deepr pattern
have evolved that one can fínd the means where computational spatial recognition through more logical, intellectual and lesssensorially led apprecÍatÍon.3These
design can be establishecL This also unfolds a surprisingly synchronous are neither visible nor readily visualisable spaces. With reference to Nigel Crossrs warnings
relationship between the mathematician/s pursuit of a solution and about failure to recognise the distinct nature of design in relation to science4 we must
the designeis computational deduction of a specifíc spatial construct. now ponder howthis computational design space is assimilated into design's own distinct
Seemingly defined by precision, it is more appropriately in the progression 'thlngs to know, ways of knowing them, and ways offinding out about them'.5
from abstraction and initial supposition to proven functionality where
computational design can be seen as acting in the mathematician'5 Representation, Perception and Mathematics
nature. Burry utilises this intersection to succinctly interject the notion Cross hás also wriíten that the central concern of deslgn is the conception and realisatíon
of intuition as both a fundamental and critically involved aspect of the of new things and at its core is the language of modelling.6 This language is changing,
mentality and practice of computational design. orchestrated through tentative appropriation from mathematícs and the prior experiences
of computer science in design computation. In being drawn Ínto this encounter with
Over the (ast half century, architecture hás been slowly adapting its representationaf multidimensional spaces and deeper tadt understanding of topology and the Erlangen
practices from the conception of objects of sensory engagement to the construction of Program/ reliving in a small way the seismic shifts in geometry in the 19th century,
systems offomnally described relationships. This shift from object descri ption to definition questions are raised for computational design thinking regarding intuitíon, aesthetics and
of a dynamícal space of design possibilities, or state space, we cal) 'computational design'. the impact of the altered relationship between representatíona! and perceptua! space.
Thís is not to say that computation in its literal sense of calcufation and use of algorithm These are questions inherited from mathematics and, in particular, late 19th-,
was ever absent from design process. Nor that our ready access to these heaving spaces early 20th-century philosophy of mathematics and the developmental psychology of
that calí on 19th- and 20th-century geometry to start to comprehend them, hás not mathematicsand space.The changes in geometryduring the 19th and early20th century
been greatly expanded by access to programmable machines based on the binary logic of were characterised by the increasing power of analytic representations, larger, more
Turing/s metaphorical machine for computatíon. general frames of reference for geometric space, and the rise and fali of an ambition to
bring the whole of mathematics withín a framework of logic. DÍscourse about whether
Computational Design Model Space Hidden from Our Eyes Cand if só, how) intuition could be dispensed with from both the foundations and the
Computation, in the electronícally fadlitated sense, hás extended the geometry of practice of mathematics was centrai to satellite questions about the continuing role
architectural representation into model spaces that no longer map to perceptual space of synthetic geometry, the value of the visual and visualisable, the cognitive nature of
in the way that the simple static Euclidean space, or even the projectíve space of the original ideation for mathematical problems. Although the rise of computational design
model once díd. The design computational model space is potentially invisible, seen only modellíng, and even Íts systems' theoretical underplnning carne after the project of
through its instances, orthe manifestations of particular í-fc[/ecto/-/e5 through thespace. It subsuming ali of mathematics within iogic had foundered, it was nevertheless bom into
is these traces that are seen, not the model ítself, whlch must be understood through more the unresolved tussle between intuition and logic.

168 Computational Design Thinking 169 PhUosophy of Mathematícs for Computationat Design
Geometry and Architecture also appear to be ordered in definite relatíonships: the form of sense-intuitíon. This form
Both geometry and architecture have the power to express and organise space by usit is not grounded in sensaíion; it is what he calls 'pure intuition'.17 Space and time are not
representations outsíde the constraints of a direct mapplng to the physical. The principal actual things, but are what make existence possibie for spado-temporal things. Things'
distínction between them lies in their leveis of abstraction and generalíty. Geometry, 35 presuppose space and time, nor are space and time mere relationships between things/
a pursuit, looks for the greatest generalities8 and, once establíshed (demonstrated or for spatial and temporal relationships exíst only where space and time are presupposed.18
proved\ offers them up as truths Cand for use). Architecture very selectively employs these Kant overturns both Newton's theory of absoluíe space (the space of the objective
general relationships constructively to underpin and create spedfic spatial relatlonships. It observer that exists independently of the thíngs in it) and Leibniz's theory of space and
is concretelyport/cL//0/-.9 During the 19th century, mathematicians arrived to a potentially time based on the relationships between things. But whíle space and time are, in Kanfs
intinite hierarchy of different geometries, spaces with different geometrical truths. Manuel construction, the foundations which make it possible for a world of the senses to be/
DeLanda, in his essay,10 describes in more detail the progressíon from Cari Friedrich mathematics (based on ypure intuition') remalns the basís for its assessment. Moving
Gauss/s application of differential geometry to consistent intrinsic description ofsurfoce smoothlyfrom space to the propositions of geometry as though there were little distinction
to the extension of such intrinsic description from the two-dimensionai space of surface between the two, he wrote: 'Ceometry is a science which determines the properties of
to n-dimensional space by Bemhard RÍemann. Gauss is also credited with developing, space syntheticaliy19 yet a priori20/21
although not publishing, partsofthe mathematics ofgroup theory.^ In 1872, Felix Klein Objects and systems, from the Kantian viewpoint, are not só much constructed by use
used group theory to tackle the problem of categorising and characterising the multiple of geometry; geometry is the necessary Íntuitive context m which objects and systems
different geometries in his infíuential manifesto known as the Erlangen Program.13 A can be conceived.
group is a set closed under a binary operation satisfying three particular axioms.13 But Kant also wrote that: 'geometrical principies are always apodictic, that is, united
From the mid-1870s to the mid-1880s, Georg Cantor was also developing the basis with the consdousness of their necessity, as, "Space hás only three dimensions^/22
of set theory. Up until this time the concept of /sef had been considered a simple one Were such truths empirically derived it would only be possibly to say, 'só far as hitherto
pertaining to fínite sets and Ít had been in implicit use dating back to the Creeks. Cantor observed, nospace hás been found which hás morethan three dimensions/.23This implies
moved infiníty from its place as a philosophical concept into the sphere of concrete Kanfs adherence to the three dimensions of space as necessary truth, for, in order for
mathematical relations. In showing that some Ínfinities are larger than others, some are other types of space to exist, geometrical knowledge would have to be a posteriori, or, in
bounded, the distinction between countable and uncountable sets, he established set other words, empirical.
theory as a foundational theory of modern mathematics. To appreciate the complexities Whether Kant would have seen the truths of hyperbolic and RÍemannian geometry
of setting the criterion that determines the members of a set, we have only to dwell for a as no less self-apparent and Ínherently true than other geometrical understandings were
few moments on Bertrand Russell/s paradox, presented in challenge to the set theory of they offered to him/ and whether higher dimensional spaces would belong, for him, to the
Richard Dedekind and Gottlob Frege in 1901: the set that contains exactiy the sets that analytic ratherthan synthetíc are questions for speculation or philosophical interpretation.
orenofmemí)ersofftíemse/ves.14MostoftheworkofthecomputationaldesÍgnerinvolves Design,whetherofarchitecture,graphicsorpoliticalsystems, unlikepuremathematics,
explidtly defíning and editing the definitíon of seis. Each varíable is a set Cpossibly an must remain concretely grounded and maintain a direct mapping between Íts models
infinítesetsuch asthesetofall positive real numbers ora random realvaluebetweenOand and their Ímagined realisation in the world. In the barely mediated leap back and forth
1); a list or an array is a set. Computational designers define sets and subsets, intersections between the imagined 'reaF Csignified?) and model context, ít is difficult to imagine the
and unions of sets without necessarily being able to conceptualise the members or the full absenceofintuition.Thecomputational design model may perform synthetically(provide
formal implications of their refationships. Rather the set defínitions are the translation of design solutions) and analytícally Cperformance of evaluations). The underlying fogic of
linguisticallyframed design Íntentions. the model comes from the analysis of the context but its implementation seems wholly
synthetic and largely intuitive. However, computational design is not a quest fortruth.
Space and Intuition
In his Critique of Pure Reason, 1781, Immanuel Kant gave us the proposltion that our Pure Intuition and 19th- and 20th-Century Geometry
minds possess, independently of experience, the forms of space and time. He called At the beginníng of the 20th century the renowned mathematician Henri Poincare
these forms intuition^5 suggesting that the intellect does not só much dlscover laws from turned to the issues raisect above with Ínterpreting Kants synthetic a priori ÍntuitÍon in
nature as impose its laws on nature.16 His analysis of knowledge from sensible intuition its application to geometry, to which geometry and how. In Science ana Hypothesis he
led to a distinction between what he termed the matter and the form of sense-intuition. writes that were the geometrical axíoms synthetic a priori Íntuitions/ as Kant affírmed,
Objects affectíng our senses glve us the matter of sensible intuition but these objects there would be no non-Eudidean geometry.24 He wrote that the geometrical axíoms were

170 ComputationalDesignThinking 171 Phüosophy of Nlathematics for Computational Design

neither synthetic a priori intuitions nor experimental facts. They were conventions. Ong While Russell for a while endorsed the Ídealist traditíon in which sensibility had
geometry could not be more true than another, it could only be more conveniente a dominating role, the foundational research of Karl WeÍerstrass, Dedekind/ Cantor
However, Poincaré sat on the fence in the intuitionism-logicism debate. Subsequt and their followers led in Russell/s writing to /the collapse of the ídea that analysís

in Sdence and Method, he dismantles the logidsts' assertion that al) mathematical truths can must rely on the imagination in arder to deal with the basic space-time notions of

be demonstrated using the principies of loglc without making o fresh appea! to intuition.~^ jnfinity and continuit/.32
The logicísts were those who worked to reconcile the whole of mathematics with logic, Poincaréalsopickeduponthedistinction between intuition inthechoiceofaxiomatic

notably Dedekínd and Frege, in the case of Poíncaré's argument specifically RÍchard Courant context and intuition in the process of inference. For him there are fresh appeals to

and GÍuseppe Peano. They were concerned with the primacy of deductive reasoning. jntuition in the mathematidan/s ongoing work.
5o were the lógica! positivists who followed much later in the early application of In the case of mathematical discovery, we shall see how closely mathematical and

computational design to architecture in the 1960s. By his own subsequent analysis, design thought processes can be seen to align. Computational design uses /off the shelf
Christopher Alexander'5 Notes on the Synthesis ofForm, 1964, while Ít raised the value of geometrical relationships but can make use of, and reconfigurethem, only byfresh recourse

systematic contextual analysis, faíled ultimately to proffer any real connection to formal to intuition. How does such spatial Íntuition operate except through visualisation?
deslgn solutions from the analysis advocated in his design method.27
Bertrand Russell famously thought that his own logicism confíicted with Kanfs Minds of Mathematídans and Computational Designers
philosophy of mathematics although none of his own writing, at least up until 1912, In Jacques HadamarcTs book The Mathematidanï Mina, Poincaré appears once more as a

refuted the claim that mathematics is synthetíc Cthe apparently obvious conflict, as strict protagonist. He provides an example of a problem that had occupied him consdously for
logícism would imply that it is analyticY Russell held two doctrines simultaneously known a fortnightwithout success before one night drinking black coffee, 'Ideas rose in crowds;

as standard and condltiona! logidsm. Mathematical theoríes for which there appeared to l felt them collide until pairs interlocked, só to speak, making a stable combination/ He
be no altemative (ie, arithmetic) were to be reduced to logic in the standard sense; those then went away on a geological expedition for some days and in a moment of certainty

for which there were several legitimate alternatives Ceg, geometr(Íes)) were to be reduced in the instant he boarded a bus, 'the idea carne to me, without anything in my former
to logic only in the conditional sense.2BThis last sometimes known as '[f-thenisirT focuses thoughts seeming to have paved the way for Ít, that the transformations l had used
not on whether one or other confíicting set of axioms is tme but on the truth of the to define Fuchsian functíons were identical with those of non-Euclidean geometry ... It
implications from a particular sei of axioms. In other words, the mathematidanrsjob is to seems, in such cases, that one is present at his own unconscious work, made partially
study ínference of certain given truths. perceptible to the over-exdted consciousness ../33 This process, most associated in
Só the confíict between Russell and Kant tums on a more subtle point of interpretation Hadamard'5 book with those of rare talent - Mozart, Gauss, Helmholtz, for exgmple - is

than whether mathematics is synthetic or analytíc. One Ínterpretation is that the role of a form of unconscious thought in which many combinations of Ídeas are shaken together

Íntuition for Kant would be restricted to the mathemaíícal contexf (the basíc axioms of an Cthe etymological root of cogn/fo) until a chance flt occurs and the result is thrown dose

intemally consistent geometry, for example). Once chosen, theorems could result from the to the surface of 'fringe' consdousness.
axioms by formal logícal deduction. But Íf the role of intuition is not limited in this way and This idea of mathematical clarity arising out of crowded, murky and disorderly

Íntuition, not formal logíc, directs the whole of geometric reasoning, then even the result space is a familiar one. John Stillwell wrote that: 'Rigor and preclsion are necessary

of a geometric construction for a proof would be determíned not by the defínition of the for communication of mathematics to the public but they are only the last stage in
context but by spatial intuition.39 It is this aspect of the Kantian view that mathematical the mathematícian/s own thought. New ideas generally emerge from confusion and
reasoning is not strictly formal, but always uses intuition, the a priori knowledge of space obscurity, só they cannot be grasped predsely untii they have fírst been grasped

and time, that Russell refutes unequivocally, based on progress in symbolic logic.30 vaguely and even inconsistently/
A very concrete objection to a priori intuitions providing methods of reasoning This intuitive search wíthin a very ill-defíned space -the outcomes only acquiring their
iogical analytical structure in the proofs, which come much later - makes mathematical

l
that could not be substituted by formal logic/ was the necessity of the figure, real or
imagined to ali geometrical proofs.31 It can be argued that the same questíon hás hung discovery appear indistinguishable from design process in ali but its aims. Thus although
over computational desígn modelling. To what extent can a real or imagíned fíguratíve mathematics is ultimately a search for truth or at least consistency, and design a search

engagement with the computational design model be sacrificed in order to embrace the for form (whether of space or object), the cognitive journeys are not necessarily very

greater sophistication of relationships within the representation? The figurative sampling distinct. In computational design, the searching wíthin obscurity may be for the form

of the model remains at some levei central to its value in communicating but is it equally of the computational deslgn model system. Directed mechanised computation then

central to designers' own desígnerly ways of knowíng the space in which they work? underplns the search for the form of the designed space itself.

172 Computationat Design Thinking lï3 Phitosophy of IVIathematics for Computational Design
Geometrical, Visual and Tactile Space
Strangely Poincaré segregated and defined for us the contrasting nature of fírst,
geométrica! space,35 second, visual, and third, tactile space, giving very physical,
embodied but mechanistic descriptions ofthe lattertwo, while Emst Mach, who devoted
a whole book to quite a similar exercise wrote that: 'No sharp line can be drawn between
the instinctive, the technical, and the scientifíc acquisition of geometric notions',36 Jean
pjaget, In critique of both Kant and PoÍncaré, and whlle stating the evldence for a chilcTs
early topological rather than metrical perception of space, raises an intriguing aspect of
the relationship between representational and perceptual space. During "the development
of representational space, representational activity is, in a manner of speaking, refíected
or projected back on to perceptual actívíty'.37 In other words the perceptual space is not
the space of untouched sensory inputs but shaped progressively also by representational
input. Once the Cartesian axial representation of space hás been assimilated and
superimposed onthespacewe perceive, ourspatial perception is changed ever after. This
'discovery' in experimental psychofogy is paradoxically dose to Kantrs own statement
of our imposition of our laws on nature. In design for realisation in physícal space, the
designer strives to recreate in imagination and in representation the fui! sensory as well
as abstract ordering of the space they are crafting. There is a reiationship between the
visual and more abstract geometrical understandlng in oscillating between the space of
constructing the representation, a potentiallyturbulent ocean ofdifferentstates, and the
imagined experience of possible realisations in physical space. But we can hardly expect
Ít to be a continuous or consistent one.
Logidsm, while it won many battles, lost the war in mathematics. Perhaps it can
be said to have had lived until fínding Íts nadir in GodeFs incompleteness theorems in
1931. But it took other blows, in particular Brouwer/s 1923 essay y0n the Significance
of the Principie of the Exduded Middle in Mathematics, Especially m Function Theory',
that questioned the bilateral assumption Ínherent in proofs based on the impossibility or
absurdityofthe impossibility ofa propertyofa system,while perhaps neglecting a 'third
wa/. On the other hand, unconditional adherents to Kanfs doctrine of a priori synthetic
intuition In the aesthetic must also now be rare.
In more recent decades, computational designers hgve followed their own logidst
quests, inthe beliefthatthewell-structured logical modelwill bringconsistency/overcome
the vagaries of geometrical and computational surprise within the model space, or resolve
the numerous interrelated conditions bearing on the design.
Jane Burry, Lorenz Attractor created using 'Chaoscope' (www.chaoscope.org), RtvIIT, 2010.
3D strange attractors rendering software offers an experience ofthevolatility ofdynamic system
spaces; spaces similar to those first identifíed by Poincaié in the 1880s as nonpenodic but not Conclusion
mcreasmg or approachmg a fíxed point, These images based on a simple equation created
by Edward Lorenz reveal local regions of intensity that exínbit briefly only m multiple short,
Computational design hás sponsored a shíft in design representatlon and thinking from
parameter variable trajec-tories fi-om these mtensities and vast voids of near inactivity for much object models to dynamic system models, from the visible to abstract mappings. Dynamíc
ofthe space, The sense ofDeleuze's multiplicities becomes more palpable m the mteraction and
system computational design model spaces.exhibit geometrical phenomena that require
contmuous transition belween states. Few computational design modellmg experiences offer
the same mtuitive immediacy m sensing the form ofthe model or state space - nor could they, the geometrical developments of the 19th ancf 20th centuries to understand them. The
given the heterogeneity and potenüalty vast nmnbers ofvariables definmg the space - but Üieir
19th-century developments in geometry and mathematics - non-Euclidean geometries,
meanmgful operational states or useful áreas of intensity may be equaüy localised. © Jane Burry.
the classiflcation ofgeometries using grouptheory, newset-theoretical basisforinfínities,

174 ComputationatDesignThinking 175 Philosophy of Mathematics for ComputationaL Design

9 LionelMarch, 'Mathemaücs and Architecture since 1960', inNexusJV; KunWiIliams (ed), I<im
topology as the larger geometrical context - led to phiiosophical debate on the subjects of Williams Books (Italy), 2002, p 9.
intuition and logic with regarei to space. The nonvisual geometrical and spatial character \Q Manuel DeLanda, 'The Mathematics ofthe Virtual: Manifolds, Vectors Fields and Transformation
Groups' in Intensive Science and Virtual Pïúlosophy, Contmuum (NewYork; London), 2002, pp
of computational design models raises similar questions about the status of intuition and
11-12.
the figurative versus deductive logical reasoning within these spaces, in particular for 11 Group theory, sometimes also caüed tïie tïieory of symmetry, is attributed to the published papers
designerly ways of knowíng. of the young mathematician Èvariste Galais as a student before his untimely death at the age of 20
m 1831.
Are the arguments from philosophy of mathematics applicable to computational
12 Só named because it was delivered at Erlangen.
design? Mathematics and design are two different things. One is concemed with 13 To unpackthis: closure intïiis case, means thatthe result of a particular operation on two members
generalitíes and constructing proofs, the other with particulars and constmcting ofthe set (Üiis two beingthe 'binary') gives a resultthatis also a member ofthe set. Thethree
axioms relate to associativity (the arder in which the operation is applied to the elements does
solutions, though both are driven by aesthetics. IVIathematical proofs are constructed
not matíer), the identity element (there is one element in the set which operated on with any other
logicaily applying deductive reasoning but often rest at base on axioms with an intuitive element results m no change to the other element) and the inverse element (for each element
foundation. Computational design models also use logic and deductive reasoning but there is an inverse element which combined with the first results in the identity element). For
instance, the group with the set whose elements are aU translations in the plane, rotations about
the particular form the design description takes among infínite possibilities is intuitívely
a point and reflections m a line (transformations that leave the dimensions and angles in a rigid
based. Design, itself, and mathematical discovery, however, are portrayed as similarly figure unaltered) with the operation meaning 'foüowed byl defines Euclidean geometry; it is the

obscure and largelysubconscious combinatorial processes. Euclidean group. This group hás subgroups - the first mcluding direct isometries that preserve
orientation (translations, rotations) and the second, indirect isometries that reverse orientation
While the figure hás been relinquished from contemporary mathematical spaces,
(reflections), The identity element is the translation (0,0). Every element hás an inverse element,
seemingly far removed from possible visualisation/ computational spatial design practice for instance, the inverse ofthe rotation 90° about a point is the rotation -90° about the same point,

leans heavily on the figurative throughout even its most abstract modelling processes, whichis also anelement ofthe group. DeLanda provides afullerexplanationoftherelationship to
geometries in hierarchy according to their symmetry m 'The Mathematics of the Virtual: Manifolds,
always maintaíning its pathways back into the physical world of appearances, at least Vectors Fields and Transformation Groups', Intensive Science and Virtual PMosopïiy, Contmuum
partiallythrough imagÍnation.The intuitionism-logicism debate in spaceand mathematics (NewYork;London), 2002, pp 11-12.

sheds at least a pencil of torchlight into the foundational nature of the perceptual and 14 Also identifíed although not published by Emst Zermelo in 1900.
15 'Intuition' here is theEnglishíranslation ofAnscAaumg, which m contemporary dictionary
geometrical challenges of bridging between logical stmcture and concrete mappings translation gives 'opmioTi', or even according to one's own 'experience', Yet Leonard Nelson
within computational desígn space. affirms that intuition in this context means 'not írom tiiought' , These are ideas that we believe,
opmions that we hold, views that we have, that, according to Richard Robmson, we cannot help but
Notes hold, They come not from thought but from some more fundamental source. (WH Walsh, Kants
l Gaston Bachelard, Tbe Poetics ofSpace, translatedby Maria Jolis, 1969 edition, Beacon Press Cnticism ofMeíaphysics, p 11.)
(Boston), 1964 (trans), p xxx, 16 MorrisKline, M.atïiematics asdíheSearchforKnowledge, Oxford University Press (NewYork;
2 Donald A Schõn and Glenn Wiggins, Department of Urban Studies and Plarming, Cambridge, MA, Oxford), 1985, p 16.
'ICindsofSeeingandtheirFunctionsinDesignmg', in Design Studies, 1992, pp 135-56. 17 Leonard Nelson; Julius Kraft (eds), trans Humphrey Palmer, Progress and. Regress in Pïiilosopïiy,
3 In the early days of the Joint Center for Urban Studies at Harvard and MIT, and the Centre voH,BasilBlackweU(Oxford), 1971.p 118.
for Land Use and Built Form studies (LUBFs) at Cambridge, UK, despite a culture of scienüfïc 18 Ibid, pp 119-23.
analytical methodology m bringing computation mto design problem-solving that had no 19 Synthetic statements arethose that mtroducenewmformationintheprediGate. Intheir antonym,
time for the consideration of 'appearances' m architectural design, Lionel March hás written analytic statements, the predicate can be derived from the concept alone. For instance, 'the model
that his primary mterest and motivation was aesthetic. (üonel March, 'Modem Movement to is morphogenetic' is synthetic as the idea of morphogenetic cannot be derived analytically from
Vitruvius; Themes ofEducation andResearch', Royal Institute ofBntishArchitects}oum.alQ\, the idea model but 'the shortest distance between two points is a straight line' is only synthetic as
1972, pp 101-9) (Sean Blair Keller, 'SystemsAesthetics: ArchitecturaiTheory attheUniversity long as a straight line is in fact distinct from the shortest distance between two points. If 'is a straight
ofCambridge, 1960-75', in Architecture, Landscape Arcïútecture, and. Urban Planníng, Harvard line' can be deduced from 'the shortest distance', it is analytic. That the geometriGal understandmg
University (Cambridge, MA), 2005, p 172.) For one reflection on the nature of aesthetics in ofspace is synthetic, ratherthan analytic, is significant m separating the foundations of mathematics
mathematics. see Godfrey Harold Hardy, A Mathemaúcian 'sApology, Cambridge University from the process of logical deduction. Analytic judgements can be formed from concepts alone;
Press (London), 1940. synthetic ones cannot.
4 NigelCross, 'Designerly Ways ofKnowmg', in Design Studíes, vol 3 no 4, 1982. p 221-227. 20 A posteriori knowledge comes from experience and is based on the perception of some object or
5 Ibid. phenomenon. A priori knowledge is independent of experience, and só is not based on perception
6 Dsid. at ali. Thus synthetic a priorí intuition, the nature, for Kant, of spatial knowledge and the foundations
7 Felix Klein's 1872 research programme and manifesto under the title 'Vergleichende of geometry, is knowledge which cannot be logicaUy deduced, is not derived from empirical
Betrachtungen uber neuere geometrische Forschungen' proposing a coherent comparative expenence, and is not from thought.
system to categorise ali different geometries on the basis oftheir degree ofsymmetry using group 21 Immanuel Kant. trans Norman Kemp Smith, Cntique ofPure Reason, Macmíllan (London;
theory. It was written whüe he was at Erlangen. Basmgstoke), 1970, first edition 1929, fírstpublished (German) 1781, p 70.
8 It is generally considered poor mathematical practice to prove a theorem for one class when the 22 Kant, transJMD Meiklejohn, Critique ofPure Reason, Bamés andNobIe Books (NewYork), 2004
same applies to a much broader categoi-y, for instance a property of certain types of quadratic (1781),p5.
equation when the same tniüi applies to au quadratic equations, 23 Ibid, p 69.

176 Computational Design Thinking lïï Philosophy of Mathematics for ComputationaL Design
24 HenriPomcaré, Science and Hypothesis, Dover Publications (NewYork), 1952 (1905 firstEnglish
translation), p 48. Assocíatíve Design
25 HenriPoincaré, trans Francis MaiÜand, Science et métïiode: Science andMethod, Dover
Publications (NewYork), 1952 (1908) (1914 mEnglish), p 50. From Type to Population
26 Ibid,p 149.
27 Christopher Alexander, Notes on ïhe Synthesís ofForm, Harvard University Press (Carabridge, peterTrummer
MA), 1964, Alexander's preface to the 1971 paperback edition already refutes his own
mathematical methods and caUs for a more natural metliod based on designer experience.
28 Alberto Coffa, 'Russeü andKant', m Syntíiese (1981), p 252,
Computation, specifíed by parametric means, defines processes by their
29 Evert Wiüem Beth, TíieFowidations ofMatïiematics,~NorïhHoÏ[3nd{!\msteTdam), 1959,p57. degrees of variation, or freedom, and the associations between the
30 BertrandRusseU, The Principies of Mathemaücs, WWNorton (NewYork), 1937 (1903), p 4. parameters. In this text, PeterTrummer elaborates upon this computational
31 Ibid, pp 456-7,
32 Coffa, 'Russell and Kanf, p 254. principie, arguing for materiality as the fundamental premise for variational
33 Jacques Hadamard, The Matíiematician 's Mina: tïie Psychology oflnvenfion in íïie Matïiematical and relational processes. Trummer succinctly traces the possibility for such
Field, Princeton University Press (Princeton, NJ), 1996 (1945), p 14.
an approach through a histórica! trajectory, in both science and architecture,
34 John Stülweïl, Numbers and Geomeïry, SAxler, FWGehrmg, andKARÍbet (eds), Undergraduate
TextsmMathematics, Springer (NewYork), 1998, p 65,
from type-based thinking to a population-based approach. This shift restates
35 He defined it as continuous, ínfínite, ofthree dimensions, homogeneous and isotropic. In this the character of a system from a presumptive descri ption based upon a single
exercise he constraíned himselfto homogeneous three-dimensional space although he is no
type to a characterisation bom of a multitude of variants. This funrtions,
adherent of Kant s necessityofthree áimensions m space. Poincaré is the father of chãos theory.
He found, in the l 880s, while studymg the three-body problem, that Ihere are orbits which are as Trummer states, in the same fashion as a morphogenetic process, where
nonperiodic and yet not ever increasmg or approaching a fixed pomt. the 'possibilities' inherent in the genotype are given specificity through
36 Ernst Mach, Space and Geometry, Open Court Publishmg (Chicago), 1960 (1906), p 69,
internai reaction induced by interaction with environment. While variation is
37 JeanPiaget and Barbei Inhelder, trans FJLangdonandJLLunzer, The Child's Conception ofSpace,
Routledge & KeganPaul (London: NewYork), 1956,p 4, a key component in this, it is only realised as a necessitated repercussion to
the dynamic nature of context (environments). Trummer presents, through
© 2011 John Wiley & Sons Ltd. this argument, a series of concepts by which information, elaborated as
constraints, can be instituted within a computational process, and executed
as a strategy for defíning the whole of a functioning system.

The assumptions of popuiatíon thmking are diametricaíly opposed to those of

the typologist. The popuíationist stresses the uniqueness of everything m the
organic world. What is true for the human spedes, that no two individuais are
aiike, is equalty true for ali other species of animais and piants ... ali organisms
ana organic phenomena are composed of unique features ana can be described
collectiveíy only in statistica! terms. Individuais, or any kind of organic entities,
form populations ofwhich we can determine the arithmetsc mean ana the statistícs
of vahation. Averages are merely statistica! abstractions, oniy the individuais of
whích the popuíation are composed have reaiíty. The ultimate conclusions of
the populatíon thinker ana of the typoiogist are predseíy the opposite. For the
typoiogist, the type (eidos) is real and the vanation an iilusion, whiie for the
popuiationist the type (average) is an abstraction ana oníy the variation is real.
No two ways of lookíng at nature could be more different
(Ernst Mayr)

tf we were to reptace the essences as the explanation ofthe identity of matéria!

objects and natural kinds we need to spedfy the wqy in which multiplicíties relate
to the physicai processes which generate those material objects and kinds.
(Manuel DeLanda)

178 ComputationatDesignThinking 1Ï9 Associative Design

The Encydopedia of Computers ana Computer HÍstory defines computer-aided desígn as matter, on the one side, and morph, meaning form, on the other. Só when we design a
programs that are based on objects. Such objects ycan be straight geometric constructs, table by means of the hylomorphic model, we take a form Çmorph) - the Ímage of the
polygons, b-splines, cubes or they can be complex high levei objects such as bathtubs, table we would like to design - and press it ínto the wood ?y/e) - the material by which
semiconductor drcuits, and car engines'.1 Today we call them geometrical or architectural the image should be realised. The result is a copy, a representation of what we ímagined
primitives. The Oxford Dictíonary of Computíng adds to the above that there is an the table shoufd look like.
individual; an author needed who, equipped with technical knowledge, puts ali these What l wlll argue in this essay is that the real potential of computational techniques
primitives together só that they make sense as a digital design system. is to overcome any Ídea of typological thinking and to come up with a thesis of a deslgn
This definition, being based on geometrical objects on the one side and the requirement practice based on morphogenetic processes. Such morphogenetic processes would
for an author on the other, provoked the discipline of architecture to develop a series of require an understandíng ofthe architectural object as population ratherthan as a type.
theoretical frameworks that have become essential for toda/s architectural practice. They would move away from a hylomorphic schema towards what Deleuze called a
From William Mitchell we have leamed that these geometrical or architectural inachínicphyium.To develop this argument l wantto introducethe Deleuzian idea ofthe
primitives can be defíned as parametric shape grammars, a grammar of geometrical muítiplidty presenteei as an altemative to an idea of essentiafism.
forms based on métrica! proportions. Applied to the aesthetic rules of Palladio/s villas, Inthissense, l will argue inthe essaythatarchitecturalthinking is affected bya much
he demonstrated through a set of algorithms the emergence of completely new forms of wider production of knowledge. Thus l wiil show how ideas of other disciplines could
variations.2 Bernhard Cache made us understand the importance of Eudidean geometry become originated in architecture. In this way, the essay can be called a transdisciplinary
and demonstrated that the discipline of architecture hás to become aware of the fact approach focusing on the question of how the concept of population thinking could
that ali computationa! techniques define variations of geometrical Ínvariants.3 With his become reoriginated within computational design in architecture and, in doing só, what
revisit to geometry as a disciplinary knowledge, he reminded us that every geometrical effect this would have on our design practices.
primitive hás to be understood, as Felix Klein formulated ít in hís Erlanger Program in
1872, as a member of a group defíned by its invariants while belng transformed. Greg Assodative Design Techniques
Lynn introduced to us the understanding of form by means of animation. in his most To develop an architectural practice based on the notion of a machJnic phylum, we have to
explicit example, the embryological house, he carne up wíth topological bodies that are move away from understanding architectural formations as a problem of geometry, We have
based on geometrical primitives that become transformed through a fíeld of forces.4 to shift our attention to the question of how, as DeLanda stated above, physical processes
The intention that of ali these examples basically shared is not only to extend the can start to generate the form of inanimate objects. In this essay l would líke to show how
pure technical deflnition of what is stated under CAD In the dictíonary, but to search this could be done and thought of through means of assodatíve design techniques.
for the /new/ or for novefty through means of computational techniques. But this search The contemporary computational discourse Is contaminated by the tem parametric
for the new was, at the 5ame time, an attempt to overcome the problem of the author. design.7 Parametric design techniques define geometrical or architectural primitives as
It therefore lays in continuation with Peter Eisenman/s critique on Modemism5 and its objects that are variable by theír metrical or dimensional parameters. This suggests that
anthropocentric world. an object hás endfess variations due to its geometric parameters. But the right description
While ali of these attempts were concerned with processing new forms of variation, of these computational applications would be one that defines ali these parametric
they were stiil based on the notion of essentialism. What l mean is that while using techniques as associative design techniques. The word 'associative' focuses not only on
computational techniques to overcome the uniqueness of the object as a masterpiece the variation of objects, but rather on their assemblage: the way these geometric primitives
with Íts particular author, they still defíned an essence for ali its variations. In the case are linked. This linkage becomes suddenly relevant through the way an architectural
of Mitchell, it is hls analogy to typological thinking in language and architecture ofthe assemblage can change. If architectural objects are thought of as assemblies we can start
18th century. In the case of Bernard Cache, ít is the truth of the deductive description to understand them as physical systems, whereby the additíon of parts defines a space
of Greek geometry. And in the case of Greg Lynn, who probably was one of the fírst of possible change. Manuel DeLanda hás given a name to that. He called it the degrees
to argue for the emergence of new forms by means of a morphogenetic process, still of freedom an object hás.8 The degree of freedom defines the possibilities inherent in
rejected the introduction of material consíraínts and therefore idealised the process of an objecfs transformation. As a consequence/ the architectural object is no longer to be
the architectural object itself. understood as a geométrica! construct but as a physical entity.
In ali the above examples essentialism is formulated through a hylomorphic model; Só if we take the example of an architectural object with ali its geometrical or
that is a model that hás a dear directionality of how matter moves into fomn.6 The term architectural primitives, we can say that every primitive defines a particular degree of
'hylemorph' Índicates what is needed to design an object. It derives from hyle, meaning freedom depending on its material or physical constraints. Ali assemblies together would

180 Computational Design Thinking 181 Associative Design

form a whole described by its space of possible changes. These changes are no longer
defined by the identity or essence of an object, but rather would allow us to understand
theobjectaspossibleformsofappearancesthroughtheinteractionofthe populationofits
parts. Thus the architectural object would be understood by means of its morphogenetic
process and defíned as a multíplicity rather than a type.
Hanju Chen, Sebastiano
Mansen/isi, Alessandra
IVlultiplicities Martmelli, MetricalProperties,
Berlage Instilute (Peter
To understand what is meant by the idea of population thinkmg and Íts origination in
Tmmmer), Rotterdam, 2008.
archítecture, it seems to be important to define the idea of multiplicity in this particular The matrbe shows prototypical
context An often used example to define a multiplicity is the definition of a geometrical conditions ofurban block
formations and the specific
manifold. Such a manifold is, for example, a sphere, as demonstrated by the sphere of diversities ofland subdivision.
the globe. Every region of the earth can be described as a planar map, or as it is called in This subdivision of land and its
block SÏZQS can be potentiaily
geometry, a Eudidean space, In which on the edge of one map, the other map begins. 5o
found within the urban fabric
the earth can be desaibed as a manifold defíned by a sei - a many - of maps. These maps of México City. They reach
from perimeter blocks, simüar
define a metrical space on the one hand and Íts assemblies define a manifold, like the
tothe 113/113 metreblock in
earth, that is non-metrical. It cannot be defíned as a whole, but only as an entity that is Barcelona (on the right comer
assembled by a certain amount of measurable planes or maps. It can only be understood below) to block sizes of 200/40
metres (upper right comer),
as a topological sphere which defines one particular orientation of an endosed space but Each node inside the blocks
is able to take on various forms that are embedded within the assemblies of Íts parts. represente the stmctural
startmg point of the vertical
]f we relate this deflnition of a geometrical manifold, or more generally speaking,
agglomerations and defines
the Ídea of a multiplicity, to the above description of an architectura! object based on the centre point ofthe 'conic

associative design techniques, then each of the architectural primitives defines the frames'. © Associative Design
Program/Berlage Institute,
metrical constraint components of an architectural object. These components are for the Rotterdam/Peter Trummer.
architecíural objects what the maps are for the skin of the earth, each being constrained
Hanju Chen, Sebastiano
to very precise measurements. The way these architectural primitives are assembled Manservisi, Alessandra
define St as a manifold. Now the way these assemblies are connected and how they are Martineüi, Non-Metrical
Properties, Berlage Institute
linked affect their appearance. The processes whlch inform their appearance define the
CPeter Tmmmer), Rotterdam,
morphogenesis ofthe population ofarchitectural objects. 2008. The matrix shows the

Deleuze's concept of multiplicities is particularly radica! as he actually applies it to íopological organisation of

vertical joins within each
ali our material realities, the animate emes as well as the inanimate ones. For him ali block configuration, The
"ideasare multiplicities: every idea is a multiplicity ora variet/.9The consequence ofthis diversity ofjoms is based on
the variation of earthquake
thought can be best visualised when it comes to defining a group of things or objects. A
mtensity, mcreasing from top
morphogenetic process accounts that a 'multiplicity must not designate a combination of to bottom (which means that
the conic frames increase).
the many and the one, but rather an organization belonging to the many as such, which
Each line between two nades
hás no need whatsoever of unity in order to form a system'.10 inside the blocks represents
Thísisinstarkcontrasttotypologicalthinking,wheretheessenceofan object defines an angled assembly oftowers.
Each ofthese assemblies
its identity; here, is based on a generative
process offinding after each
interaction neighbouring joms
o species is not defíned by essentia! traits but rather by the morphogenetic
that give the best structural
process that gave ríse to it. [...] While an essentíaíist account of species is performance. © Associative
Design Program/Berlage
basicatty static, a morphogenetic account is mherently dynamic. And while an
Institute, Rotterdam/
essentialist account may re!y on factors that transcend the realm of matter Peter Trummer.

182 Computational Design Thinking 183 Assodative Design

Han Ju Chen, Sebastiano Manservisi, Alessandro MartineHi, Population of Vertical Assemblies, Hanju Chen, Sebastiano Manservisi, Alessandro
Berlage Institute (Peter Trummer), Rotterdam, 2008. The image stiows the emergmg MartmellL Actualised Population, Berlage Institute
population of vertical building agglomerations based on the relationship between their (PeterTrummer), Rotterdam, 2008. Emerging
extensive properties such as the geometrical constraints of structural stabüity and the metrical population of vertical assemblies due to the
properties of block size and their intensive property of earthquake energy based on the interrelationship ofthe intensive and extensive
relaüonship of acceleration (g) and wave periods (s). The population of towers is based on properties of space. Associative Design Program/
the component of angled stmctures that vary in terms of their metrical properties in length Berlage Insti-lute, Rotterdam/Peter Trummer. ©
and angle, and are assembled due to the topological orgamsation ofneighbourmgjoms. Berlage Institute, Rotterdam - Peter TRimmer.
Associative Design Program/Berlage Institute, Rotterdam/Peter Trummer. © Berlage Institute,
Rotterdam - Peter Trummer .

184 Computational Design Thinking 185 Associative Design

ana energy (etemal archetypes, for instance), a morphogenetíc account gets Before Mayr/s groundbreaking concept was introduced, two kinds of typological
nd of ali transcendental factors using exdussveiy form-generatmg resources thoughts had emerged in biology, which described the genesis of individuais. One was called
which are immanent to the material world. (Manuel DeLanda)11 transmutationism, the other transformationísm.17 Both of these concepts were based on a
typological approach, but had slightly different explanations on howtypes change throughout
For Deíeuze, any material reality is based on ideas that do not transcend ever-countir history. The transmutationist believed, as Mayr telis us, that change could only occur by the
properties. In oppositiontothetypoiogicaithought, in whichíhetype resembles ali other origin of new types. 'SÍnce a type (essence) cannot evolve gradually Ctypes are consídered to
members of a species/ a multiplicity is based on the effects produced by the accumulation be constant), a newtypecan originateonlythrough an instantaneous^mutatiorTorso/totíon
of the individual members of the species in relationship to each other. In this sense. of an exísting type, which thereby gives rise to a new class ortype.r18Whilethe transmutationist
we have to understand Deleuzian ontology, as he states in his book Difference and rejected any forms of gradualism or better gradual transformation, the transformationist
Repetition, that 'everything is a multiplicity in só far as it Íncarnates an idea ... instead believed in the idea of evolution and formed the concept of gradual evolution.
of the enormous opposition between the one and the many/ there is only the variety of The discipline of biology distinguishes betweentwo forms oftransformationism. Each ofthem
multiplidties - or in otherwords, difference'.12 had a different understanding of what infíuences changes within a species. One movement of
The important reason for explaining the difference between the idea of the one and transformationism/postulatedthattypes(e5sences)weresteadilyimprovedbyanintrinsicdrive,and
the mony is the fact that the typological variant was thus far the only accepted theory to that evolution was believed to take place not bythe origin of newtypes, but bythetransformation
explain the genesis of architectural forms. The most infíuential and constantly recurring of existing types'.19 It became known in biology as finalism. This kind of transformation due to
example ofthis typological thought is the one defíned by Quatremère de Quincy, given a strive for perfection sounds very familiar to us as archltects, reminding us of Quatremère de
attheend ofthe18th and begínning ofthe19th century, which defines the one to many Quincy, who understood the evolution of primitive huts advandng Ínto complicated constmction
relatíonship as one that is based on the distinction between the type as the idea of an in the same way as primitive society advanced to high forms of civilisation.
object and its model as its application.13This essential type defínition means that every The second group of transformationists are known as followers of the Lamarckian
architectural design or urban plan is the application of a model based on an ever-countíng theory. Evolution is understood as the gradual change of organisms, in which the change
essence cafled the idea or type. We have to keep in mind that this defínition of the type is due to the directinfíuenceoftheenvironmentonthegenetic material. We ali knowthe
versus the model served as the argument for Aldo Rossio work on Ttíe Architecture of example of giraffes reaching to the highest accessible tree branches for food, and giving
the üty, which was and still is today the most dominating architectural theory governing this information to lengthen their neck further to be inherited by the next generation.
practice for the last 50 years. Typological thinking therefore defines its genesls bytranscended essencesthat classify
a species by its morphological characteristics. In contrast, the populationist defines the
PopulationThinking in Biology Ídentity of a species through its morphogenetic process. Thafs why it is radically different
Ernst Mayr introduced the idea of population thinking in biology based on Darwin's On to those views based on typologies. Population thinking not only replaced typological
the Orígin of Specjes by Means of Natural Selection, or the Preservation of Favoured thinking, it basicaliy erased Íts roots.
Roces Ín the Struggle for Life. When he writes that, 'The assumptions of population To understand what population thinking is or how it distinguishes itself from'
thínking are diametrically opposed to those of the typologist^14 and continues that, other species concepts, we can look at the way it defines the relationship between the
'For the typologist, the type Ceidos) is real and the variation an illusion, while for the individual and Íts related species. It does it by two forms of differences. One states that
populationist the type (average) Is an abstraction and only the variation is real',15 each individual hás to be different from the other, but at the same time these differences
he basically replaces one species concept by another. A species concept is used to among individuais sustain its difference as a spedes. The diversity of the individuais is
designate a class or a famlly of similar objects or thlngs to which a name hás been criticai to the entire species. And this is the reason why it is called population thinking. It
given. Most of the time, the term spedes is applied to describe living organisms since Ít needs a criticai mass of different individuais in arder to diversify Íts gene pool.
origínated in biology, but it hás also been used to describe inanimate objects as in the In that sense, a population refers to a multiplicíty as deflned by Deleuze. It does
world of physícs, and even to describe human artefacts.16 not define Ítself as the relationship between the one and the many, it rather defines the
Population thinking became the only accepted species concept within bioíogy. organisation ofthe many itself.
Two aspects are tremendously relevant for the argument presented here. One is the To introduce population thinking into architecture, we haveto understand how biology
difference in defining 3 group of objects or individuais as a species by the typologist or explains the morphogenetic process of this spedes concept, particuiarly in regarás to the
the populationist; the second, perhaps more Ímportant, aspect is how the differences difference between genotype and phenotype.Theterminological distinction between the
between species are understood Ín terms of their morphogenesis. phenotype and the genotype is that the:

186 Computational Design Thinking 187 Associative Design

Human

U! na

Monroe W Strickberger, Skeletal Structures ofForelimbs, 2000. Skeletal structures ofthe

Ernst Haeckel, Embryos, 1870. Haecl<el's classic illustration ofdifferent vertebrate embryos at
ï.mvanous ïePresenïaïlve íeírapods (liferaüy animais with four pods or feet) to show the
comparable stages ofdevelopment. Alti-iough Haeckel took some liberties in drawing these
h.omology among bones at different Prox™odistal leveis m the Iimbs (wher~ep7oximans"nearL
figures, the earlier stages are more similar to one another than later stages are to one another.
ÏlbodLanddistal furthestfrom Üiebody) -homology that is preserveddeïtíe"changes±mul
Each embryo begins with a similar number ofpharyngeal arches, paired eyes and interna]
s^shapemd_function of the forelimbs during the course of evolution. Compare"for'example,
structures such as a smgle midline heart (not shown). Some biologists caU this a phylotypic
LSJ"gJeupper a™ bme/íhe^umerusm whitej m each limb, or the distai digïtsmgreen"Êven stage representing the distinctive developmental substructure of the body plan shared by
when digit number varies, homology of the digits remains (HaU 2008), As Da?m~noted" "What ~
individuais in a particular phylum. In lafôr stages of áevelopment, these and other structures are
.canbe more curiou.s thmthat the hand of ™toed for grasping, -that~ofa~molefor~'diggmg,'
modifíed to yield the charactenstic fomi of each subgroup. The embryos m the different groups
3, the paddle of a poipoise and the wing of a bat, should au be constmcted on
have been scaled to the same approximate size só that comparisons can be made among them.
! mclude simiiar bonés and in fhe same relatíve positions?'
The image md its subtiüe is taken from BK Hall, B Hallgrimssom, Síríckberger's Evoluïion,
Ihejmage andÍtssubtiüe is taken ï!om BK HallJ B HaUgrimssom, Stríc3<berger-s EvoMion, fourth edition, 2008, Jones & BartlettLeammg (Sudbury, MA), www.jbleaming.com. Reprmted
fourthediíÍonl 2008^0nes,&BartIettLeammg (Sudtouiy, MA), w,vwjbÏearTmg.com"Reprmted with permission. ©Jones & Bartlett Leaming,
with permission, © Jones & BarÜett Leaming.

188 Computational Design Thinking 189 Assodative Design

genetic matéria! itself ;s the genome (hapioid) or genotype (diploid), which çchema as it ;sees matter as homogenous and form as fíxed'. As mentíoned above, the
contraís the production of the body of an organism ana ali of Its attributes, the conceptualisation of homogenous matter on the one side, and fixed form on the other,
phenotype. Thss phenotype is the result ofthe Ínteraction ofthe genotype with the defínes the direction of how ideas move Ínto form. But what actually happens, SÍmondon
environment duríng development. The amplitude of varíation of the phenotype continues, is that matter and form define 'a zone of médium and intemediary dimension,
produced by a given genotype under different environmentaí condítions is called ofenergeticand molecular dimensions' between them.27Thls zone can be referred to asa
its norms of reaction, (Ernst Mayr)20 morphogenetic space, in which both matter (tíy/e) and ideas (form) are in dynamic relations.
Based on Simondon/ Deleuze gave thls dynamic condition a name; machinic phylum.28
It is the phenotype which is exposed to natural selection, never the genotype. This In his article The Machinic Phylum',29 Manuel DeLanda unfolds this philosophical
distinction between a mortal body and an Jmmortal transmitter of hereditary instmctions concept and examines its significance in what he calls 'the emergence of noveity", where
is what revolutionised biology.21 matter is seen as an active material with 'morphogenetic capabilities' to generate different
This is what biologists call morphogenetic process. Withín the aeation of mankind structures through the constraints of its material properties. To demonstrate what he
and other vertebrates this is called embryology. It is not just a metaphor of fertilised means by thís, let us take an example he uses in his article. In physics, he explains,
eggs with ali the information given for the development of an organism. It defines the /a population of interacting physical entities, such as molecules, can be constraíned
genesis of its form, the process of an organism unfolding under 'differentiated tissues energetically to force ít to dísplay organised collectlve behaviour'.30 The example he
and organs'.22 This unfolding is called 'progressive differentiatiorT,23 It is this process of often used to explain this behaviour is the fomnation of soap bubbles. He explains that
differentiation, which defines the physical processes of its emerging form, that allows us the molecules of the soap collectively minímise the tension of the surface. Each soap
to think of it as a transdisciplinary concept. bubble díffers according to the way the molecules organise themselves to perform the
most minimal surface tension. Another example can be given by the formatíon of crystal.
Population Thinking in Architecture ;lf you hold a crystal, made out of carbon, Ít could take the shape of a diamond with its
To understand the idea of population thínking in relation to associative design techniques, beautifully regular tetrahedral form. But it could be graphite, whose hexagonal sheets
two important defínitíons are described above. One is that population thinking does sheer off as it is rubbed over paper/31 One substance can occur in many forms.
not define the relationship between the one (fype) and the many (motíe/), but that it Both examples show that such processes of progressive differentiation embedded
belongs to the organisation of the many as such. The many of an architectural object is within a machinic phylum can be defíned by its physicality: the interrelationshíp between
therefore the assembly of its components or geometrical primitives. The second important íts metrical and non-metrical properties. The metrical properties of the soap bubble ar the
statement is that a morphogenetíc process is a process of differentiation. Thls is needed in crystals are given form by the non-metrical properties of the minimal surface tension in
order to understand an assembly of architectural primitives as a dynamical system. the case of the soap or the collectively minimised bonding energy to form crystals. Ali of
Various fíelds of knowledge are dealing with the dynamical potential of interacting which occurred as the collective behavlour of a population of molecules.
systems like population thinking. From economics, sociology, physics and polltics, Ít If we relate thls knowledge to our computational design techniques and the
occurred in geography through the work of the Chicago SchooFs Burgess, Park and organisation of the architectural object as an assembly of its parts, we can say that the
McKenzie24 and theír study of human ecologies in the dty of Chicago. It entered urbanism process of differentíatíon occurs between the constraints of the architectural primitives,
by the centrai place theor/ of Walter Christaller, for whom the formation of densities of and the externai forces that effect the negotiation between them. What the molecules
citíes emerged as the accumulation of individual behaviour based on economic forces. are in physical processes, the geometrical primitives are to our assembled objects. Each of
Ali these examples have in common that, within their discipline, they have their metrical properties of space such as lengths, widths, heights, surfaces and volumes
replaced the hylomorphic model as the dominatíng model that defined the relationship called extensive properties are constrained by their materiality. But to produce a collective
between matter and form.35 behaviour, they have to be confronted or tested wlth the non-metrical properties of our
The hylomorphíc model demes the interrelation between morph (form) on one side build environment. These properties are defíned by degrees of temperature, pressure,
and hyle (matter) on the other side.26 It is GÍlbert SÍmondon, in his book LWvidu et sá tension and capadties called intensive properties.32 These non-metrical properties
génese physico-biotogíque, who introduces the common notion ofthe hylomorphic model affect the way one assembly relates to the other. The metrical material constraint of the
to a wider fíeld of disciplines when he says; 1t is not only the clay and the brick, the marble architectural primitive defines the degrees offreedom ofthe system.
and the statue which can be thought according to the hylemorphic model, but also a The negotiation between the non-metrical propertíes of architectural components
great number of formal genetic and compositional actualitíes, in the living world and the andtheeffectsoriginatingfromtheirassemblydriven bythe non-metrical properties, can
psychic domain/ In the chapter on ïorm on Matter', Simondon criticises the hylomorphíc be called the morphogenetic process from which a fonn emerges.

190 Computational Design Thinking 191 Associative Desígn

Mika Watanabe and Lin Chia-Ying, Intensive Property, Mika Watanabe and Lin Chia-Ying, Interrelationship between Extensive
Institute (Peter Trummer), Rotterdam, 2008. The image shows Properties and Intensive Properties, Berlage Instiíuíe (Peter Trumraer),
the raáiation effect ofthe sou condition ofthe desert section of Rotterdam, 2008, The diagram shows the metncal effect ofthe intensive
Maricopa County, Phoenbc. Due to the vegetation density the sou property of radiation. Depending on the amount of curvature of the build
is either more or less resistant to heat-up. This technique is used structure, the radiation of a building envelope can be reduced due to the
to indicate heat island effects within the urban layout of Phoenix, metrical changes made possible by the chosen material constraints of the
Arizona. Associative Design Program/Berlage Institute. Rotterdam/ building structure. Associative Design Program/Berlage Institute, Rotterdam/
Peter Trummer, © Berlage Institute, Rotterdam - Peter Trummer, Peter Trummer. © Berlage Institute, Rotterdam - Peter Trummer.

192 Computational Design Thinlíing 193 Associative Design

Even though it was not designed by computational techniques but definitejy is
on thinking associatively, an interesting example of thls is the repetitive beam;
delivery zone of Utzon/s Opera House in Sydney.33 These repetitive beams first
strangely curved surfaces. They seem to be a merely aesthetlcally driven cons+rnri.;.
wíthout any particular purpose or constraints. But upon closer examination, we see
these beams perform a kind of geometrical transformation through a metrical
the curvature of the beam is defined by the position of a point moving along a dic
line. In eme region, the section ofthe beamtakesthe form of a square, while in
it becomes a T-shape. The diversíty of these sectíonal shapes of the beam follows
distribution of structural forces, or, in other words, the Íntensive property of space. in this
way, the form of the beam is differentiated by the metrical property of the constraints of
Íts plywood formwork in relation to the non-metrical properties of the force distribution
of the structural diagram. The beam structure behaves as a population whereby
element is dífferent in arder to generate a whole.

Condusion
What l have tried to argue in this essay is the way population thinking can become re-
originated in architecture. l have done that by demonstrating that population thinking
performs as a multiplidty, which redefines the relationship between the one and many as
an organisation of the many. l have further tried to outline what this could mean to our
understanding of computational techniques and how the potential of associative design
techniques supports the Ídea of thinking in morphogenetic processes rather than in types.
In this instance, the morphogenetic process is driven by the constramts and infíuences of
the metrical properties of architectural components and non-metrical properties of spaces.
The aim of this artide is to support a materialíst viewpoint in architecture. It is a
critique on the disciplínary knowledge of architecture with its bías towards typologícal
thinking, which needs to exclude true complexity in order to work. But, as Robert Venturi
in Compiexity ana ContradÍctíon^ stated: 'More is not less.'
More is not a necessity; it is material.

zon. The sculptural concourse beams of the

1 House viewed from below, Syc^ey,
rósTMaxDupam Associates, photograph by Max
Dupam, Oct. 1962. ©MaxDupam.

195 AssociativeDesign
194 CompLrtational Design Thinking
fncyclopedia of Computers and Compuíer History, vol l A-L, Raul Rojas (ed), Fitzroy Dearbom
); London), 2001, p 179.
William MitcheU, 'A Grammar of Paüadian Villa Plans', in The Logic ofArcïiitecture - Design,
Comp^ation, and Cognítion, MIT Press (Cambridge, MA), second edition, 1990, p 152.
^-cïiiïecture of Geornetry - Geometry of ArcMecture, lecture byBemard Cache, altheBeriage
Institute in Rotterdam 2003/04.
ÇregLynn, Ammaíe Forrn, Princeton Architectural Press (NewYork), 1998, p 30.
'IVlisreadmg Peter Eisenman', in Peter Eisenman Aura und ExzeB; Zur Uber/vin.d.ürig der
Metaphysik der Architektur, Uürich Schwarz (ed), Passagen Verlag (Vienna), 1995, p 109,
Thedefmitionofthehylomorphistakenfrom 'Simondon andthePhysico-BiologícalGenesis ofthe
isH^^^l^ Íiiliiiii^ll Individual', Chapter One: 'Formand Matter, Section I - Foundaüons ofthe Hylemorphic Model-
Tecïmology ofthe Capture ofForm', byGiïbert Simondon, Original in French; L'inüividu et sã
vico-bioïogique, Presses Universitaires deFrance CParis), 1964, pp 27-39, original
translation by Taylor Adkins.
Patrik Schumacher, 'Parametricism - A New Global Style for Architecture and Urban Design',
Archítecturaí Design, vol 79, no 4, July/August 2009. pp 14-23 and 'The Parametricist Epoch: Let
the StyleWars Begm', The Ai-cMects' Joumal.. no 16, vol231, 6 May 2010.
Manuel DeLanda, Inïensive Science and Virtual Philosophy. Continuum (NewYork). 2002. p 13.
Guies Deleuze, Difference and Repeíition, Athlone Press (London). 1997, p 182.
Ibid. p 182.
DeLanda, Inïensive Scíence ana Virtual Pïiilosopïiy, p 9.
Deleuze, Difference and Repetiïion, p 182.
See Sylvia Lavin, Ouatremère de Quincy and the Invention ofa Modern Language ofArcïutecíwe,
MIT Press (Cambridge, MA; London), 1992.
Emst Mayr, Popi-dations, Species. ana Evoluíion, Belknap Press of Har/ard University Press
(Cambridge, MA; London), 1963, p 4.
Ibid, p 4.
Ibid, 'Species Concepts andTheirApplication', p 10.
Ernst Mayr, Wïiat Evolutíon Is, Phoemx (London), 2002, p 83,
Ibid,p85.
Ibid, p 89.
Mayr, What EvoMion Is, p 98.
See Brian Goodwin, How íhe Leopard. Cíianged Its Spots, Phoerux (London), 1995, p 27,
DeLanda, Intensive Science and Virtual Pïnlosophy, p 16.
Ibid.p 17.
Robert Park, Emest W Burgess and Roderick D McKenzie, The City, University of Chicago Press
(Chicago), 1925.
See GíUes Deleuze and Felix Guattari, A Thousaiid Pïaíeaus - Capiíalism and. Scïiizopbienia,
JT^ZOnLglomeTofaconcoursebeaml Sydney Opera House, Sydney, 1973. The in
Continuum (NewYork) andAlhlone Press (London), 1987, pp 408-9.
^lÍ^cr°Ï:secÏS ofvarious concourse beamschemes:The^ge(b)'^ov;.Zic
'Simondon and fhe Physico-BioÍogicai Genesis offhe índividuaï', pp 27-39.
^ms^sïlno.fthe?^^ Deleuze and Guattari, A Thousand Plateans, p 409,
d.6mïfrates.&emterrelatlomNpbetweent^^^^
^ S^ÍÏ ^^SS^.^Ï''^!^^ ^^^e',he^^^^ ? Jbid.
Manuel DeLanda, 'The Machinic Phylum', in TecïmoMorphica.. V2 ^Publisher (Amsterdam), 1997.
Ï^amJ,he.Tgedefínes the vanatlon onhe^ctionoftheb^b7mo™1thuep^u
S.a.wÏ.delïd.5POÏC Ibid,
?omt.g^concurrent wlth pomts c the section ^smto-aÏshïrTheÏpeeS byï£^mt Goodwin, How the Leopaïd Cïianged its Spots, p 9.

^:^ ^hlÏnlcfd^ends.on ^^-^^^eS^^X" The defínition of extensivo rind intensive properties of space is taken from Deleuze, Dífference and
^wlan^ïsmtër^^^^ Repetítíon, p 223.
rï!image (c) shows a sectlon of the círand staircase-isometnc view^f ^c^cuoulr^ïm See Fi-ançoise Fromonot. Christopher Thompson, Jórn Utzon - Tíie Sydney Opera House, Electra/

^CTaTeT^aZ^^lrrf°^^^^^
tes ofplywood, Redrawn by Sean Alüquist,
Gingko Press (Corte Madem, CA), 1998,
Robert Venturi, Complexüy ana Conïradiction in Architectnre. Museum of Modem Art (New York],
1966.

© 2011 John Wiley & Sons Ltd.

196 Computational Design Thinking

19Ï Associative Design
Integral Formation and Materialisatíon While these examples of deferred impact refer mainly to advances in material

Computational Form and Material Gestalt hnology/ one can still trace an Ínteresting paraliel with the current employment of
r-aided design and manufacturíng technoiogies. The by-now-ubiquitous use of
Achim Menges technologies in architecture serves, more often than not, as the facilitative
affordable means to indulge in so-called free-form architecture as conceived at the
In this excerpt from the book Manufacturing Material Effects: Rethinkir of the last century. Although this may lead occasionally to innovative structures and
Design and Msking in Architecture (2008), Achim Menges cleark spatial qualities, it is important to recognise that the technology used in this way provides
establishes a system-centric approach to both design and computation a'mereextension ofwell-rehearsedand established design processes.
a thread which binds both the mechanisms of a virtual process with the particularly emblematic is the one-dimensional reference to the notion of digita!
behaviours of physical material processes. In doing só, he pinpoints both an morphogenesi5. By now almost a cliché, this term refers to various processes of form
opportunity and a necessity to capture materialisation within the realm of tion resulting in shapes that remain elusive to material and construction loglcs.
computation. Computation is a system offunctions which operate In foregrounding the geometry of the eventual outcome as the key feature, these
interaction and variation of information. Natural systems, as Monges techniques are quintessentially not dissimilarto more conventional and long-established
describes, operate in a similar fashion reconciling in material structures representational techniques for explidt, scalar geometric descriptions. As these notational
that result from both internai physical constraints and externai infíuences cannot integrate means of materialisation, production and construction, these
and forces. Performative morphologies, in computational design, can be crucial aspects need to be subsequently pursued as top-down englneered material
derived similarly. As IVIenges describes in relation to representational solutions. Being essentially about appearance, digitai morphogenesis dismisses both the
techniques, form of a static nature emerges from processes concerned capacity of computational morphogenesis to encode logic, structure and behaviour, as
primarilywith shape. Form as a performative, dynamicsystem emerges from well as the underlying principies of natural morphogenesis.
principies based on the behaviour of material, methods of manipulation
and assembly, and interaction with environment. In computing material as Integrating Formation and IVIaterialisation
a property in the generation of form/ it becomes the criticai nexus at which Natural morphogenesis, the process of growth and evolutíonary development, generates
performance in spatiality, environmental mediation and structural capadty systems that derive complex articulation, specific gestalt and performative capadty
become issues of novel design, well beyond the mere generation of shape. through the interaction of system-intrinsic material characteristics, as well as externai
stimuli of environmental forces and infíuences. Thus, formation and materialisation
Archítecture, as a material practice, attaíns social cultural and ecological relevance are always ínherently and ínseparably related in natural morphogenesis. Such integral
through the articulatíon of material arrangements and stmctures. Thus, the way we processes of unfolding material gestalt are particulariy striking when one considers
conceptualise these material Ínterventions - and particularly the technology that enables architecture which, as a material practice, is (by contrast) still mainly based on design
thelr construction - presents a fundamental aspect in how we Cre)think architecture. approaches characterised by a hierarchical relationship that prioritises the definitíon and
In many ways, the progress over decades of computer-aided design and manufacturing generation of form over Íts subsequent materialisation. This suggests the latent potential
(ü/\D/CAMX or rather the greater availability and affordabílity of these technologles, can of the technology at stake may unfold from an alternative approach to design, one that
be seen in the lineage of other technícal advancements. In the history of architecture and derives morphological complexity and performative capacity without differentiating
construction, groundbreakíng technologíes have often been employed initíally to facilítate between fonn generation and materialisation processes.
projects that were conceíved - and indeed embraced - through well-established design The underlying logic of computation strongly suggests such an altemative, in which
concepts and construction logícs. There is ample evidence of thís inertia in design thínking in thegeometricrigourandsimulationcapabilityofcomputationalmodellingcanbedeployed
the context oftechnological progress. For example, the design ofthestructure and connection to Íntegrate manufacturing constraints, assembly logics and material characteristícs in
detaíls of the first cast-iron bridges of late ISth-century England were modelled on timber the definítion of material and construction systems. Furthermore, the development of
constructions. SimÍlarly, the early reínforcËd-concrete structures of the late 19th century versatile anafysis tools for structure, thermodynamics, light and acoustics provídes for
mimicked previous iron and steel frame buíldings. In fact, almost half a century had to pass Íntegrating feedback loops of evaluating the system/s behaviour in interaction with a
between the first patent for reinforced concrete and its signifícant infíuence on design through simulated environmentas generative drivers in the design process. Far beyond the aptitude
the conceptualisation ofits innate material capacitles as manifested in Robert Maillarfs bridges of representational digital models, which mainly focus on geometry, such computational
and the shell structures ofvarious 20th-century pioneers such as Franz Dischínger. models describe behaviour rather than shape. This enables the designer to conceíve of

198 Computational Design Thinking 199 Integral Formation and IVlateriatisation

Nico Remhardt 3D Gewirke-Verbmid, Umversity ofArt and Design Offenbach (Achim
Menges\offènbach; Germany- 2006-7. The '3D Gewirke-Verbtmd' projectmvestigated
ways ofutilising local form-fïndmg processes to differentiate a larger^ contmuous'material
system, The specifíc manipiüation-component (Figure l) defines the vectors anci distances of
gathermgpariicuiar points on a three-dlmensional spacer textile. NumeroLis computational'
andphysical expenments were conducted explonng the behaviour of local manipulatïon'
^eas.[ mlerdePendent manipuiaíion arrays and the resultmg overaü morpholog/(Fi^ire 2,
Figure 3), Based on the resultant computaüonaUy denved protocols, a number of füU-scale
prototypes were constmcted (Figure 4). © Achim Menges.

200 Computationat Oesign Thinking 201 Integral Fonnation and Materialisation

material and constmction systems as the synergetic result of computatir of ali, the geometric description of material systems, or rather the notation
and instrumentalising the system/s intrinsic loglcs and constraints of makinc|, the^ltin9 rticular features of the system'5 morphology needs to be established The
behaviour and interaction with externai forces and environmental influences ^ Ï,IT1/S rir definition of the system hás to overcome the primacy ofshape and related
the performatíve effects resulting from these interactions. Thus, the understflL^1 a! ^'rïc^descriptive characteristics. Therefore, the designer hás to facilitate the setup
material effects [...] extends far beyond the visible effect towards the thermolri^'^of Ka'c*omputational model, not as a particular gestalt specified througha numbe[- of
acoustic and luminous modulatíon of the Cbuilt) environment As these mnH,,i..al71ic/ 'Jrdmates and dimensions, but as a framework for possible formations affordjng
in relation to the material Ínterventions and their constructíon process, carinn °nLS' Ï,,rther differentiation that remains coherent with the behaviour observed ana
antidpated as actual behaviour rather than textbook principies, thedesign ofsl/l!e ,cted from physical experiments and exptorations of the relevant system. This
structure and climate becomes inseparable. —a'i ui spg^ ^putational framework, which essentially constitutes an openi model^ ^eferred to
Crossing a number of discíplinary boundaries, the design approach presented ^'ïa~mework/ here due to the ambiguous meaning of /model/ m a design contexa
demands that structural and environmental engineering, which hás tendedtoihp>! ^"then step-by-step informed by a series of additional parameters, restnctlons
questíon of post-dËSign optimisatíon, becomes an essential factor in the setup of th? and characteristics Ínferred from material, fabrication and assembiy logics and
?!!!g[Lp.rocess ítsel^Therefore realislng the potential of computatíonal design' ^onstraints. Principally, this includes the specific material and geometric behaviour
^Tp^er_cortrolled fabrication is twofold: tirst, it enables (re)establishing a f ar more ^"nfonnative processes, the size and shape constraints of involved^machinery, the
!m-Tedi?te relation to,the proce5ses of makin9 and ronstructíng by unfoidlng"inna^ rocedural logistics of assembly, and the sequences of construction. As these aspects
material capadty and behaviour, and, second, understanding this behaviour as a means ^ary greatly depending on the setup and construction of the material system more
of creating not only space and stmcture but also mía-o-dimatic conditíons. While the detaiÍed explanations follow describing specific projects. However, it is interesting to
latter may have a Profound impact on our conception of spatial organisation, which can note the significant shift in the way computer-aided manufacturing CCAM) processes
now _bethought °[as differentiated macro- and mícro-climatic condítions, providíng"'a are employed in this context.
.eÏogeneou,s habitat forhuman activitie5' the former will be the main focus, especíally Whes-eas the nature of CAM enables difference to be achieved, it is currently used
as the research on integral processes of computational morphogenesis and performance mainly as a means of Íncreasing speed and precision in the production of variation.^
evaluation is a substantial fíeld of its own. This basic research entalls developíng-and Symptomatic for presen/Íng the-fadlitative character of manufacturing and its related
exploring new modes of integrating design techniques, production technologies and Protocols is the term ;mass customisation'. Flourishing due to the reintroduction of
system performance. These modes are by no means similar when developed for different affordable varíation, 'rnass customisation' nevertheless remains an extension of well-
systems, but rather differentiate into a wide range of possíble material articulatíons and known and long-established design processes embradng the still dominant hierarchy
computational methods. Só, whíle sharing a common objective, a rich palette of different of prioritised shape-definition and subsequent, purely facilitative manufacturing. One
3f)proã. has been explorcd over the Pa5t tive years through varíous research projects. needs to be aware that the accomplishment of economically feasible variation through
What follows is a cross-section ofthese approaches. comp-uter-controlled production and fabrication, by manufacturers and designers alike^
does not by itself lead to strategies of instrumentalising the versatility of differentiated
DÍfferentiated Material Systems material systems. Nonetheless, the far-reaching potential of CAM technologies is
The research projects presented here seek to develop and deploy computational evident once they turn into one of the defíning factors of a design approach seeking
techniques and digital fabrication technologies to unfold innate material capacity the synthesis of form-generation and materialisation processes. At this point the
and specific latent gestalt. They commence from extensive investígations and tests of highly specifíc restrictions and possibilities of manufacturing hardware and control ing
what we define as material systems. Material systems are consídered, not só much as softwarecan become generative drivers embedded in the setup and development of
derivatíves of standardised building systems and elements, but rather as generative the computational framework.
drivers in the design process.
Extencfing the concept of material systems by embeddlng their material Computational Morphogenesis
characteristícs, geometríc behavlour, manufacturing constraínts and assembly logics Generally, Ít can be said that the inclusion of what may be referred toas system^ntnnsic
within an integral computational model promotes an understanding of form, material characteristics and constraints comprises the fírst crucial constituent ofthe computational
stmcture and behaviour not as separate elements, but rather as complex interrelations. setupïefíned through a series of parameters. The definition of the range inwhich these
This Ínitially requires disentangling a number of aspects that later form part ofthe integral parameters can be operated upon/ and yet remain coherentwith the material, fabrication
computational setup in whích the system evolves. and constmction constraints, is the criticai task for the designer at this stage.

202 Computationat Design Thinking 203 Integral Formation and IVIateriahsation

Steffen Reichert, Responsive Surface Stmcture, University ofArt and Design Offenbach (Achim
Menges)^, Offenbach, Germany, 2006-7. The 'Eesponsive Suiface Structure' project utilises woods
mherent hygroscopic properties to develop a skui structure capable of adaptmg ifs porosity m
response to changes in ambient humidily. The component consists of a moistu.re responsive veneer
composiíe element (Figure l) . Once exposed to a higher levei of humidity, the veneer composite
components autonomously open a gap between the substruclure and the veneer scales resulüng in
different degrees ofporosity (Figure 2). The local component shape and orientation, as well as the
mathematicaüy defmed surface undulation (Figure 3), evolve m contmuous feedback wilh strLictural
evaluation and thennodynamic analysis. As the logics offabrication and assembly had also been
encoded in the mitial computational setup, the evolved morphology could be dírectly construcfed
as afLLÍlyfLinctionalprotolype (Figure 4). ©Achim Menges.

204 Computational Design Thinking 205 Integral Formation and IVIateriatisatíon

griable input to the defíning parameters. This input generates a specifíc output - one
The second crucial constítuents of the generative computational fram^
ÏJdual instance of the system - leading to the registration and analysis of instance
recumng evaluation cycles that expose the system to embedded analysis took"A^i.are
inalv,'fír""structural and environmental effects. Through these effects - basically theway
plays a criticai role during the entlre morphogenetic process, not onfy in estflh ^ sis
'sis
^Lsy'stemmodulates the environment - the system-s performative capadty unfolds from
and assessíng fitness criteria relatedto structural and environmental capadty, butfl^?
revealing the system/s material and geometric behavioural tendencies.The mnrfn.^- cyclesofmanipulationandevaluation. . ., , __._,..
^esTprocesses of driving the development of^the system^through continuai
relation between constralnt and capacity, in concert with the feedback hp+^r?g
,ofits instances can be envisioned in differentways.The most immediate
stimuli and response, is consequently an operative element wlthin the comnni.-,^.
^biïity-is the direct, top-down Íntervention of the designerm the ^^met^
framework. In this way, evaluation protocoÍs serve to track both the coherency of lho
,tion and related assessment cycle. More coherent with the overall conc^pt
generative process with the aforementioned system-Íntrinsic constraints, as well as
l"areiprocessestosed on similar principies as natural morphogenesis^this
systeirTs interactíon with a símulated environment. Dependíng on the system's ir
."twoÏndïof development processes are of interest here: the growthoUhe
environmental modulation capacity, the morphogenetic development process
^dTvTdual'instan'ce aróïheevolutionof the system across generatonsof P°P"l^^n"J
to recurrently mterface with appropriate analysis applícations, such as multiphysics
'^^1 ïnstance^ln orderto facilitate the former, there are,differentcomPuteti^nal
computational fíuid dynamics CCFD) forthe investígation ofthermodynamic relatlons, o^r
nnao'w'tuhu°moderslt'hat canteimplemented, which are ali based on two criticarócto^
light and acoustic analyses. It seems important to mention that a CFD analysis provides
go'nThe onue"hland,'"the~intemalldataset or growth rules - the Senot>/Pe;_and_ont^
only a partia! insight, as the thermodynamic complexity ofthe actual environment is fgr
^'he'r,7h7vari'ab'le"gestalt that results from the interaction^of the ge"°?^h
greater than any computational model can handle at this moment in time. Nonetheless
tuhTenvironment~-- ttephenotype. The criticai task for the designei defining^the
as the main objective here lies not solely in the predictíon of precise data, but mainly in
"through the aforementioned system-intrinsic constraints. The^enerat^on
the recognition of behavioural tendencíes and patterns, the instrumental contributions
of such tools are signifícant. ^'ph^otypi^steminstances, enabled through seed ^^^ns^P^
rfe7genrotypicrewritíng mies, happens in direct interaction w^the environment^
In parallel with the environmental factors, continuai structural evaluation informs
Otntelcrittailiaspectto be considered here, and captured in the computational procesj^s
the development process, or even directly interacts with the generatlon of the system's
the7rofound~influence of goal-oriented physiological regulation mechanisms, such as,
morphology through processes of evolutionary structural optimísation. Yet, in general,
the notion of single-criteria optimisation Is opposed to the underlying principies of forexample,homeostasisonthegrowthprocess. ^ ,. ,.._,_j ..,:,
CEAaach7erÍved^tance'then forms part of a population and^is^evaluated with^he
morphogenesis. It is imperative to recognise that computational morphogenesis does not
aforementioned^nalytical tools. Driven by titness aiteria, evolutionary^ com^atio^
at ali reproduce a technocratic attítude towards an understanding of effíciency based on
^^ug:hltlhelÏmpïementation of genetic algorithms, for example) "n_thenbe employed
a minimal material weight to structural capacity ratío. Nor does it embrace the rationale of
toevoÍvevarious generatíons based on the confluent dynamics of mutation, vanation,
what 20th-century engineers called 'building correctl/. Structural behaviour in this sense
becomes one agent within the multifaceted Íntegration process. Overall, this necessitates selection and inheritance. ^ _ _ __, _, _^
"7ucontinuouis'mediation of th e stochastic evolutionary processes and goal-o
a shift In conceptuaiising multicriteria evaluation rather than an effíciency model,
Biologísts, for example, refer to effectiveness as the result of a developmental process phys^iogicaNevelopments at play, or more g^e"^the^i^,re90^^twe^
to^P^to^own processes, is a central task for the descer Furttonw^
comprisíng of a wide range of criteria. Accordingly, the robustness ofthe resulting systems
^Z'to"enabTe gïnuinemorphological differentiation (that i^changesjn kin^an^not
is as much due to the persistent negotiation of divergent and confíicting requirements as
theír consequential redundandes. ^n~degree), it is of criticai importance that the initially e5tablishedffines^ter^^
'w^lasThFdefínÍng parameter ranges - in fact, the enüre computational framework - is
As ofyet, two essential elements of a computatíonal framework for morphogenetic
processes have been íntroduced: the parametric setup based on the material systenfs caoable of evolving alongside the system-s development. ^ ^ ^ _„,„„„

intrinsíc constraínts, and the evaluatíon cycles through which the interaction of individual
capw'^^rco^ misconceptions may need to be ^dressed^F^^^y^
such'a^mputationalframework challenges the nature °^currenti^stab"shed ^
system instances with externai influences and forces are frequently analysed. In other
words, the possibility of manipulating the system's articulation in dírect relation to approaches^.tdoesnotinvoketheretirementofthe^chitectinfav^of^
understanding the consequential modulation of structural or environmental effects
^'^^^^^ importance ofthe d^neri^n^nati^^
hás been established. Therefore, the processes that trigger and drive the advancing tha^cen^toenabli^, moderating andmfluendng ^^a[úe^ pm^
development of the system are the thlrd criticai constituents of the computatíonal
^^^^U^andsen.tivities.Second.tep^hefacttha^
°dl^ig^a"pp?oa<;h''req'uiresïserious engagement with technology, as may have become
framework. The framework through these processes is able to operate, as they provide

206 Compuí3tiona[ Design Thinking 20Ï Integral Formation and Materialisation

EmTech, Component Membrane, A^
Emergent Technologles and Design ME
Programme (Michael Hensel, Achim Me
MichaelWeinstock), ArchítecturalAfisn^^'
London, UI<, 2007. The :AA Componenl -""u"'

Membrane' project is based onthe definitit

of a component that deploys a ïiyperparahr>ii:,
membrane as a load-bearing tensile
withm a íramework of steel members, The
proliferation ofthe component was evolved
in feedJaack wilh structural evaluation, as well
as environmental analysis of precipitation
(Figure l), wmd and sunlight (Figure 2). The
resulting overlapping membrane articulation
protects from rain, while at the same time
remainmg porous enough to avoid excessive
wind pressure or blocking views across
London's roofscape (Figure 3). Furthermore.
the membranas contribute considerably to the
süfíhess of the overaU structure, which £LG^S as a
cflntüever resting on just three poinls (Figure 4),
© AA EmTech -Achim Menges .

209 Integral Formation and Materialisation

208 Computationat Design Thinking
clear from the above description of the involved computational framework, ifs
The Computational Fallacy
not limited to exotic materiais and manufacturing processes. [...] In effect, as t he m.;'!
expenditure consists ofthe intellectual investment in an altemative conceptualisatÍQnlï Sá nfo rd Kwinter
material systems and related computational processes, this design approachfíourkhln
contexts of limited resources Here complexity and related performative capacityunfoi>|^ Considering the increasing pervasiveness of computation in architecture,
from the continuous evolution and differentiation of initially simple material elempn^ Sanford Kwinter, one of toda/s leading architectural thinkers ana
and fabricatíon procedures. --r-- ..."—..ui cicinents
theoretidans, questions in this artide, fírst published in the architectural
journal Thresholds (2003), the prevailing use of computerísed methods in
FromAchim Menges, 2008, 'Integral Formation and Materialísation:
design, as they are characterised by processes that are reductive in nature.
Computafíonal Form and Material Gestalt', in B Kolarevic and K Klinger
Computing, at its most basic, is the execution of a series of simplifíed linear,
(eds) Manufactïiríng Material Effects: Rethinking Design and Making in
binary operations. As a mechanistic process, Kwinter explains, the numerical
Architecture, Routledge, Taylor & Francis Books (NewYork), pp 1QQ-2ÏQ.
Reproduced by permission ofAchim Menges. © Achím Menges. processes become over-abstractions of physical processes, be it social,
material or cultural. The consequence is an imposed order whose reality can
be realised only through means of translating the representational to the
physical. Computation does not, itself, operate in the same manner as natural
material systems. Models for computation emerge when material (matter)
intelligence is equated to computer (matter) intelligence. Kwinter establishes
a crucial distinction between computation as a wilful engine for calculation,
and the results that it can derive in unfolding the logic and elegance of self-
organising, emergent systems and developmental patterns at extraordinary
and unprecedented depth. This opportunity lies with the ability to allow
computation to investigate the elemental conditions of material íntelligence,
searching for intricate nonlinear order and complex material formation.

The /mechanical/ and the 'electronic' are by themselves not paradigms and do not
represent distinct, successive, agonístic ;agesr or irredudble worlds in collision. To
continue to think of these in such worn and sterile ways can have no other effect than
to hide from ourselves their political dimensions. The mechanical and the electronic (and
most of what is denoted by these terms in present usage) are in fact expressions of two
continuous, interdependent historical-ontological modalities: those of Matter (substance)
and Intelligence (order, shape).
Everyunitof Íntelligiblematter in ourtechnicalor cultural world, regardless howsimple,
is refined or organised to a degree suffícient at least to distinguish it from the random and
disordered background flux or noise of the natural world. COf course, natural objects may
possess this same property of retinement in proportion to how closely they are formed
and organised by the processes of life, processes now commonly understood to extend
beyond the merely organic.) In this sense such matter may be said to possess a greater or
lesser amount of 'embedded intelligence'. One can understand by this a set of instmctions
accumulated overthe ages (through the application oftools and controlled processes) and
Íncorporated into this matter as a kind of permanent and continuaHy reactivated 'memory
[either through shape, rhythm, or disposition as in a tool/ or through purity or precise
proportion as in the relationships of metais in an alloy and the properties derived therefrom).

210 Computational Design Thinking 211 TheComputationat Fallacy

AH matter, even totally disorganised matter, possesses some degree of activp ir What is at stake today hás nothing to do with the eclipse of a material or mechanical
Cwhat DÍderot called 'sensitivity'), and the refinement of matter is always the pfín^!nce ^ by an increasingly electronic one, but rather the emergence of a new regime of
of the intelllgence embedded within Ít. When different types of matter, or r-fifZ, w<,hiprtion that uses the undeniable allure of an archaic revival Ca return to matter,
orientations of intelligence in matter, are brought together in a proper or svmn^knt and free development) to facilitate a repressive reorganisation of social space
rhythm or proportion, an entirely new levei of intelligence is created. As the cnmnill:!. 'well as a mastery of the ver/ conceptual lexicon with which this reorganisation will
added and engaged elements Íncreases, one approaches, then amves at, the 'mprh^ Z, ^e thought through. More bluntly: what is taking place today under the guise of such
Now the mechanical is thought to be primitive in perhaps two senses. Fir^ L- "ational' historical process is the systematic formation of a new subjectivity - a new type
relationships of intelligence are based on rudimentary, visible associations of j. of man, to use NÍetzsche^s expression - whose matter/intelligence variables are being
elements that interact at a very reduced levei of information exchange. Wh at t hic ^engineered and tinely calibrated to fít those of a new machinic workplace-society into
means is that in mechanical devices, most qualitative ínformation tends to be rÉ whích s/he is to be seamlessly integrated.
or elíminatecf in favour of a very controlled, exclusive extraction of quantitative It is therefore disconcerting to observe the direction of much discussion in the design
(rates of movement, measures, etc). Most elements in a mechanical complex have world today around the advent of new telecommunications technologies, computerisation
single functlon and a single set of relationshíps/ só that most of the embedded material and software-driven milieus.These developments are either extolled as 'exciting\ /new',
intelligence is suppressed in favour of a single quality or dimension of expression. dt is and /full of new freedoms and possibilities' (by those blissfully unconcemed that much
here that the commonly confused terms 'mechanical' and /mechanistic/ fínd a justifiable of what is being só celebrated is but an extension of ali that is oldest and most repressive
convergence. The first term means having to do with machines/ the second means in our political and corporeal history), or else are seen as posing an unavoidable or even
deterministíc and reductíonistic.) welcome challenge to an already weakened or near-obsolete domain of cultural practice,
Second, the mechanical is said to be primitive in relation to electronic processes. namely the slow, grave, viscous world of matter.
because the latter appear to manifest the same magicai qualities of material intelligence The routine disdain heaped on matter by both these points of view is in fact focused
found in fundamental, freeand unprocessed matter, asetofqualitiesthatcan besummed Ídeologically on an ofticially distorted notion of the mechanical - made now to mean
up in a single word, self-control. 'Control' here means simply the sustained application of anything that is concrete and available to intuition. Our task today l would argue, is to
intelligent - or organísing - force over time. While the mechanical complex seems over- resist these pathways of thought, and wherever possible to expand the concept of the
desígnedandthereforelimítedtoveryrigid and predetermined pathways ofdevelopment concrete and to extend the play of intuition into new domains.
(Íe, no deveiopment at ali), archaic and electronic matter-complexes arethoughtto be able To do this effectively Ít must remain within our power Cconceptual and political) to
to move and evolve in coherent yet unpredetermined ways. Their manifest intelligence is refuse the advent, not só much of the spedfic machines and techniques of contemporary
both multispectral and free-form or /complex'. development, but of the broader systems of rationality in which they come packaged or
The movement of ali (advanced) technological socleties hás been one from archalc for which they serve as Trojan horses.
matterintelligence Cempirical, qualitatíve, multispectral) to mechanical matter intelligence Communications networks, computers, microprocessor control systems are sodally
(numerical, dissodated), but only incompletely and each in its own way. toxic entities, l would arque, primarily when used correctíy, that is, in their capacity to
In the West/ mechanical matter intelligence took on an almost refígious status (as routinise interactions with people and processes in Íncreasingly enginÊered, confined
electronics is certainlyachieving today) tothe point ofannihilating archaic matter íntelligence and deterministic spaces. It is our duty and mandate to refuse this new, pseudo-material
from public and social memory. Now the way in which a society organises its systems of space entirely, and to follow the 'minora archaic path through the microchip; that is,
intuition - its sdence, its philosophy and its technics - is in every manner a political one. to make the eiectronic world work for us to reimpart the rich indeterminacy and magic
The real and possible arrangements of Íntelligence and matter In nature are one thíng Cand of matter out of the arid, cruel and numericalised world of the reductionist-mechanical
[Íkely unlimited), yet how we represent these possibilities to ourselves is another. To speak and the disdplinary-electronic.
of a mechanical paradigm of material qualities and perceptible functions and to oppose this No computer on earth can match the processing power of even the most simple
to an electronic one of immaterial processes and pure íntelligence is absurd and dangerous. natural system, be it of water molecules on a warm rock, a mdimentary enzyme system,
Absurd, because despite what cyberspace glad-handers may think, there can clearly be no or the movement of leaves in the wind. The most powerfu) and challenging use of the
shape or order (Intelligence) without matter, even if thís maíter is comprised of nothing computer (aside from the obvious benefíts of automated number cmnching in purely
more than pure photons (cinema, retina) laser projection) or molecular acoustic resonance numerical domains such as accounting) is in learning how to make a simple organisation
Cmusic), etc, and dangerous, because such clichés do little more than render us stupid and (the computer) model what Ís intrinsic about a more complex, infínitely entailed
docile in the face of disfíguring yet well-camoufíaged social and hístorical processes. organisation Cthe natural or real system).

212 Computatíonal Design Thinking 213 TheComputationatFaltacy

Implidt here is the idea of leaming how to make matter model matter, or how }ossibilities for such a scenario are almost already foreclosed, and it will certainly not

study natural or 'wilcT Intelligence in a contalned but active, refíning domain. in this come to pass with anything short of a colossal, sustained and collective act of human will.

the computer becomes metalfurgícal substance, it extends the exploratory evolution, [t is we/ the engineers of human environment and actívity, who bear the burden to ensure

process of differentiation and refinement by inventíng new leveis of order and shape. The a properly human pleading in this struggle for our fate.
computerand itssoftwaretogethercan form a Matter/lntelligence unitofaveryprimj^
Sanford Kwínter, The Computational Fallacy' , Thresholàs - Denatured, no
butusefulkínd.ButtodothÍs,thecomputer,inthetriadNature-Mind-Computer,mustpi;
26, 2003, pp 90-1. Threshoïds is the biannual journal of the Department of
only the approprigte intermediary role of interface between Nature and MÍnd. This would
Architecture at the Massachusetts Institute for Technology. Reproduced by
be in ctear contradistinction to what is more often the case today, where computationa]
permission of Sanford Kwinter and Threshoïds. © Thresholds.
environments provide a customary but imperceptible experiential envelope from which
Nature (and ali nondeterministic unfolding) is excluded and within whích the activil
horizon of Mind is Ínsidiously confined. We must not believe the narcotísing hype that an
emerging electronic world is poísed to líberate us from a mechanical one, nor even that
there exists an electronosphere fundamentally discontinuous from the mechanosphere
that hás formed us till now.
It is true that an important transítion is taking place: mechanícal relations are being
dramatically transferred to new and different leveis - like the little bali in a scam artisfs
shell game - but they are certainly not disappearíng. What is more, this transition state is
an unstable one, and one ofthe possible amns on history/s bifurcation diagram (the one
that does not lead smoothly to the total routinísation and economic subsumption of the
human organism) leads at once to the possibllity of multiple new ecologies of human
exlstence as well as to the dark, possibly unfathomable mysteries of nature Ítself.
What we neecl today is twofold. On the one hand, resistance - we need to direct
our theoretical activity away from simple-minded clichés in order to conceptualise the
proper materiality of the electronic with its brutal effect on both human energy and
the physical environment; and on the other hand, productive affirmation - to actlvely
press computatíon toward íts deep rootedness in the archaic world of natural intelligence,
which means at the very least to use computatíon just as the early moderns used the
telescope and the microscope, to engage aspects of nature whose logic and pattern had
previously remained ungraspable because they were lodged at too great a remove from
the modalities of human sense and intuition.
During the Renaissance, specifíc movements and structures of astronomical and
microscópio scale were forthe firsttime brought Ínto the purview of human thought and
perceptíon,andíntheprocesscertainformsofhistoricaltyrannybecameforever[mpossible.
Today the computer offers the possibillty of apprehendíng developmental patterns of
extraordinary and unprecedented depth and abstractíon/ offering tantalising glimpses of
the very free-form structure of time itself Cchaos, complexity, self-organisation).
Just as Lucretius^s hydraullc hypothesís in hís anclent Treatise On Nature once
proposed to free humans from the capriciousness and prejudices of the gods, só thís
new tool - among ali the horrors to which Ít is already giving place - may well bear
the potential to unlock the door on the universal laws that govern the appearance and
destruction of fonn, and in só doing to free us from the multiple tyranny of determinism
and from the poverty of a linear, numerical world. Yet there should be no illusions; the

214 Computational Design Thinking 215 TheComputationalFallacy

Select Bibliography

Abel, C. 'EvolutionaryPlaiming', Architectiiral Desigri, DecemberissueNo. 7/6,JohnWiley& Sons, Londoi mid Gescblchte emei Unlerscheídung. Transcnpt Verlag, Bielefeld, 2004,
1968.
Mayr, E. 'TypologicalversusPopulationThmking', Evolution and. tíie Diversity ofLife, FirstHarvard
Allen, S. Terminal Velocities: The Compuler in Lhe Design Sludio', Practíce: Archítecture Tecímicnie + UniversityPress, 1997.
Represcnfatíon, Routledge.Abingdon/NewYork. 2009. Mayr. E. 'VariationalEvolution'. Wïïaï Evolution Is, BasicBooks, NewYork, 2001.
Aish, R. TromIntuitiontoPrecision', Afi FiJcs 53, The Architectural Association. London, 3005. Menges, A. 'Integral Fonnaüon and Materiatization: Computalional Form and Material Gestall', B

Alexander, C. 'Syslems Generating Systems', Ãrcïiitectural Design. Decembcr issue No. 7/6. JohnVvilei Kolarevic and K lüinger (eds) , Manufaciuring Maïeríaï Eãects: Rethinking Design and Making m

&Sons,London,1968. Arcïiitecture, Eouüedge, NewYork, 2008,

Bentley, P and D Corne. Creaïive Evolutionary Systems, Academic Press, San Diego, CA, 2002. MarvmM. 'Why People Think Computers Can't', Tecbnology Review. Nov/Dec, TechnologyReviewInc,
Cambndge, MA, 1983.
BenLley, P anel D Corne, Evolutionary Desigrs by Cornpiüers. Morgan Kauímann Publishers, San
Francisco, 1999. Mitcheü, W. :A New Agenda for Computer-Aided Design', M McCuUough, W MitcheU and P PurceU
(eds), Tíie Electronic Design Studio: Aicïiitectural Knowledge ana Media m tïie Computei- Era, MIT
Bertalanffy, L v. General Sysl.em Tlieory: Foun.dalions, Development, Applications, George Braziller,
Press, Cambndge, MA, 1990.
NewYork, 1969.
Mitdiell,W. 'Reasonmg aboutDesigns'. The Logic ofArcíiitecture: Design, Computation, and Cognition,
Chu, K. 'Genetic Space: Hourglass offhe Demiurge', AD Arcfiitects in Cyberspace II, Prohle No. 136,
MITPress, Cambridge, MA, 1990.
JohnWiley & Sons, London, 1998.
MitcheU. W. 'The Uses of Inconsistency m Design', Y Kalay (ed), Prínciples of Compuïer-Ãided Design:
Cmtchfield, J. 'Is Anything Ever New? Considering Emergence', G Cowan, D Pines and D Melzner (eds)
£'vaJuafmg'cU]dPredíc^mg'DefflgnPerA?rmanc'e,Wiley-Interscience,Nev/York, 1992.
ïntegratíve Themes, Addison-Wesley, Eeadmg, 1994.
Negroponte. N, Towards a Hmnanism Through Machmes', Arcïiiíectural Desigii, September issue No.
Thompson, DW. On Gro^/vtb arid Foirn: Tïie Complete Rei/ised Ecíition, Dover PublicaLions, NewYork, 1992.
7/6. JohnWiley & Sons, London, 1969.
DeLanda, M. The Mathematics ofthe Virtual', Intensíve Scienco and Virtual Pliilosïiopy, Contmuum,
Negroponte, N. Computer Aids to Design and Arcïiitecture.. Mason/Charter Publishers, London, 1975.
NewYork,2002.
Novak, M. Transarchitectures andHypersurfaces: Operations of Transmodermty', AD Hypersurface
DeLanda, M. 'Deleuze andthe Use of lhe GeneücAlgorithminArchitecture', Neii Le?ich (ed), Designing
Arcliitectures, Profile No. 133, JohriWiley & Sons, London,1998.
fora Digital World, Wiley-Academy, Chichester, 2002.
Pask, G. 'TheArchitecturalEelevanceofCybeinetics'.^reA'terfuj-aróesjgn,SeptemberissueNo. 7/6,
DeWolfT andT Holvoet. 'Emergcnccversus Self-organisation: Different Concepts but Promising \ÏVtien
JohnWiley & Sons, London, 1969,
Combmed', SBmeckner, G DiMrirzo Senigendo, AKarageorgos, RNagpal (eds), Engmeermg self-
organismg systems: meíhodologies and applications, Springer, Berlin, 2005, Schrôter, J. 'Analog/Digital - Opposition oder Kontmuum?', J Schrõter and ABõhnke (eds), Analog/
Digital - Opposiïion ocler KontmuLim?: Zur Tïieone und Geschichte einer Unterscheidwig., Transcript
Prazer, J. An Evolntíonary Arcïiitecíure, AA Publicaüons, London, 1995.
Verlag, Bielefeld, 2004.
FrazerJandMRastogi, 'ANew Canvas', AD Arcïiiïecïs m Cyberspace H, ProffleNo. 136JohnWüey &
Spiller, N. VciciUatmg Objects', AD Arcïntects in Cyberspace 11, Proffle No. 136,JohnWüey & Sons,
Sons.London. 1998.
London, 1998.
Gero, J. 'Creativity, emergence and evolution m design: Concepts and framevrork', Knowïed.ge-Based. Steadman, P. 'Evolutionary Design by Computer', The Evolution ofDesigns: Biological Arialogy m
Systems9(7), 1996. ArcMtecïure and the Applied Arts. Rev. ed., Routledge, London/New York, 2008.
Goethe J W. trans by Bertha MueUer. 'Formation and TrarLsformation' , Goethe 's BotanicdS Wrítings, OX
Terzidis.K. 'AlgonãïmicPoTm', Expressive Form: A Conceptual App-roacli to ComputaíionalDesign,
BOW Press, Woodbridge, 1989.
Spon Press, New York, 2003.
Goodwm. B. 'Developmental Complexity dnd Evolutionary Order', G Cowan, D Pines, D Meltzer (eds),
Turmg.A, 'ComputingMachinery andlntelügence', Mmd, New Series, vol. 59, no, 236, Oxford
Complexily: Metapiiors, Models and.Reality, Westview Press, Bouider, CO, 1994.
UniversityPress, 1950.
Holland. J. Emergence: írorn Chãos to Order, Oxford University Press, Oxford, 1998, Weibel, P. 'Algonthmus undKreativitat', WBerka, E Brix and C Smekal (eds), Woïier kommt das Neue?
Howard, R, Compiitingm Consímction: Pioneers ana tíie Future, Butterworth-Heinemann, Oxford/ Kreativitat in Wissenschaft i.md Kunst, Bohlau, 2003.
Wobum. 1998.
Wemslock. M. 'Morphogenesis andthe Mathematics ofEmergence', MHensel, AMenges and M
Kwmter S. 'The CompLitationalFaHacy', Thresholds - Denatui-ed, No. 26, MIT Deparünent of Weinstock (eds), AD Emergence: Morphogenetic Desigri Sïraïegies. Proffle No. 169,JohnWüey & Sons,
Ai-chiïecïure., Cambnáge, MA, 2003. London, 2004.

Kwinter, S, 'Who isAfraidofFonnalism'; CDavidson (ed), Far From EquSibirum: Essciys on Tecïuiology Wolfram, S. 'HowDo Simple Programs Behave', M Silver (ed), AD Progi-ammmg Cultures: Art ana
andDesign Cnlture, Actar, NewYork, 2008. ArchitecInreintheAgeofSoSware.Promeï-io. 182, JohnWiley & Sons, London, 2006,

LoleU, S. "'TheMere Digital Process ofTummgOverLeaves'', ZurWort-undBegriffsgeschichíe von Wemer, N. Cybemeïics: or Conïrol and Communicaïion in tïie Animal and ïíie Macïime, MIT Press,
"Digita]':'JSchrõterandABõhnke (eds). Analog/DígitaI - Opposítíon oder Kontinuum?: Zur Tïieone CamJoridge, MA, 1948.

216 Computationat Design Thinking 217 Setect Bibliography

Index
Figures in italJcs indicate captíons.
collective 164-65
itre for Land Use and Built Form studies (LUBFs), cybemetics 10,11, 12,16,51, 54,68-76,158, 161,163
emergent 149 Czech Mational Library Internatianal Competition
3D modelling and rendering taols 148 Cambridge176
SdIVIaxScnpt 97
responsíve 160, 161
chãos 100n4,178n35,214_ (2006)17
Bentley, PeterJ 28, 123-26
a posteriori knowledge 177n20 Chen, Han Ju, Manservisi, Sebastiano/ Martinelli,
Bentley Peter J and Come, David W: Creatíve Damin, Charles 21, 42, 43-44, 48, 49,151,159,164,^88
a priori knowledge 177n20 Alessandra
Evolutionary Systems 120 Actualised Populatian 185 On the Orígin ofSpecies 43, 44, 186
AASchool ofArchitecture 73
Bergson, Henri 40n4 Metrical Properties 183 DarwinianTheory44
adaptiogenesis21
Bertalanffy, Ludwig von 13 J 5 [Mon-IVIetrical Properties 183 Darwlnism 44, 48
adaptiveness 55, 56
General System Theory: Foundatjons. population of Vertical Assemblies 184 Davíes, Paul: The Cosmic Blueprínt 149
ADFs (Automaticslly Defined Functíons) 124
DeveSopment, Applications 50 Dawklns, Richard 150
ADIs (Automatically Detined Iterations) 124 Chicago School 190
'An Outline of General System Theory' 50 Christaller,Walter190 de Duve, Christian: Biueprint for a Ce!! 149-50
ADLs (Automatically Defíned Loops) 124
best-flt aptimisation 113, 115,116,117 Church-Turing hypothesis 151 deVries, Hugo 45
Ahlquist, Sean 14, 27
Bichat, Marie-François-Xavier 40n4 climatic madulation27 dedsíon theory 51
Deep Surfaces25
Bildung 16,30,31 CNC (Confígurable Networkíng Computing] machines 166 Dedekind,Richard170,173
Ahlquist, Sean and Fleischmann, Moritz: Evolutionary
biochemistry 77n7 Coates,P125 deformation 35, 38, 160
Spatíal Subdivision, AA EmergentTechnologies and biological causation 38 degreesoffreedom 142, 145,146,181
Desígn Programme22 Colloquy of Mobiles 77n9
biology42, 134, 15Gn1, 159, 190 combinations [binary operatíons) 86 DeLanda, Manuel 20, 26,1 70,177n13,179,181,1 82,186
Akin,0mer92
developmental 10,48, 125 combustion 59 The [Vlachinic Phylum' 191
Aleksander, lgor155
evolutionarylBS common descent 44, 45 Deleuze,Gilles20,774,181,182,187
Alexander/ Christopher 13, 15, 73, 76n2, 102.154
evolutionary developmental 48-49 communication theory 52 Difference ana Repetition 186
Notes on the Synthesis of Form 172
molecular 47, 48, 77n7 complex adaptive systems 163 Dennett, Daniel 90
algorithmsW, 11,24,28, 142, ~I68
population thinkíng in 186-87, 190 complexity100n4,149, 151 J 55-56,158,160, Desargues, Gerará 169
algoríthmicdesign 97
Biomorphic School 156n1 163-64, 210,214 Descartes, René 13, 35
algorithmicform 94-100
biopopulations 42, 49 generated 135 design computation 168-69
evolutionary (EA) 22, 27, 121, 123. 1 26
'Blind Watchmaker' argument 150 computation 10, 11,19, 26, 28, 94, 96, 98, W2,103, analogue109-10
genetic(CA) 95,113, 123,124,131, 148, 164, 207 Bonet i GarT, LluTs 110 115, 118,166n1, 168, 179, 198, 199, 211 through drawing 110, H Ï, 112
inductive 95
Boolean álgebra 74 desígn space 120
search 20-21 evolutionary 20-21, 121, 207
Boolean networks 164 computational fíuid dynamics CCFD) 206 determinism 98, 214
Alice in Wonderland (Carroll) 84
Boolean subtraction 114 deterministic chãos 146
analytic statements 177n19 compute, defmed 102
Borges, Jorge LUÍS 152 Diderot, Denis212
arches103, 704, 105-6. 118 computer science 125, 1 69
Borkin, Harold 91 differential equations 51
Architectura! Design (joumal) 58, 68, 78 romputer-aided design (CAD) 68, 81 ,83, 112,180
Brahé,Tycho34 a new agenda for 86-92 differentiation 158, 164, 190,210, 21 4
Emergence: Morphogenetic Design Stmtegies
Brodey, Warren 74, 80 computer-aided design and manufacturing (CAD/CAM) progressive 190, 191
issue 158
The Design of Intelligent Environments' 154 198, 199 diffusion 59
Architecture Machines 78, 79, 83. 84
area/volume ratio 14
Brouwer, LEJ: /0n the Signifícance ofthe Principie of computer-aided manufacturing (CAM) processes 203 digitísation 96, 98
Aristotle13, 15, 152
the Exduded Middle in Mathematics, Especially in computerisation 10, 13, 94, 96, 98,101n6 directiveness 55
Function Theory' 175 Dischinger, Franz 198
Art Nouveau 72 computers, digital 154
Bruneí, IK70 DNA155
articulated massing22 Computen/ision™ 112
Brunelleschi, Filippo 102, 103 Constraíned Generating Procedures (CGPs) 131, 133, noncoding Cjunk') 46
artificial intelligence 73, 81, 84, 90, IQInlO, -I58. 163
BUILDsystem 100n1 134,135, 736,137, 139 structure 48
artificial Iife152
Burgess, ErnestW190 Dobzhansky, Theodosíus 20, 46
Ashby, Ross 56 constraints 12
Burton, D 70 Durand, Jea n-Ni rol as-Lo u is: Precfs dês Leçons
associative design 179-94 contemplative cognition 31 ni
Bygningen blev tegnet og opfórt som grovsmedie, convergence 128 d'Architecture 152
techniques 181-82
Copenhagen 104 Conway, John 137 dynamic systems 161, 163,165, 175
assoaativity 12
astronomy 34 Conwa/s automaton CLife'} 137, 138, 139,140, 141
Cache, Bemhard 180 Eastman, Charles IVI 91
automated design 11 Coons, Steven 28
CADD55112 coordinates 35, 37, 38 emlogy10,158,165,166
Avery, Oswald 47, 48
Cain,Arthur4G ecosystems 149,153
Copernican system 62
Cantor, Georg 170, 173 cotangent ares 105 Eisenman, Peter 98,101 ni 9,180
Bachelar-d, Gaston 169
capacities 147 Diagram Diaríes 99
Barcelona 106 Courant, Ríchard 172
Cartesian axial representation of space 175 electronosphere 214
BASIC113 creative evolutionary systems 120-29
Cartesian coordínates 38. 160 creativity 26, 28, 86, 94-95, 97,120, 126,149, 155 ellipses33,103, 704,105,106
Bauhaus 72
Cartesian grid 19 Crick, Frands48 embryogeny 120, 121 J 24-25, 126, 727,1 28
Baur, Marco, Kron, Sebastian, Schmidts, Chrístisn. and Case Based Design 21 embryology 190
Weitzel, Christian: Evolutionar/ Design Strateqií Cross, Nigel 169
catastrophetheory14G emergence 24, 131, 135, 137,139, 141, 1 58-68
Project 02 14 crossover 164
catenary curve 34, 103 Crutchfíeld, James P 24 EmTech: CompDnent Membrane 200, AA Emergent
Bayesian methods 95
causality 55 Crystal Palace, London 69, 76n4 Technologies and Design Master Programme 200
Beer, Stafford 68
cellularautomata 99, 131, 137, 138, 139, 740. 141 Cummings, R/V163, 165 Encyclopedia of Computers ana Computer History 180
behaviour 147,164 ' ' --' • •" • ••'

custom-designed modular f unctions 126 energy52, 54,145, 153J54, 157n15, 161

218 ComputationalDesignTMnking
219 Index
engineeríng 69, 76n1,77n7, 159. 160, 202 Galilean principies 98 Jakob;,N125
grouptheory170, 175,176n7
entropy 52
GalileoGaIilei 33, 90 CUI windows 113 Java 25
environment 42, 44, 68,75, 149, 153,160, 161, 166. GameofLife 147-48 Jencks, Charles 153,156n1
214,215 ' ' - ' '"'
game theory 51 Hadamard, Jacques: The Mathematidan's Mind 173 Julia, Gaston 96
environment processing 25
Gardner, M 137 Haeckel, Ernst 45
envíro n mental humanism 84
GaudT, Antoni 72, 74, 112, 118 Embryos 189 Kalay, Yehuda E 91
epistasis 124 Kant, lmmanuel55,170-71,172,175, 177n20,178n35
Casa Mila apartments 106 Hancock, P 126
equifínality 50, 55 Critique o f Pu re Reason 170
Colónia Cüell 109, 110 Harris, Britton 79
Erlangen Program 169, "170, 180 Kauffman,SA164, 165
Sagrada Família Church 102, 106, 707. Harvard Graduate School of Design 12
essentialism 42, 43, 180
109, 110, 777, 112, 113, 774:115,/ïï 7 Haivard University: Joint Centerfor Urban 5tud;es176n3 Kazakov,V124
Eudidean space182 n9n11 .•-.-, ,.^, n/^
Kepler, Johannes 34
Harvey, 1124
Eu;er, Leonard 744 Klein,Felíx170,176n7, 180
Gauss, Cari Friedrich 170, 173 Haüy, René-Just 33-34
evaluation 120. 125
Gehry, Frank 98 heat Island effects 792 Knight, Terry Weissman 88
evolution 19, 20, 21, 42, 49,71, 75, 120, 124. 126. Gehry Partners 115 Koza,John123,124,125
128,129,150,151, 187 ' ' ~ ' '"'
Hebb'srule132
gene activity 164 helicoid 106 Krampe, Florian andVoss, Christopher: Integrated
basic theory of44, 45 Urban fVlorphologies27
general system theory 13,15, 56, 158 Helmholtz, Hermann von 173
biological 148 Kron, Sebastian 14
aimsof 50-51 Hemberg-Extended Undenmayer Systems 12
CGntínuous2W Kuhn,TS100
see o/so system theory Hense!, Michael 22, 208
convergent 163 Kumar,S124,125
generatmg systems 58-67 heredity 42
discontinuous 46 Kwinter, Sanford 10, 16
generative systems 152 Heylighen, Francis 165
gradual 187
genes hierarchy134,155,156n10,170
natural 121. 165. 166 Lamarck, Jean-Baptiste45
regulatory48 HsH Climbing optimisation algorithm 115, 715
stylistic 87 'selfísh' 150 Lamarckiantheory 187
holism13,71
variational 43-44
genetic coding 161 holistic phenomena 58, 59-60, 62, 64,66, 67 Langdon,W124
Evolutionary Desígn Strategies Project 02 14 laser-cutting modelling machines 166
genetic networks 165 holistic properties 58, 59
evolutíonary paradigm 149
genetic pragramming (CP) 123, 124 Holland,JohnH152 Latham,W126
evolutionaryprogramming 123 Lê Corbusier 149
genetics 19, 47 Adaptation in Natural and Artificia! Systems 151
evolutionary strategies 123 Leibniz, Gottfried 171
developmental 48 Emergence: From Chãos to Oírier 131
evalvability125 Leonardo da Vinci 102, 152
molecular 48 Holmes, OliverWendell 39
genome47, 190 homeostasis, homeostat 54, 55, 56, 207 life cycles 164
Fechner, GustavTheodor40n2
genometry, orthogonal 159 homologies 32, 33, 159-60, 188 Liggett, Robin 87, 88
Fechner/s Law40n2
genotype 21, 42, 44, 47, 48, 49, 120, 123, 124,1 25. Hooke,Robert103 limit cycle 146
feedback 15,16, 20, 24, 50, 52,53, 54,55-56, 68. 74. 127, 164, 179, 187,190, 207 ' "' Lin Chia-Ying 192, 193
Hooke'5 Law of Elasticity 25
161, 163, 199, 204,200 • " --" "
genotype representation 123-24, 126 Huxley, Aldous 56 Líndenmayer systems 125
Fermat, Pierre de 144
geometríccontounng 14 Huxley,TH45 Lindgren, Nilo 80
Feynman, Richard 143-44
geométrica! phyllataxis 161 Huygens, Christiaan 34 Linnaeus, Cari 33
Fibonacci series 161, 762
geometry hylomorphic model 180-81, 190-91 Lisp 88
fíeld of rapidities and slownesses 145 Littlewood, Joan 73
and architecture 170 hyperboias 112, 113
fínalism 187
computational 97 hyperbolic paraboloíds 106, 109 logic
flnite element analysis 78 fuzzy 95, 126
descriptive 119n7 hyperboloidsHJ,113,n5
fínitesets 170 and intuition 169, 172, 176
differential 170 ofrevolutionlOG, Ï07J09, 112J19n7
fítnessfunctions 121, 123, 125.126 bgicism 175
Eudidean 95, 99, 100n3, 145, 177n13, 180
Fleischmann, IVioritz22
hyperbolic 171 illuminance analysis 78 LOKAT11,12
Florence Duomo dome 103 Lorenz, Edward 174
and morphogenesis 161-63 incompleteness theorems 175
Fogarty,T123 Lorenz Attractor using 'Chaoscope' 174
form
Riemannian 171 inductíon 99
second arder 103 Industrial Revolution U8,153 Louvre, Paris 100n5
and behaviour160 Lucretius: Treatise on Nature 214
synthetic 169 infinity 170
and process 160
Gero, JS 124, 125 infGrmationn,52,120,151,212 Lynn,Greg100n2J80
formalism 97, 98
gestalt 16, 30, 31, 55, 199, 202, 203, 207 information theory 51, 159
formation 199 McCulloch,Warren78
Ginsberg, Mathew L 92 inheritance 21, 47, 48, 49,124,207
FormZ 97 Mach, Ernst175
glíderguns 148 Mendelian40n10
Fortran 86
giiders 139, 740, 141 instabilityGO machinic phylum 181, 191
Frame Based Design 21 McKenzie, Roderick D 190
Gõdel, Kurt175 Instituto for Computatianal Design, Stuttgart University
Frazer, John 28 macros 112
Goethe, Johann Wolfgang von 16, 19,30, 32 9, 14, 25,27
An EvoSutionary Architecture 20, 149 Maher,ML24
'Ideen über organische Bildung' 30 integration 164
free-form architecture 199 Maillart, Robert198
'Schriften zur IVIorphologie' 30 inteiligence211, 212
Frege, Gottlob170 Mandelbrat, BenoTt 96
Goldberg, DE 124, 126 material 211, 212
Frowd, C 12 G Manservisi, Sebastiano 183, 184.185
Could, StephenJ 21 intensivo properties 191
Fuchsian functions 173 March, Líonel 151, 176n3
gradient-basedtraining 95 interactions147,158J59
Fun Palace projert73, 75 Architectonícs of Humanism 98
gradualism 44, 45, 46. 187 interruptibilitySI, 83
functionalism 70, 74
GRASP11,12 íntuition 155, 170-71, 175, 177n15, 212 Markov chains 99
funct]ans142, 143 Martinelli, Alessandra 783,184, 185
gravity 39-40, ~\03 andlogic169,172,176
Futurism 70 mass customisation 203
lawof143, 144 intuitionism-logicism debate 176

220 Computational Design Thinking 221 Index

Massachusetts Institute ofTechnology (MIT) 12, morphogenetic process 179, 187, 190, 191, 194, 206 ParetooptimalityU3 pureforms40n5
100n1,17Gn3 morphogenetic space 191 park,Robert190 Pythagoras 33
Joint Center for Urban Studies 176n3 morphogenetics 151, 182 Parmee, 1123,126 Pythagorean theorem 100n3
material systems 202-3, 206, 21 0, 211 morphogens 162, 163 Parque Güell, Barcelona 72
materialisation 198, 199 morphological differentiation 207 ParticleSystem25 quadratic equations 176n8
materiality26 morphologicsl-phylogenetic research 49 Pask, Gordon 68, 81 quantum computation 99
mathematics morphology 16, 27, 30, 32, 34,35,151, 156, 160. pattern158,160,161 Quatremère de Quincy, Antoine-Chrysostome 186, 187
applied [mixed) 119n2 200,203,204n ' "" pattern-form models 163
criticai role in architecture 159 multimoda[ity123 pattems of activlty 160 radiation 792, 193
defíned 102 multiple objectives 123 pattems of behaviour 159 randomness 99
and design 17G multiplidties 20, 181, 182, 194 Paxton, J 70 rationalism 97, 98
Eudidean 119n8 IVIusicoloursystem 77n9 Peano, Giuseppe 172 reaction-diffusion model 162
as a kind of software 142 mutation21,22,49,123,164,165,207 perception 155, 177n20 real virtuality 143, 146,147,148
pu re 119n2 mutualism 70, 74 performative morphologies 198 recombination 165
as a searchfortruth 173 Perrault, Claude 100n5 recursion 99
Maver,ThomasW90 n-dimensionalthinking 118 reductionism 56
perspectivism 56
Maya Embedded Language (MEL) 97 Nagakura,Takshiko 88 PERT pragramming 68 region identification 123
Mayne,Thom 98 Magasaka, l 125 Pevsner, Nikolaus 76n5 regression 99
Mayr, Ernst 21, 42, 179, 186, 187. 190 Nash, Jahn 76n3 phase space 146 Reichert, Steffen: Responsive Surface Structure 204
What Evolution !s 42 natural science 150 phenotype 21, 42, 44, 46-49,120,121, 123, 125, 126, Reinhardt, Nica: 3D Gewirke-Verbund^OO
mechanical,the212 natural selection 21, 42-47, 150, 151,159, 190 128,164,165,187,207 Reitman, Walter R 91
mechanism-independent 145 Nature 30, 31, 34, 39,149,151, 152,154,170, 211, 214 representation 121, 125-26 Renaissance 103, 214
mechanisms 131-32,133, 134-35, 736,137,145,159.207 negligibility, principie of 34 phanology 65 reproduction 22, 159
causal 145, 146 asexual 163
Negroponte, Nicholas 73, 77nQ phylogeny 48
mechanistic2~f2 Riemann, Bernhard 170
The Architecture Machine 154 physical chemistry 159
world view55, 57 Refiections on ComputerAÍds to Design ond phy5ics52,55,56,142,143-44,191 Romanes, George John 44
mechanosphere 214 Archítecture 78 Piaget, Jean 175 Romantics 88
meiosis 49 Heo-Lamarckians 45 Pitteway,MichaelLV704,105 Rosin, Paul L 104, 105
Meldahl, Ferdinand 704 Neodarwinism 44 Plato 33 Rossi. Aldo: Thè Architecture of the City 186
Mendelian mheritance 40n10 net theory51 Meno 90 Rotring pen 112
Menges, Achim 14, 22, 25, 27, 208 neural networks 95, 132, 148,155.1 63 Platonian classes 43 Roysl BotanicGardens, Kew: Tropical House 69
Manufacturing Matéria! Effects: Rethmking Newcastle 73 Platonii: primitive shapes 95 rule-based systems generation 12
Des/gn and Making in Architecture 198 Mewton, 5ír [saac 34. 171 Poincaré, Henri 33, 145,146, 173,174, 175 Russell, Bertrand 170, 172-73
Morphogenetic Desígn ExperimentOI 12 Newtonian mechanics 55 Scfence ana Hypothesis 171 -72
Morphogenetic Design Experíment 02 - AA Newtonian principies 98 Soence and Method 172 St Paul's Cathedral dome 103
Strawberry Bar 22 IMietzsche, Friedrich 213 point digitisers 112 saltation46,187
Morphogenetic Desígn Experimentos - Paper Mirenberg, Marshall 48 Poli, R 124 Santa Fe Institute 164
Strip Morphologies 7S noise 123 polygeny 124 scalability 12S
meta- prefíx 101 ni 5 non-Eudidean geometry 118, 171, 173, 175 polymorphism 46-47 Scheffler+PartnerArchitects 17
meta-structures 99 non-Eudidean mathematics 119n8 Poon,J24 Schmidts, Christian 14
metabolism 151 nonuniform rational B-spline (NURBS)-based surfaces 97 PoppecKarllOlnlG 5chnier,T125
metalanguage 69,7Q, 73, 76n5 norm ofreaction 47 population thinking 21, 42, 43,179,182, 194 Schõn, Donald 169
Method ofCoordinates 35 Novak, Marcos 98, 99 in architecture 190-91, 194 scripting 112
Mi[ler,James15S in biology 186-87,190 searchspace20,21, 124
Mílne, Murray 13 Ocean North 17 populational speciation 44, 45 selection 21, 22, 42, 44, 49, 152,207
mineralogy 33 one-to-one mapping 124 populations20,21,42,43,49 byfítness164,166
MitcheII,WilliamJ88. 152. 180 opensystem 15 geametric 22 self-organisation 149,151,154, 164, 165,166,211, 214
The Eíectroníc Design Studio 86 operations research 51 va ria bi e 43 self-shading 25
mixed forms 40n5 organic principies 98 positivism 143 set theor/ 170
Modern Movement, Modernism 87. 110. 180 organisation 158 PostScript128 shadow casting 25
modern synthesis42 shape grammars 97, "125
organised wholes 15 prediction 16
molecular dynamics 164 Oxford Dictíonary of Computing 1 80 premature canvergence 123 Sheppard, PM 46
molecular principies 98 Simon. HerbertA: The Sdences ofthe Artificial ~\S6n'\0
Price, Cedric73
nnonotonidty 91, 92 Simondon, Gilbert: L"mdividu et sá génese physico-
Page,J124 Prigogine, llya 151, 161
Montessori system 64 Paley, William 150 primitives 135,180,181 J 82, 190, 191 bioSogique 190-91
Morgan, TH 45 Palladio, Andrea 180 probability 99 Simplex optimisation algorithm 115, 116
morphism 97, 98 papershapes 727,128 process 160,162 5íms,K126
morphogenesis 10, 12, 18, 19-20, 26, 159. 161-65. parábola 103 programmatic adjacencies 22 singularities 143-47
186,202 paradigm shift 100, ~I01n19 programmatic aliacations 22 Sketchpadll-13
computational 199, 202, 203 parametric design 1 ~\ 2-1 3, 114,181 progressive mechanisation 54 Soane, SirJahn 76n3
digital 199 parametricequation 32 Prolog 88 Socrates 90
natural 199. 207 parametric systems 78 Ptolemaic system 62 soft architecture 154
morphogenetic materialisation 149 parametrics 12, 19, 32 Puig Boada, Isidro 110 solar system 62, 53

222 Computational Design Thintíng 223 Jndex

solid/void ratio 14 topology 86, 98, 101 ni 2,145, 169, 176, 184
spatial voids22 trajectories 168
spatiality 26 transcendency 98
^utationai Desígn Thínkmg is the fourth AD Reader. This reader series
spedation 123 transformation35, 38-41, 163
species 42 Transformation-groups 40n8 John Wiley & Sons, the publishers of Architectura! Dessgn G4D), invites
stabilityGO, 151 transformationlsm 43, 44, 187
statespace146, 168
influential architectural thinkers and educators to compile anthologies on
Eransformations (una ry operations) 86
statetrajectory134 transition functions 131, 134,135, 137 core topics of study for students of architecture and design.
Steadman, Philip 87 transmutationism 43, 44, 187
Stillwell,John173 tripie points 110, 115
Stiny, George 88 Turing, Alan 90, 151, 161, 162, 1 68 previous titles in the AD Reader series include:
stochastíc models 51 Turing Machine 148, 151
Strickberger, Monroe W: Skeletal Structures of type 21,179,186
Forelimbs 783 typologicai thinking 182, 187, 194
'string loop and two nails' method 103, 104, 105
structural analysis 98 Ulam, Staníslaw 137
subroutines 134 uncertaínty 151
substitution-groups 40n8 Universítat Politécnica deCgtalunya 113 The Diagrams of The Post-Modern Reader
Space Reader:
Sullivan, Louis 149 UNIX113 Architecture Edited by Charles Jencks
supercomputers 148 unselfconsdous process 58
Heterogeneous Space in
surface scanners 112 LJRBAM5 (motion picture) 82 EdÍted by Mark Garcia CSecond Edition)
Architecture
Surrealism 72 Utzon,j0rn194, 195, 196
Sutherland, Ivan 12, 92
Edited by Michael Hensel,
Swift, Jonathan 152 van Doesburg, Theo 11 8 Christopher Hight and
Sydney Opera House 194, 195, 195 variation 21, 42, 43, 46-47. 179,207
symmetry-breaking 151,161, 165 natureof 48-49
Achim Menges
synthetic statements 177n~I9 variational versíon 144
system theory IO/ 15, 16, 158, 1 63 Vavak,F123
see a/so general system theory vectorfíelds 17
Venturi, Robert: Comptexity ana Contradiction 194 w
Tan, Milton 88 verticality 16
Taura,T125 Viço/ Giovanni Battista: The NewScíence 100-101n5
READER.
technology 52, 54, 57,121, 153, 1 98,199, 207 virtual reality142, 143, 147, 148
tectonics 159 volumetric subdivision 22
Teicholz, Eric 12 von IMeumann, John 137, 151, 156
teleology55, 56 Voss, Christopher27
Temple Meads 69 voxel representation 124

Spáçe Reader
tendendes 147
Terzidis, Kostas 10, 26, 28 Wakefíeld,JP124
Expressíve Form 94 Wallace, Alfred Russe[43,44 Heterogeneous Space in Archjlecture
Thackara, John 153 Watanabe, Mika and Lin Chia-Ying
Eiuicdtiy' • Michael Hensel ';^^>;.,
Th eo r/ of G ro u ps 40n8 Intensive Property 792
Christopher Higbt ^^
Theory ofTransformations 35, 40n8 Interrelationship between Extensive Properties
Achïm Monges:
thermodynamics 151, 161, 163, 199, 204 and Intensive Properties 793
thernnoregulatíon 54 Watson, James D 48
ThInkADot 59-60 Weibel, Peter: 'Algorithm and Creativity' 26, 28
Thompson, D'Arcy Wentworth 16, 19,159-60 Weierstrass, Karl 173
The coordinate systems of the fírst and second Weinstock, Michael22, 208
figures with three intemnediate systems Weismann, August 44
interpo!ated37 Weítzel, Christian 14
The first Íntermediate coordinate network, with its Whitehead, Alfred North 159, 160
corresponding inscribed pelvis37 Whorfían effects 98
On Growth and Form 32 Wiener, Norbert16, 54, 160
ISBN 15BN 15BN
Pelvis of Apatornis 36 Wolfram, Stephen 164
978-0-470-51945-5 Cpb) 978-0-470-74866-4 Cpb)
Pelvis ofArchaeopteryx36 Wood,Peter113, ?ï6,n9n10 978-0-470-51943-1 Cpb)
Theory ofTransformations" 19 word system 58, 59, 60 978-0-470-51944-8 (hb) 978-0-470-74867-1 (hb)
978-0-470-51942-4 (hb)
Threshoíds (architecturaljournal) 211 Wren, Sir Christopher 76n3, 102
Thue, Axel 11 Wright, FrankLloyd 149
Todd,S126 Wrights organic thesis 71
tools 97 W5CC 68
topological invariants 146
topological mapping 25 Xhyper program 113, 115,116
topological properties 97

224 Computational Design Thinking