Dec 20, 2007
Originally posted on Half an Hour, December 20, 2007.
Text of my presentation to SURF Education Days, 13 November 2007, Utrecht, the Netherlands. Slides, audio and video are also available.
1. Introduction
I don’t have fancy slides today. I don’t have nice pictures or anything like that. I’ve spent some time over the last few days looking at a paper by Paul Kirschner, John Sweller and Richard E. Clark, which describes the “failure of constructivist discovery problem-based experiential and inquiry based teaching.â€
For those of you who are familiar with my work you’ll know that a great deal of the work that I’ve done is constructivist discovery problem-based experiential inquiry based teaching. And so this sort of paper is published in Educational Psychologist, which I guess is an important journal. It’s been widely cited.
This sort of paper and the criticisms associated with that sort of paper raise questions. I do work in educational theory and I also do work in software development and I’m never completely sure that I’m doing the right thing. I raise questions. Is the work grounded in research? Is this really the way learning happens? And so I question myself and I question myself not just because it’s a good thing to do, but just because I’m not positive.
So I worry when I see papers like this. I worry that maybe the foundations of the work that I and other people are doing is not well-grounded, but then I look at a paper like this and I realize that what really needs to be done is that these arguments need to be drawn out. They need to be made explicit and it needs to be shown very clearly and not ambiguously why these are not good arguments and that is going to be my task today.
And for you who are listening to this task today, what you’ll be able to take away from this, I believe, is, first of all, an outline an idea of the theory of learning that underlies learning with blogs and wikis and other web 2.0 technologies, but also, more importantly, a way to respond to people who think that instructional technology and the use of the Internet in online learning ought to be nothing more than the presentation of instructions to students telling them what to do. That is control learning. That is the old way of learning. The web way of learning, Web 2.0 learning, is the new way. It is free learning and that is what I advocate.
2. Connectivism (Free Learning)
Sometimes this theory is known under the heading of connectivism. Connectivism is a name that was coined originally by George Siemens. Connectivism is essentially, and this is my take on it not necessarily George’s, the theory that knowledge and learning can be described and explained using network principals.
Now what do I mean by that? What I mean is: knowledge itself, to know something, is to be organized in a certain way, to have a certain pattern of connectivity in the mind, a certain neural connection in the mind. To learn on that theory is therefore to acquire that pattern of organization. To learn is therefore not to have things pushed in your head but to grow and to develop in a certain way and specifically to grow and develop in such a way that you are able to recognize patterns in the environment.
Connectivist learning theory, therefore, is based on the theory of how networks learn, that is to say, how networks grow, how networks develop, how networks form structures of connections between neurons. There are four major ways in which networks grow. And I’m not going to say that these are the only ways, these are all the ways, that this is the definitive statement. But these are ways that we have observed through history that networks grow.
One way is simple Hebbian association. What that means is that if two neurons fire at the same time and don’t fire at the same time, a connection tends to be drawn between them. That’s it. Very simple!
The second way is accidental association. If two neurons are beside each other a connection tends to join them.
The third way connections are formed is back propagation and this comes from the theory of connectionism in the field of computer science. The networks form their connection and then feedback is sent into the network according to the output that the network produces. If the network produces good output the connection will be reinforced. If the connection produces bad output the connection will be broken.
And then finally, Boltzmann learning, which is a theory based on thermodynamics, which says essentially that connections will tend to form at the most stable configuration. If you think of it as like throwing a stone into a pond, the water will settle out. Well connection in brains work in much the same way, according to this theory, and the brain settles out.
The main thing to understand here is that connectivist learning theory is about how connections form in the brain, and for that matter how connections form in networks generally. Because connectivists talk about not simply networks in the brain, they also talk about learning networks in society at large, networks of people in society who are connected to each other. The two theories work out to be two parts of the same theory.
In connectivist pedagogy, therefore, to teach is to model and demonstrate. To teach is to present experiences to people so that they can begin to form these connections in their mind. And then to learn is to actively form these connections by practicing, by repetition, and by reflecting on that practice.
Both of these imply what might be called participation in an authentic community of practice. The idea here is that to learn is to put oneself in a situation where you are practicing in the way that whatever discipline you are in is practicing. For example, you learn physics by doing physics. You learn how to take care of forests by going to a forest.
The role of the teacher in this model is to practice one’s work in an open manner. This has been a challenge, I think, for pretty much all of society, but the idea here is that instead of doing your work in secret in back rooms without being open about what you do, you do your work in an open and transparent manner so that people can see what you’re doing.
In preparation for this talk, for example, I created a summary of the Kirschner, Sweller and Clark paper and I put that on my Website and I collected notes from other people who wrote about this paper and I put it on the website and what I wanted people to see is how I go about assembling my thoughts in order to prepare for a talk like this.
To work, on this theory, is to engage in a community. Most of us, when we do our work, in whatever profession, don’t do it simply all by ourselves. We are involved in a community of practice. We have shared ways of doing things. We have a shared vocabulary, a shared understanding of what constitutes success, a shared understanding of how we test for that success. And it’s to be openly reflective to think about what we’re doing in this community and to think about whether it is the best way to do it, why we’re doing the things that we’re doing.
In this model the role of the learner is to themselves in some sort of environment like this. It may be the actual community of practice itself, which is what I recommend, or it may be a simulation of that community, perhaps a role playing game, perhaps electronic performance support systems, perhaps the actual community of practice itself. You can imagine all the different ways a learner can place themselves into one of those communities of practice. It is to observe the way people who are successful in that practice actually conduct their practice and it is to be reflective to engage in conversations about that practice.
You can see why Web 2.0 and Internet technologies like blogging and wikis and things like social networks play such an important role in these theories. These are technologies that make this sort of activity possible. These are technologies that make it possible for a person to practice their discipline in an open manner. These are technologies that make it possible for a community of practice to develop on a worldwide basis connecting people from many different countries together. They make it possible for the learner to observe experts, to interact with experts, and to learn by doing.
3. Instructivism (Control Learning)
The other view of the world is known as, well, there are different names for it, but I will call it control learning or instructivism, and it is the approach that’s characterized in the Kirschner, Sweller, and Clark’s paper.
It’s a model of learning, and especially on line learning, that is what we might call traditional on line learning. What this learning is, is learning based in the theory of learning objects and based in the theory of the learning management system.
And the model here is that our on line learning environments basically emulate the practices and the processes of the content in a traditional classroom. Learning objects contain content in the core, and pedagogy that is wrapped around that content and the idea of this traditional learning is that the content must be explicitly instructional. The content must actually guide the student or the learner through a series of instructional steps.
The idea of learning object theory is that the learning object needs to have, for example, things like learning objectives. It needs to have guided practice. It needs to have assessment of some sort.
The basis for this theory, the basis for this approach and where it is practiced the most is in environments like the U. S. military where the Sharable Courseware Object Model or SCORM was developed. And SCORM is basically a model of personal learning where you go step by step by step through the learning manual and you follow the instructions that they tell you to follow, you do what they tell you to do and that is what is supposed to produce learning.
The learning management system and we’re familiar with the learning management system, we’re familiar with systems like blackboard, for example, or even open source systems like Moodle and Sakai. The intention of these programs is to present this material step by step by step.
In the introduction (to this talk) the Buntine oration that I gave in Perth is mentioned. In that talk, there are three major locuses or loci of control; three places in which control of the learning process, control of the learning content is exerted in this traditional picture.
One is in content packaging. And if you think about content packaging what content packaging is, is you take a bunch of learning objects, put them all together, put a wrapper around them and then you compress them with a zip archive or something like that, thereby making them useless even to a browser. Content packaging is a picture of learning as a package that you put on a shelf like a book in a library.
Federated search, second. Federated search is not like Google. Federated search, which is the approach that is recommended by the people who created SCORM by advanced distributed learning is a mechanism where when you search you search this library, this library, this library, and this library and that’s it. You search only from recognized authoritative sources who have ‘the knowledge’, whatever that happens to be, authoritative sources, I guess like Educational Psychologist, the journal.
And then, third, learning design, which is a third major component, which actually has its origins here in the Netherlands with Rob Koper’s educational modeling language or EML is a mechanism for stepping you through the presentation of learning materials. Now Learning Design will put you in roles and Learning Design will branch and present different materials in different circumstances, but it’s still that theory of presenting material, presenting material, presenting material, and the theory behind that is if you are told what to do you will learn. And that’s the basics, the basis of instructivism.
4. The Argument
And so now I turn to the paper, that I intend to criticize in this talk; why minimal guidance during instruction does not work and the corollary of that is why maximal guidance does work.
And what the authors do in this paper is they set up two alternatives and on the one side, say the authors, are those people advocating the hypothesis, people like me, that people learn best in an unguided or minimally guided environment.
(Just as an aside every time I quote from them I’ve very sensitive to their use of language. Their use of language is very often loaded or prejudicial and I’m trying to be careful not to be drawn aside by that. “In an unguided or minimally guided environment.†What does that mean?)
Generally defined as one in which learners, rather than being presented with essential information, must discover (now they throw that word discover in there very deliberately because they want to tie it to discovery learning) or construct (there you go with constructivism right) essential (notice that word ‘essential’ - that word has a whole philosophy that comes with it, the whole Saul Kripke essentialism view of the world that there are certain innate natures of things) information for themselves. So that’s the bad thing.
On the other side, this is their view now, are those suggesting that novice learners should be provided with direct instructional guidance on the concepts and procedures required by a particular discipline (we’ll come back to that) and should not be left to discover, (notice how it’s changed a little bit from the first presentation) those procedures (again a very careful word) by themselves.
Just as an aside before I get into the main criticism, I was uncertain, and of course they do not discuss in the paper, what a novice learner is. There’s a certain sense in this paper that a novice learner is any person who has not learned what needs to be learned. And if you read the paper a certain way the paper says if you already know what is being taught you have no problem being taught it, but if you don’t know, then you have to be instructed, properly so called. That would be an unfair reading of the paper, but they leave it open because they don’t tell us what a novice is. Is a novice a baby? Is a novice a 10 year old? Is a novice a first year student in a college or university? We don’t know.
So what do they mean by minimally guided learning? Everything! The old theories of discovery learning, problem based learning, inquiry learning, experiential learning, and of course, the most recent thing, constructivism and, if they had thought of it and felt so inclined, they would have included connectivism and on line learning probably all in the same breath.
There are two assumptions, they say, to this unguided kind of learning. The first assumption is that students should be challenged to solve authentic problems (notice the scare quotes) or inquire information, acquire knowledge (again notice how that is phrased, ‘acquire knowledge,’ I’m going to get it from here and I’m going to put it in here) in information rich settings. And they say (notice again the loaded term, the assumption right, as though we’re just making it up) the assumption is that having learners construct their own solutions, whatever that means leads to the most effective learning experience. And then the second assumption, they say, is that the non-guided people assume that knowledge can best be acquired through experience based on the procedures (the discipline). We’re going to come back to that.
Now their argument, this is their main argument here, it’s a nice categorical syllogism. I like categorical syllogisms because they’re so easy to work with. Any instructional procedure (i.e., ours) that ignores the structures that constitute human cognitive architecture (now there again, ‘human cognitive architecture’, as though the mind is like a house) is not likely to be effective. There, you can see where this is going, right? Minimally guided instruction appears to proceed with no reference to the characteristics of human cognitive architecture. So we’re just one of those things. And indeed, what they mean now (we’ll come back to this later in the talk) by ‘human cognitive architecture’ is the characteristics of working memory, long-term memory and the relations between them. Okay, fine. We’ll come back to that. Thus, they conclude, minimally guided instruction is unlikely to result in effective learning.
You see why I certainly worry about papers like this is because what they’re saying basically is the approach that people who are talking what 2.0 learning, 2.0 blogs, wikis, social practice, communities practice and all that, has no grounding in the theory of how the mind works. And that’s a very serious charge. Turns out to be false but it’s still a very serious charge.
So this is their credo, their manifesto, a.k.a. their conclusion: after a half century of advocacy of minimally guided learning (people like me) it appears there is no body of research supporting the technique. Now, off in the distance you can hear the howls and the wails of protest from the people who have been studying this stuff for 50 years and have found that it works, but we’ll leave that aside.
Insofar that there is any evidence, they argue, it almost uniformly supports direct strong instructional guidance rather than constructivist based minimal guidance. Not only is non-guided instruction less effective (they even talk about this for a bit) it may produce negative results.
And you ask, “How can that be?â€
5. The Reality Check
Well let’s do a reality check first about their conclusion. Their argument, first of all, is simply inconsistent internally. These aren’t major issues, but it’s a bit of a problem.
On the one hand when they’re busy criticizing the minimally guided research they say instructors can’t apply it, they always cheat. They always do some scaffolding. They won’t let students discover things for themselves. They’re always suggesting, telling them what to do and all of that. Okay, fair enough.
But on the other hand, they say that minimally guided instruction is failing. Well either they do it and it doesn’t work, which is bad, or they don’t do it. You can’t say both. You can’t say they’re not doing it, and it doesn’t work.
And then also the section later on in the paper they talk about how minimally guided learning, discovery learning, constructivism, and problem based learning especially, are used in 50 medical schools in the United States. Now I hadn’t heard any particular criticisms about the quality of doctors in the United States. Maybe that’s just a fluke.
But they examine this and they say the strongest criticism they find in those doctors, because the doctors turn out to be fine, but in their paper, they recommend, “unnecessary procedures,†and it strikes me knowing the American medical system, that the last cause of the recommendation of unnecessary medical procedures is the kind of learning that we do. This is the most litigious society in the world. If you drop your pencil you will be sued. That is why they recommend unnecessary medical procedures (whatever those might be) not because they were taught to use problem-based learning.
But be that as it may. Their conclusion is simply not plausible. It doesn’t make sense. It’s not believable. We know that people learn using problem based learning and inquiry based learning. There’s a huge body of research and Hmelo-Silver, Duncan, and Chinn cite numerous studies in their response to the Kirscher, Sweller and Clark paper. And even without that research we know that people learn without guidance because we see it all the time. We have the evidence of our own senses.
Nobody went to school to learn how to build the Internet. Nobody was instructed (‘they said well first you get a website, get it on the web, get some HTML’). People discovered that all by themselves, and it turns out that the mechanisms of computers are things that even small children can learn by themselves.
Many examples, I’ll just point to one. It’s called the ‘hole in the wall gang’ and what they did in cities in India is they would take a computer, they put it literally in a hole in a wall and so the make it assessable to the children in the community. These are children in India, so it’s not like they’re growing up working on their laptops at home. And so they get this computer. They’re not instructed in any way. They learn how to operate it, they learn how to program it and they learn all kind of things about this computer without being told to do anything. So we can see that this works.
Even more to the point, instructivism is a kind of learning by telling; it’s a kind of learning by giving people the information, the concepts, the facts. But we know that people have to learn by practice. Learning is not being told. Learning is doing.
Examples are all over the place. Deanna Kuhn writes that we can hope to impart the smallest fraction of knowledge in any science. How could we possible teach science by teaching facts when there is not literally, not enough time to put all the facts in people’s heads? Even if we were putting them in one after another every second of every day, there’s too many facts in science. We have to do it a different way.
Think about what you would want in medical practice. Would you want a doctor who was told about medicine or a doctor who practiced medicine?
Their argument is based on a straw man. Inquiring learning, problem based learning, are not examples of ‘minimally guided learning’ and again, Hmelo-Silver, Duncan, and Chinn talked about this at some length. They are based on the process of scaffolding, they are based on direct instruction when needed. Indeed, my criticism of problem based learning and inquiry learning is the instructor is too involved. I think there’s too much instructing happening in these kinds of learnings and that there should be less. But that’s a separate argument for a different day.
Their argument is a false dilemma and this is the easy and obvious criticism. They are offering (remember at the very beginning where I presented their argument) the choice between minimally guided instruction or strong instruction. And it begs the question who is doing the choosing, doesn’t it?
This is the example I like to use for this. Imagine my first visit to Rome. I remember getting off the train and walking out and I’m in Rome, in the train station. I have no idea what I’m doing and so a number of things present themselves here, right? I could take a guided tour. I could get a map and walk around myself. I could just walk around aimlessly and never find my hotel or… these are all things where I am choosing what to do, right? And the alternative is, to be kidnapped and to be told where to go.
Now the alternatives are not ‘being kidnapped’ or ‘being lost’. There’s all kinds of ground in between, all kinds of ways that a person can receive guidance that is not in the form of direct instruction. And so their argument is a straw man, yet again.
People who are minimally instructed are in no sense cast adrift. So in preparing this paper for today, nobody told me how to prepare this paper, but that doesn’t mean that I’m sitting all alone. I put out messages in e-mail, I log onto websites, I got all kinds of information back, really helpful useful information that mapped out the territory for me, told me about resources I hadn’t found, pointed to me through objections. I got lots of guidance when I was preparing this talk, but I was not instructed.
And this is a general criticism. This isn’t just me. There’s a social dimension to much learning and Miles Berry points to that in his criticism of the article. A lot of learning, even traditional learning in the classroom, contains a large social dimension as you interact with the other people in your class.
6. Scientific Practice
There are some deeper misunderstandings in this paper as well and I want to explore them. Let’s turn to Kirschner, Swellers, and Clark’s explanation of why learning turned out to be the way it is. And their explanation is inexplicably US-based, but we’ll leave that aside.
They identify the curricular reform that happened after Sputnik (it’s kind of neat, Sputnik happened and then I was born; I came into the world roughly the same time as Sputnik, so I am a product of these reforms, maybe that’s what really scares them). And what the reform is, they say, and it’s repeated throughout the paper, is that it’s based on the assumption that knowledge is best or can only be acquired or learned through experience that is based on the procedures of the discipline. And they repeat this four or five times, I didn’t count them, in the article.
And so they’re saying the assumption here is, if you want to learn physics you should practice physics the way a physicist does, which is what I said at the beginning of this talk. And this has led, they say, to this unguided project work (I did lots of projects when I was in school: Ecuador, the Danube River…) and a rejection of instruction (this is the key phrase, it’s repeated several times in the paper), a rejection of instruction based on the facts, laws, principles, and theories that make up a discipline’s content.
That’s a pretty common view, isn’t it? There’s probably fewer advocating wikis or weblogs or something like that, Wikipedia. People are saying, “what about the facts, what about the laws, what about the content that actually is the discipline, physics, or mathematics or whatever?â€
And they say it may be, they say it is, an error to assume that the pedagogical content of the learning experience is identical to the methods and practices of the discipline being studies. So what they’re saying is that the basic fundamental assumption of my own theory is in error. And it’s a mistake to assume that instruction should focus exclusively on application (well it’s one of those weasel words, ‘exclusively on’, it’s not what they mean, what they mean is it’s a mistake to say that instruction should be application).
Well what do they think? How do they think science works? Because if it’s a mistake to adopt that method, then the nature of that method is pretty important, don’t you think?
Well, happily they explain it in a couple of places for us. One place they explain it is when they discuss Kolb (1971), and Kolb and Fry (1975) and they present a process where a person carried out an action and sees the effect, and then they see this and they understand this effect and begin to anticipate the consequence, and as a consequence of that, they generalize. They understand the general principle.
It sounds like discovery learning. And they extend this to other types of unguided or minimally guided instruction. It’s one of two major components of problem-based learning, they say in their paper: explicit teaching of problem solving strategies in the form of the HD (hypothetical-deductive) method. Barrows & Tamblyn 1980. 1980 is kind of significant because (Thomas) Kuhn was only about 1974 or so.
And then teaching, it’s the same principle. Teaching of the basic content in the context of specific case or instance. So again the same sort of set up here. You have the specific and as a student you’re supposed to generalize. That’s the law. And there’s the problem. That sort of way of going about learning, they argue, might not be the most appropriate way to solve problems. In fact, it’s really difficult, especially in clinical settings, especially in “information rich environments.†It’s really difficult to come up with generalization. You’ve got all kind of convenient hypotheses (that reminded me of Chomsky when he talks about the poverty of the of stimulus). There’s so much information it’s really hard to pick the right generalization.
But the thing is, real science - the stuff that real scientists do in real labs - is not the hypothetical deductive method. Hasn’t been since the ‘50’s. It was developed by Carl Hempel, and the ink wasn’t even dry on the page when Hempel and other logical positivists like A.J. Ayer were being criticized all over the place. Karl Popper, right off the bat, not verification but falsification, which is very much not the HD method. And numerous others. Kuhn - Thomas Kuhn. Lakatos. Lauden. Feyerabend. All pointing to the fact that scientists in practice do not practice the HD model. Nobody does it.
What is science? Science according to Thomas Kuhn (The Structure of Scientific Revolutions) is a community process. The process is not argumentation and it’s specifically not inductive argumentation. It’s explanations. The criteria for explanations are theory based or theory-bound and include things like simplicity, parsimony, testability. And explanations aren’t stand-alone facts. They depend on your expectations; they depend on your theories. As Bas van Fraassen says in The Scientific Image, the explanation of something depends not only what caused something to happen, but also in your expectation of what could have happened instead. Obviously.
7. The Prestige
Let’s turn now to the core of the Kirschner, Sweller and Clark paper, which is the ‘cognitive load argument’. This is where you’re going to give you cognitive psychology and tell you, here’s how learning works.
So their theory is based on the theory of long term and short term memory. They’re not going to be interested in sensory memory (I find that a very interesting statement I’m not going to linger on) and the manner in which our cognitive structures are organized. You’ve heard this before, right? You have short term memory, you have long term memory. Short term memory is the stuff that we are consciously aware of. Long term memory is the stuff we aren’t.
Long term memory is - and these are their words - a ‘massive knowledge base’ and anyone who understands cognitive structures knows how bad a statement that is. If you looked at the structure of the mind, it does not resemble a knowledge base at all. And I won’t get into the details of it, but neural structures and databases are two very different things.
And they say, you’re skillful in an area because your long term memory contains huge amounts of information, so it’s a theory knowledge based on quantity, piled fact on fact on fact. And then they cite – astonishingly – DeGroot’s work on chess expertise. But, you know, I play chess. And there’s this fiction that chess players who are really good chess players can predict ahead, and they can keep all these different board positions in their head. But you cannot defeat a computer like that. The computer will always predict ahead further than you. But people still can beat computers because they visualize, they recognize, successful formations. They don’t memorize a whole bunch of chess positions; they see what is going to work and what is not going to work.
This is Kirschner, Sweller and Clark again: they say the alternate means of instruction is justified by this cognitive theory of long term memory. And the aim of all instruction they say is to alter (nice word there, ‘alter’) long term memory. If nothing has been changed in long term memory they said nothing has been learned. It harkens back to the logical positivist principle of verificationism, that a difference that makes no difference is a difference at all. Or in Bergmann’s formulation, in the dark, all cows are black.
And then the other side of the theory is that working memory, short term memory (and we’ve seen the research on this) has limited capacity, has limited duration, only lasts for a few minutes and is restricted to a small number of elements - the famous number seven elements. We’ve done lots of work on that. Seven elements doesn’t mean seven digits because if you taught to cluster things you can actually remember more than seven. That’s going to form an important part of their theory.
Here’s where the prestige comes in (that’s from magic, right, the movie, you do your trick and then you do the unexpected thing and it all comes together and all the crowd goes ooh ahh). Here’s where it all comes together.
Most learners can, to use their words, construct knowledge. People can actually spot the generalizations. And to do this you must construct a representation. Now this representation, they say, is equally good whether you’re using full information or partial information. So if you have the full information you’re really not losing anything, but if you have partial information, part of your short term memory is occupied in trying to search for possible theories or hypothesis. And because you’re searching for what theory this data could fit into, you’re overloaded, your short term memory is overloaded and you are not able to focus and actually think about what you’re supposed to be learning.
And so, constructivist – because it’s making you search for these theories - is not a good prescriptive instructional designed theory. It’s too hard.
8. Personal Knowledge
Well it’s all in the search, isn’t it? And this is the putative ‘worked example effect’, where you’re doing a search for the hypothesis (imagine a logic problem, “oh what principle am I going to apply, am I going to apply transposition, composition or whatever?â€). So you’re searching these hypotheses, and it’s unrelated to learning, you’re trying to find the best fit, but if you’re studying a worked example, the theory has already been picked for you and you’re looking at what has been done. You’re learning by being told.
The problem is, problem solving - real human problem solving - doesn’t work that way.
We don’t do this mental search of our internal hard drive and try to find the right theory. That’s not how it works at all. We don’t have a whole bunch of general principles or theories stored in our head. Rather we looked at the data and we recognize patterns in that data. It’s a process of seeing rather than searching. This is why is so interesting that at the beginning of this they dropped perceptual memory as though it was completely irrelevant. But perceptual memory is totally relevant because perceptual memory is a process of recognition.
There’s even in the literature there’s discussion of this (I’m not just not making this up). Stephen Kosslyn, for example, image-rotating examples, the theory where you’re going through algorithms and processes does not explain reaction time when you’re asking people to visualize the rotation of images. We work on a sub-symbolic level, not a symbolic level. Cognition is based on a process of pattern recognition at that sub-symbolic level.
And we know this. There has been a lot of discussion about experts and schemas and how an expert will acquire a frame – to use Lakoff’s term – or knowledge of organization. And they even admit, the authors even admit, cognitive research has shown that to acquire expertise in a domain the learners must acquire the necessary schemata.
But what are schemata? What is this ‘picture of domain’ that a person must have. They (Kirschner, Sweller, Clark) think it is facts, laws, principles, and theories that make up the discipline’s content. But this is simply wrong. It is an incorrect understanding of science. And they ought to know that.
Bloom’s Taxonomy talks about types of learning that are not facts, data, and all theories. Or Michael Polanyi, his book Personal Knowledge, talking about personal knowledge, the difference between ‘knowing that’ and ‘knowing how’. And what’s important about Polanyi’s work is that he says, the bulk of our knowledge, even our conceptual knowledge, is ineffable. That means it cannot be represented in words, which means that a statement (which is a theory) is not a good expression of that knowledge, and that a law is not a good expression of that knowledge. The configuration of connections in your brain, that is a good representation of that knowledge.
Knowing a discipline is knowing the practice of that discipline. It’s learning to think like a scientist or a forester or a hockey player, and learning to recognize, to see the way they see, to speak they way they speak their words, to judge the way they judge.
And that’s what we are producing in the read-write web. The read-write web, the web in which we talked about what we’re doing, we reflected, we practiced, is developing these new kinds of literacies, these new ways of people being able to express how people think - that is, how they talked, how they practiced, how they judged, how they evaluated. There’s a nice picture of this, the revision of Bloom’s verbs that includes verbs tied to the social infrastructure skills and abilities that help people learn and help people grow new mental configurations, grow in knowledge in this environment.
The pedagogy in this environment is based on personal learning. It’s based on the acquisition, the developing, the growth of capacities and aptitudes - like recognition - rather than laws and facts and theories and data. I tried to talk about this a little bit in a paper I presented the last time I was here in Holland called Things You Really Need to Learn. And I talked about the general principle, how to predict consequences. How to stay healthy. How to live a meaningful life. These are the core principles, the core things that people need to learn, not facts and data.
I’ve talked about the semantic principle, the mechanisms, the ways we design our networks in order to make the most reliable system for recognizing patterns in the environment and these principles I put under the headings of Autonomy, Diversity, Openness, and Interactivity. And I’ve talked about the principle of personal learning, the idea here that you develop your own learning yourself. You make your own learning and the way you make your own learning is you find the relevance in the environment. And that has to do with similarity or salience and Amos Tversky, interaction, communicating with your community, and then usability, just being able to speak the language, use the interactions.
That’s the theory that is really supported by, if you will, the cognitive architecture. And so that’s my refutation of Kirshner, Sweller and Clark and my presentation of the alternative connectivist theory of knowledge. And I thank you very much for your time.
---
1. Paul A. Kirschner, John Sweller, Richard E. Clark. 2006. Why Minimal Guidance During Instruction Does Not Work: An Analysis of the Failure of Constructivist, Discovery, Problem-Based, Experiential, and Inquiry-Based Teaching. Educational Psychologist, 41(2), 75-96. Lawrence Erlbaum Associates, Inc. http://www.cogtech.usc.edu/publications/kirschner_Sweller_Clark.pdf
2. George Siemens. 2005. Connectivism: A Learning Theory for a Digital Age. International Journal of Instructional Technology and Distance Learning, Volume 2, Number 1, January, 2005. http://www.itdl.org/Journal/Jan_05/article01.htm
3. Stephen Downes. 2007. Kirschner, Sweller, Clark (2007): Summary. Half an Hour. http://halfanhour.blogspot.com/2007/11/kirschner-sweller-clark-2006-summary.html
4. Stephen Downes. 2007a. Kirschner, Sweller, Clark (2007): Readings. Half an Hour. http://halfanhour.blogspot.com/2007/11/kirschner-sweller-clark-2006-readings_12.html
5. Advanced Distributed Learning. Website. http://www.adlnet.gov/
6. Stephen Downes. 2004. Buntine Oration: Learning Networks. Delivered to the at the Australian College of Educators and the Australian Council of Educational Leaders conference in Perth, Australia, October 8, 2004. http://www.downes.ca/post/20
7. Cindy E. Hmelo-Silver, Ravit Golan Duncan, and Clark A. Chinn. 2006. Scaffolding and Achievement in Problem-Based and Inquiry Learning: A Response to Kirschner, Sweller, and Clark (2006). Educational Psychologist 42(2), 99-107. http://www.cogtech.usc.edu/publications/hmelo_ep07.pdf
8. Mark Rickards. 2005. Teach-yourself computing for kids. BBC News, may 2, 2005. http://news.bbc.co.uk/1/hi/world/south_asia/4498511.stm
9. Deanna Kuhn. 2007. Is Direct Instruction an Answer to the Right Question? Educational Psychologist 42(2), 109-113. http://www.cogtech.usc.edu/publications/kuhn_ep_07.pdf
10. Miles Berry. 2006. Why minimally guided instruction does not work. Weblog. http://elgg.net/mberry/weblog/124841.html
Text of my presentation to SURF Education Days, 13 November 2007, Utrecht, the Netherlands. Slides, audio and video are also available.
1. Introduction
I don’t have fancy slides today. I don’t have nice pictures or anything like that. I’ve spent some time over the last few days looking at a paper by Paul Kirschner, John Sweller and Richard E. Clark, which describes the “failure of constructivist discovery problem-based experiential and inquiry based teaching.â€
For those of you who are familiar with my work you’ll know that a great deal of the work that I’ve done is constructivist discovery problem-based experiential inquiry based teaching. And so this sort of paper is published in Educational Psychologist, which I guess is an important journal. It’s been widely cited.
This sort of paper and the criticisms associated with that sort of paper raise questions. I do work in educational theory and I also do work in software development and I’m never completely sure that I’m doing the right thing. I raise questions. Is the work grounded in research? Is this really the way learning happens? And so I question myself and I question myself not just because it’s a good thing to do, but just because I’m not positive.
So I worry when I see papers like this. I worry that maybe the foundations of the work that I and other people are doing is not well-grounded, but then I look at a paper like this and I realize that what really needs to be done is that these arguments need to be drawn out. They need to be made explicit and it needs to be shown very clearly and not ambiguously why these are not good arguments and that is going to be my task today.
And for you who are listening to this task today, what you’ll be able to take away from this, I believe, is, first of all, an outline an idea of the theory of learning that underlies learning with blogs and wikis and other web 2.0 technologies, but also, more importantly, a way to respond to people who think that instructional technology and the use of the Internet in online learning ought to be nothing more than the presentation of instructions to students telling them what to do. That is control learning. That is the old way of learning. The web way of learning, Web 2.0 learning, is the new way. It is free learning and that is what I advocate.
2. Connectivism (Free Learning)
Sometimes this theory is known under the heading of connectivism. Connectivism is a name that was coined originally by George Siemens. Connectivism is essentially, and this is my take on it not necessarily George’s, the theory that knowledge and learning can be described and explained using network principals.
Now what do I mean by that? What I mean is: knowledge itself, to know something, is to be organized in a certain way, to have a certain pattern of connectivity in the mind, a certain neural connection in the mind. To learn on that theory is therefore to acquire that pattern of organization. To learn is therefore not to have things pushed in your head but to grow and to develop in a certain way and specifically to grow and develop in such a way that you are able to recognize patterns in the environment.
Connectivist learning theory, therefore, is based on the theory of how networks learn, that is to say, how networks grow, how networks develop, how networks form structures of connections between neurons. There are four major ways in which networks grow. And I’m not going to say that these are the only ways, these are all the ways, that this is the definitive statement. But these are ways that we have observed through history that networks grow.
One way is simple Hebbian association. What that means is that if two neurons fire at the same time and don’t fire at the same time, a connection tends to be drawn between them. That’s it. Very simple!
The second way is accidental association. If two neurons are beside each other a connection tends to join them.
The third way connections are formed is back propagation and this comes from the theory of connectionism in the field of computer science. The networks form their connection and then feedback is sent into the network according to the output that the network produces. If the network produces good output the connection will be reinforced. If the connection produces bad output the connection will be broken.
And then finally, Boltzmann learning, which is a theory based on thermodynamics, which says essentially that connections will tend to form at the most stable configuration. If you think of it as like throwing a stone into a pond, the water will settle out. Well connection in brains work in much the same way, according to this theory, and the brain settles out.
The main thing to understand here is that connectivist learning theory is about how connections form in the brain, and for that matter how connections form in networks generally. Because connectivists talk about not simply networks in the brain, they also talk about learning networks in society at large, networks of people in society who are connected to each other. The two theories work out to be two parts of the same theory.
In connectivist pedagogy, therefore, to teach is to model and demonstrate. To teach is to present experiences to people so that they can begin to form these connections in their mind. And then to learn is to actively form these connections by practicing, by repetition, and by reflecting on that practice.
Both of these imply what might be called participation in an authentic community of practice. The idea here is that to learn is to put oneself in a situation where you are practicing in the way that whatever discipline you are in is practicing. For example, you learn physics by doing physics. You learn how to take care of forests by going to a forest.
The role of the teacher in this model is to practice one’s work in an open manner. This has been a challenge, I think, for pretty much all of society, but the idea here is that instead of doing your work in secret in back rooms without being open about what you do, you do your work in an open and transparent manner so that people can see what you’re doing.
In preparation for this talk, for example, I created a summary of the Kirschner, Sweller and Clark paper and I put that on my Website and I collected notes from other people who wrote about this paper and I put it on the website and what I wanted people to see is how I go about assembling my thoughts in order to prepare for a talk like this.
To work, on this theory, is to engage in a community. Most of us, when we do our work, in whatever profession, don’t do it simply all by ourselves. We are involved in a community of practice. We have shared ways of doing things. We have a shared vocabulary, a shared understanding of what constitutes success, a shared understanding of how we test for that success. And it’s to be openly reflective to think about what we’re doing in this community and to think about whether it is the best way to do it, why we’re doing the things that we’re doing.
In this model the role of the learner is to themselves in some sort of environment like this. It may be the actual community of practice itself, which is what I recommend, or it may be a simulation of that community, perhaps a role playing game, perhaps electronic performance support systems, perhaps the actual community of practice itself. You can imagine all the different ways a learner can place themselves into one of those communities of practice. It is to observe the way people who are successful in that practice actually conduct their practice and it is to be reflective to engage in conversations about that practice.
You can see why Web 2.0 and Internet technologies like blogging and wikis and things like social networks play such an important role in these theories. These are technologies that make this sort of activity possible. These are technologies that make it possible for a person to practice their discipline in an open manner. These are technologies that make it possible for a community of practice to develop on a worldwide basis connecting people from many different countries together. They make it possible for the learner to observe experts, to interact with experts, and to learn by doing.
3. Instructivism (Control Learning)
The other view of the world is known as, well, there are different names for it, but I will call it control learning or instructivism, and it is the approach that’s characterized in the Kirschner, Sweller, and Clark’s paper.
It’s a model of learning, and especially on line learning, that is what we might call traditional on line learning. What this learning is, is learning based in the theory of learning objects and based in the theory of the learning management system.
And the model here is that our on line learning environments basically emulate the practices and the processes of the content in a traditional classroom. Learning objects contain content in the core, and pedagogy that is wrapped around that content and the idea of this traditional learning is that the content must be explicitly instructional. The content must actually guide the student or the learner through a series of instructional steps.
The idea of learning object theory is that the learning object needs to have, for example, things like learning objectives. It needs to have guided practice. It needs to have assessment of some sort.
The basis for this theory, the basis for this approach and where it is practiced the most is in environments like the U. S. military where the Sharable Courseware Object Model or SCORM was developed. And SCORM is basically a model of personal learning where you go step by step by step through the learning manual and you follow the instructions that they tell you to follow, you do what they tell you to do and that is what is supposed to produce learning.
The learning management system and we’re familiar with the learning management system, we’re familiar with systems like blackboard, for example, or even open source systems like Moodle and Sakai. The intention of these programs is to present this material step by step by step.
In the introduction (to this talk) the Buntine oration that I gave in Perth is mentioned. In that talk, there are three major locuses or loci of control; three places in which control of the learning process, control of the learning content is exerted in this traditional picture.
One is in content packaging. And if you think about content packaging what content packaging is, is you take a bunch of learning objects, put them all together, put a wrapper around them and then you compress them with a zip archive or something like that, thereby making them useless even to a browser. Content packaging is a picture of learning as a package that you put on a shelf like a book in a library.
Federated search, second. Federated search is not like Google. Federated search, which is the approach that is recommended by the people who created SCORM by advanced distributed learning is a mechanism where when you search you search this library, this library, this library, and this library and that’s it. You search only from recognized authoritative sources who have ‘the knowledge’, whatever that happens to be, authoritative sources, I guess like Educational Psychologist, the journal.
And then, third, learning design, which is a third major component, which actually has its origins here in the Netherlands with Rob Koper’s educational modeling language or EML is a mechanism for stepping you through the presentation of learning materials. Now Learning Design will put you in roles and Learning Design will branch and present different materials in different circumstances, but it’s still that theory of presenting material, presenting material, presenting material, and the theory behind that is if you are told what to do you will learn. And that’s the basics, the basis of instructivism.
4. The Argument
And so now I turn to the paper, that I intend to criticize in this talk; why minimal guidance during instruction does not work and the corollary of that is why maximal guidance does work.
And what the authors do in this paper is they set up two alternatives and on the one side, say the authors, are those people advocating the hypothesis, people like me, that people learn best in an unguided or minimally guided environment.
(Just as an aside every time I quote from them I’ve very sensitive to their use of language. Their use of language is very often loaded or prejudicial and I’m trying to be careful not to be drawn aside by that. “In an unguided or minimally guided environment.†What does that mean?)
Generally defined as one in which learners, rather than being presented with essential information, must discover (now they throw that word discover in there very deliberately because they want to tie it to discovery learning) or construct (there you go with constructivism right) essential (notice that word ‘essential’ - that word has a whole philosophy that comes with it, the whole Saul Kripke essentialism view of the world that there are certain innate natures of things) information for themselves. So that’s the bad thing.
On the other side, this is their view now, are those suggesting that novice learners should be provided with direct instructional guidance on the concepts and procedures required by a particular discipline (we’ll come back to that) and should not be left to discover, (notice how it’s changed a little bit from the first presentation) those procedures (again a very careful word) by themselves.
Just as an aside before I get into the main criticism, I was uncertain, and of course they do not discuss in the paper, what a novice learner is. There’s a certain sense in this paper that a novice learner is any person who has not learned what needs to be learned. And if you read the paper a certain way the paper says if you already know what is being taught you have no problem being taught it, but if you don’t know, then you have to be instructed, properly so called. That would be an unfair reading of the paper, but they leave it open because they don’t tell us what a novice is. Is a novice a baby? Is a novice a 10 year old? Is a novice a first year student in a college or university? We don’t know.
So what do they mean by minimally guided learning? Everything! The old theories of discovery learning, problem based learning, inquiry learning, experiential learning, and of course, the most recent thing, constructivism and, if they had thought of it and felt so inclined, they would have included connectivism and on line learning probably all in the same breath.
There are two assumptions, they say, to this unguided kind of learning. The first assumption is that students should be challenged to solve authentic problems (notice the scare quotes) or inquire information, acquire knowledge (again notice how that is phrased, ‘acquire knowledge,’ I’m going to get it from here and I’m going to put it in here) in information rich settings. And they say (notice again the loaded term, the assumption right, as though we’re just making it up) the assumption is that having learners construct their own solutions, whatever that means leads to the most effective learning experience. And then the second assumption, they say, is that the non-guided people assume that knowledge can best be acquired through experience based on the procedures (the discipline). We’re going to come back to that.
Now their argument, this is their main argument here, it’s a nice categorical syllogism. I like categorical syllogisms because they’re so easy to work with. Any instructional procedure (i.e., ours) that ignores the structures that constitute human cognitive architecture (now there again, ‘human cognitive architecture’, as though the mind is like a house) is not likely to be effective. There, you can see where this is going, right? Minimally guided instruction appears to proceed with no reference to the characteristics of human cognitive architecture. So we’re just one of those things. And indeed, what they mean now (we’ll come back to this later in the talk) by ‘human cognitive architecture’ is the characteristics of working memory, long-term memory and the relations between them. Okay, fine. We’ll come back to that. Thus, they conclude, minimally guided instruction is unlikely to result in effective learning.
You see why I certainly worry about papers like this is because what they’re saying basically is the approach that people who are talking what 2.0 learning, 2.0 blogs, wikis, social practice, communities practice and all that, has no grounding in the theory of how the mind works. And that’s a very serious charge. Turns out to be false but it’s still a very serious charge.
So this is their credo, their manifesto, a.k.a. their conclusion: after a half century of advocacy of minimally guided learning (people like me) it appears there is no body of research supporting the technique. Now, off in the distance you can hear the howls and the wails of protest from the people who have been studying this stuff for 50 years and have found that it works, but we’ll leave that aside.
Insofar that there is any evidence, they argue, it almost uniformly supports direct strong instructional guidance rather than constructivist based minimal guidance. Not only is non-guided instruction less effective (they even talk about this for a bit) it may produce negative results.
And you ask, “How can that be?â€
5. The Reality Check
Well let’s do a reality check first about their conclusion. Their argument, first of all, is simply inconsistent internally. These aren’t major issues, but it’s a bit of a problem.
On the one hand when they’re busy criticizing the minimally guided research they say instructors can’t apply it, they always cheat. They always do some scaffolding. They won’t let students discover things for themselves. They’re always suggesting, telling them what to do and all of that. Okay, fair enough.
But on the other hand, they say that minimally guided instruction is failing. Well either they do it and it doesn’t work, which is bad, or they don’t do it. You can’t say both. You can’t say they’re not doing it, and it doesn’t work.
And then also the section later on in the paper they talk about how minimally guided learning, discovery learning, constructivism, and problem based learning especially, are used in 50 medical schools in the United States. Now I hadn’t heard any particular criticisms about the quality of doctors in the United States. Maybe that’s just a fluke.
But they examine this and they say the strongest criticism they find in those doctors, because the doctors turn out to be fine, but in their paper, they recommend, “unnecessary procedures,†and it strikes me knowing the American medical system, that the last cause of the recommendation of unnecessary medical procedures is the kind of learning that we do. This is the most litigious society in the world. If you drop your pencil you will be sued. That is why they recommend unnecessary medical procedures (whatever those might be) not because they were taught to use problem-based learning.
But be that as it may. Their conclusion is simply not plausible. It doesn’t make sense. It’s not believable. We know that people learn using problem based learning and inquiry based learning. There’s a huge body of research and Hmelo-Silver, Duncan, and Chinn cite numerous studies in their response to the Kirscher, Sweller and Clark paper. And even without that research we know that people learn without guidance because we see it all the time. We have the evidence of our own senses.
Nobody went to school to learn how to build the Internet. Nobody was instructed (‘they said well first you get a website, get it on the web, get some HTML’). People discovered that all by themselves, and it turns out that the mechanisms of computers are things that even small children can learn by themselves.
Many examples, I’ll just point to one. It’s called the ‘hole in the wall gang’ and what they did in cities in India is they would take a computer, they put it literally in a hole in a wall and so the make it assessable to the children in the community. These are children in India, so it’s not like they’re growing up working on their laptops at home. And so they get this computer. They’re not instructed in any way. They learn how to operate it, they learn how to program it and they learn all kind of things about this computer without being told to do anything. So we can see that this works.
Even more to the point, instructivism is a kind of learning by telling; it’s a kind of learning by giving people the information, the concepts, the facts. But we know that people have to learn by practice. Learning is not being told. Learning is doing.
Examples are all over the place. Deanna Kuhn writes that we can hope to impart the smallest fraction of knowledge in any science. How could we possible teach science by teaching facts when there is not literally, not enough time to put all the facts in people’s heads? Even if we were putting them in one after another every second of every day, there’s too many facts in science. We have to do it a different way.
Think about what you would want in medical practice. Would you want a doctor who was told about medicine or a doctor who practiced medicine?
Their argument is based on a straw man. Inquiring learning, problem based learning, are not examples of ‘minimally guided learning’ and again, Hmelo-Silver, Duncan, and Chinn talked about this at some length. They are based on the process of scaffolding, they are based on direct instruction when needed. Indeed, my criticism of problem based learning and inquiry learning is the instructor is too involved. I think there’s too much instructing happening in these kinds of learnings and that there should be less. But that’s a separate argument for a different day.
Their argument is a false dilemma and this is the easy and obvious criticism. They are offering (remember at the very beginning where I presented their argument) the choice between minimally guided instruction or strong instruction. And it begs the question who is doing the choosing, doesn’t it?
This is the example I like to use for this. Imagine my first visit to Rome. I remember getting off the train and walking out and I’m in Rome, in the train station. I have no idea what I’m doing and so a number of things present themselves here, right? I could take a guided tour. I could get a map and walk around myself. I could just walk around aimlessly and never find my hotel or… these are all things where I am choosing what to do, right? And the alternative is, to be kidnapped and to be told where to go.
Now the alternatives are not ‘being kidnapped’ or ‘being lost’. There’s all kinds of ground in between, all kinds of ways that a person can receive guidance that is not in the form of direct instruction. And so their argument is a straw man, yet again.
People who are minimally instructed are in no sense cast adrift. So in preparing this paper for today, nobody told me how to prepare this paper, but that doesn’t mean that I’m sitting all alone. I put out messages in e-mail, I log onto websites, I got all kinds of information back, really helpful useful information that mapped out the territory for me, told me about resources I hadn’t found, pointed to me through objections. I got lots of guidance when I was preparing this talk, but I was not instructed.
And this is a general criticism. This isn’t just me. There’s a social dimension to much learning and Miles Berry points to that in his criticism of the article. A lot of learning, even traditional learning in the classroom, contains a large social dimension as you interact with the other people in your class.
6. Scientific Practice
There are some deeper misunderstandings in this paper as well and I want to explore them. Let’s turn to Kirschner, Swellers, and Clark’s explanation of why learning turned out to be the way it is. And their explanation is inexplicably US-based, but we’ll leave that aside.
They identify the curricular reform that happened after Sputnik (it’s kind of neat, Sputnik happened and then I was born; I came into the world roughly the same time as Sputnik, so I am a product of these reforms, maybe that’s what really scares them). And what the reform is, they say, and it’s repeated throughout the paper, is that it’s based on the assumption that knowledge is best or can only be acquired or learned through experience that is based on the procedures of the discipline. And they repeat this four or five times, I didn’t count them, in the article.
And so they’re saying the assumption here is, if you want to learn physics you should practice physics the way a physicist does, which is what I said at the beginning of this talk. And this has led, they say, to this unguided project work (I did lots of projects when I was in school: Ecuador, the Danube River…) and a rejection of instruction (this is the key phrase, it’s repeated several times in the paper), a rejection of instruction based on the facts, laws, principles, and theories that make up a discipline’s content.
That’s a pretty common view, isn’t it? There’s probably fewer advocating wikis or weblogs or something like that, Wikipedia. People are saying, “what about the facts, what about the laws, what about the content that actually is the discipline, physics, or mathematics or whatever?â€
And they say it may be, they say it is, an error to assume that the pedagogical content of the learning experience is identical to the methods and practices of the discipline being studies. So what they’re saying is that the basic fundamental assumption of my own theory is in error. And it’s a mistake to assume that instruction should focus exclusively on application (well it’s one of those weasel words, ‘exclusively on’, it’s not what they mean, what they mean is it’s a mistake to say that instruction should be application).
Well what do they think? How do they think science works? Because if it’s a mistake to adopt that method, then the nature of that method is pretty important, don’t you think?
Well, happily they explain it in a couple of places for us. One place they explain it is when they discuss Kolb (1971), and Kolb and Fry (1975) and they present a process where a person carried out an action and sees the effect, and then they see this and they understand this effect and begin to anticipate the consequence, and as a consequence of that, they generalize. They understand the general principle.
It sounds like discovery learning. And they extend this to other types of unguided or minimally guided instruction. It’s one of two major components of problem-based learning, they say in their paper: explicit teaching of problem solving strategies in the form of the HD (hypothetical-deductive) method. Barrows & Tamblyn 1980. 1980 is kind of significant because (Thomas) Kuhn was only about 1974 or so.
And then teaching, it’s the same principle. Teaching of the basic content in the context of specific case or instance. So again the same sort of set up here. You have the specific and as a student you’re supposed to generalize. That’s the law. And there’s the problem. That sort of way of going about learning, they argue, might not be the most appropriate way to solve problems. In fact, it’s really difficult, especially in clinical settings, especially in “information rich environments.†It’s really difficult to come up with generalization. You’ve got all kind of convenient hypotheses (that reminded me of Chomsky when he talks about the poverty of the of stimulus). There’s so much information it’s really hard to pick the right generalization.
But the thing is, real science - the stuff that real scientists do in real labs - is not the hypothetical deductive method. Hasn’t been since the ‘50’s. It was developed by Carl Hempel, and the ink wasn’t even dry on the page when Hempel and other logical positivists like A.J. Ayer were being criticized all over the place. Karl Popper, right off the bat, not verification but falsification, which is very much not the HD method. And numerous others. Kuhn - Thomas Kuhn. Lakatos. Lauden. Feyerabend. All pointing to the fact that scientists in practice do not practice the HD model. Nobody does it.
What is science? Science according to Thomas Kuhn (The Structure of Scientific Revolutions) is a community process. The process is not argumentation and it’s specifically not inductive argumentation. It’s explanations. The criteria for explanations are theory based or theory-bound and include things like simplicity, parsimony, testability. And explanations aren’t stand-alone facts. They depend on your expectations; they depend on your theories. As Bas van Fraassen says in The Scientific Image, the explanation of something depends not only what caused something to happen, but also in your expectation of what could have happened instead. Obviously.
7. The Prestige
Let’s turn now to the core of the Kirschner, Sweller and Clark paper, which is the ‘cognitive load argument’. This is where you’re going to give you cognitive psychology and tell you, here’s how learning works.
So their theory is based on the theory of long term and short term memory. They’re not going to be interested in sensory memory (I find that a very interesting statement I’m not going to linger on) and the manner in which our cognitive structures are organized. You’ve heard this before, right? You have short term memory, you have long term memory. Short term memory is the stuff that we are consciously aware of. Long term memory is the stuff we aren’t.
Long term memory is - and these are their words - a ‘massive knowledge base’ and anyone who understands cognitive structures knows how bad a statement that is. If you looked at the structure of the mind, it does not resemble a knowledge base at all. And I won’t get into the details of it, but neural structures and databases are two very different things.
And they say, you’re skillful in an area because your long term memory contains huge amounts of information, so it’s a theory knowledge based on quantity, piled fact on fact on fact. And then they cite – astonishingly – DeGroot’s work on chess expertise. But, you know, I play chess. And there’s this fiction that chess players who are really good chess players can predict ahead, and they can keep all these different board positions in their head. But you cannot defeat a computer like that. The computer will always predict ahead further than you. But people still can beat computers because they visualize, they recognize, successful formations. They don’t memorize a whole bunch of chess positions; they see what is going to work and what is not going to work.
This is Kirschner, Sweller and Clark again: they say the alternate means of instruction is justified by this cognitive theory of long term memory. And the aim of all instruction they say is to alter (nice word there, ‘alter’) long term memory. If nothing has been changed in long term memory they said nothing has been learned. It harkens back to the logical positivist principle of verificationism, that a difference that makes no difference is a difference at all. Or in Bergmann’s formulation, in the dark, all cows are black.
And then the other side of the theory is that working memory, short term memory (and we’ve seen the research on this) has limited capacity, has limited duration, only lasts for a few minutes and is restricted to a small number of elements - the famous number seven elements. We’ve done lots of work on that. Seven elements doesn’t mean seven digits because if you taught to cluster things you can actually remember more than seven. That’s going to form an important part of their theory.
Here’s where the prestige comes in (that’s from magic, right, the movie, you do your trick and then you do the unexpected thing and it all comes together and all the crowd goes ooh ahh). Here’s where it all comes together.
Most learners can, to use their words, construct knowledge. People can actually spot the generalizations. And to do this you must construct a representation. Now this representation, they say, is equally good whether you’re using full information or partial information. So if you have the full information you’re really not losing anything, but if you have partial information, part of your short term memory is occupied in trying to search for possible theories or hypothesis. And because you’re searching for what theory this data could fit into, you’re overloaded, your short term memory is overloaded and you are not able to focus and actually think about what you’re supposed to be learning.
And so, constructivist – because it’s making you search for these theories - is not a good prescriptive instructional designed theory. It’s too hard.
8. Personal Knowledge
Well it’s all in the search, isn’t it? And this is the putative ‘worked example effect’, where you’re doing a search for the hypothesis (imagine a logic problem, “oh what principle am I going to apply, am I going to apply transposition, composition or whatever?â€). So you’re searching these hypotheses, and it’s unrelated to learning, you’re trying to find the best fit, but if you’re studying a worked example, the theory has already been picked for you and you’re looking at what has been done. You’re learning by being told.
The problem is, problem solving - real human problem solving - doesn’t work that way.
We don’t do this mental search of our internal hard drive and try to find the right theory. That’s not how it works at all. We don’t have a whole bunch of general principles or theories stored in our head. Rather we looked at the data and we recognize patterns in that data. It’s a process of seeing rather than searching. This is why is so interesting that at the beginning of this they dropped perceptual memory as though it was completely irrelevant. But perceptual memory is totally relevant because perceptual memory is a process of recognition.
There’s even in the literature there’s discussion of this (I’m not just not making this up). Stephen Kosslyn, for example, image-rotating examples, the theory where you’re going through algorithms and processes does not explain reaction time when you’re asking people to visualize the rotation of images. We work on a sub-symbolic level, not a symbolic level. Cognition is based on a process of pattern recognition at that sub-symbolic level.
And we know this. There has been a lot of discussion about experts and schemas and how an expert will acquire a frame – to use Lakoff’s term – or knowledge of organization. And they even admit, the authors even admit, cognitive research has shown that to acquire expertise in a domain the learners must acquire the necessary schemata.
But what are schemata? What is this ‘picture of domain’ that a person must have. They (Kirschner, Sweller, Clark) think it is facts, laws, principles, and theories that make up the discipline’s content. But this is simply wrong. It is an incorrect understanding of science. And they ought to know that.
Bloom’s Taxonomy talks about types of learning that are not facts, data, and all theories. Or Michael Polanyi, his book Personal Knowledge, talking about personal knowledge, the difference between ‘knowing that’ and ‘knowing how’. And what’s important about Polanyi’s work is that he says, the bulk of our knowledge, even our conceptual knowledge, is ineffable. That means it cannot be represented in words, which means that a statement (which is a theory) is not a good expression of that knowledge, and that a law is not a good expression of that knowledge. The configuration of connections in your brain, that is a good representation of that knowledge.
Knowing a discipline is knowing the practice of that discipline. It’s learning to think like a scientist or a forester or a hockey player, and learning to recognize, to see the way they see, to speak they way they speak their words, to judge the way they judge.
And that’s what we are producing in the read-write web. The read-write web, the web in which we talked about what we’re doing, we reflected, we practiced, is developing these new kinds of literacies, these new ways of people being able to express how people think - that is, how they talked, how they practiced, how they judged, how they evaluated. There’s a nice picture of this, the revision of Bloom’s verbs that includes verbs tied to the social infrastructure skills and abilities that help people learn and help people grow new mental configurations, grow in knowledge in this environment.
The pedagogy in this environment is based on personal learning. It’s based on the acquisition, the developing, the growth of capacities and aptitudes - like recognition - rather than laws and facts and theories and data. I tried to talk about this a little bit in a paper I presented the last time I was here in Holland called Things You Really Need to Learn. And I talked about the general principle, how to predict consequences. How to stay healthy. How to live a meaningful life. These are the core principles, the core things that people need to learn, not facts and data.
I’ve talked about the semantic principle, the mechanisms, the ways we design our networks in order to make the most reliable system for recognizing patterns in the environment and these principles I put under the headings of Autonomy, Diversity, Openness, and Interactivity. And I’ve talked about the principle of personal learning, the idea here that you develop your own learning yourself. You make your own learning and the way you make your own learning is you find the relevance in the environment. And that has to do with similarity or salience and Amos Tversky, interaction, communicating with your community, and then usability, just being able to speak the language, use the interactions.
That’s the theory that is really supported by, if you will, the cognitive architecture. And so that’s my refutation of Kirshner, Sweller and Clark and my presentation of the alternative connectivist theory of knowledge. And I thank you very much for your time.
---
1. Paul A. Kirschner, John Sweller, Richard E. Clark. 2006. Why Minimal Guidance During Instruction Does Not Work: An Analysis of the Failure of Constructivist, Discovery, Problem-Based, Experiential, and Inquiry-Based Teaching. Educational Psychologist, 41(2), 75-96. Lawrence Erlbaum Associates, Inc. http://www.cogtech.usc.edu/publications/kirschner_Sweller_Clark.pdf
2. George Siemens. 2005. Connectivism: A Learning Theory for a Digital Age. International Journal of Instructional Technology and Distance Learning, Volume 2, Number 1, January, 2005. http://www.itdl.org/Journal/Jan_05/article01.htm
3. Stephen Downes. 2007. Kirschner, Sweller, Clark (2007): Summary. Half an Hour. http://halfanhour.blogspot.com/2007/11/kirschner-sweller-clark-2006-summary.html
4. Stephen Downes. 2007a. Kirschner, Sweller, Clark (2007): Readings. Half an Hour. http://halfanhour.blogspot.com/2007/11/kirschner-sweller-clark-2006-readings_12.html
5. Advanced Distributed Learning. Website. http://www.adlnet.gov/
6. Stephen Downes. 2004. Buntine Oration: Learning Networks. Delivered to the at the Australian College of Educators and the Australian Council of Educational Leaders conference in Perth, Australia, October 8, 2004. http://www.downes.ca/post/20
7. Cindy E. Hmelo-Silver, Ravit Golan Duncan, and Clark A. Chinn. 2006. Scaffolding and Achievement in Problem-Based and Inquiry Learning: A Response to Kirschner, Sweller, and Clark (2006). Educational Psychologist 42(2), 99-107. http://www.cogtech.usc.edu/publications/hmelo_ep07.pdf
8. Mark Rickards. 2005. Teach-yourself computing for kids. BBC News, may 2, 2005. http://news.bbc.co.uk/1/hi/world/south_asia/4498511.stm
9. Deanna Kuhn. 2007. Is Direct Instruction an Answer to the Right Question? Educational Psychologist 42(2), 109-113. http://www.cogtech.usc.edu/publications/kuhn_ep_07.pdf
10. Miles Berry. 2006. Why minimally guided instruction does not work. Weblog. http://elgg.net/mberry/weblog/124841.html
