Our first example is a single scene from Jurassic Park, set entirely in the control room of Isla Nublar. Apologies in advance for repeating some material already covered by the book and website, but it is necessary to focus on the aspects that are of interest to this study.
The eponymous Jurassic Park is heavily automated, with the entire park designed to be controlled from the computer systems in this room. Villainous computer system designer Nedry took advantage of this to shut down systems across the entire park, releasing all the dinosaurs, to cover his industrial espionage. Most of the park staff had already been evacuated due to a storm warning, and the small team of core technical staff who remained have, by this point in the film, all been killed by dinosaurs. (Including Nedry—who, had he been given time for extrospection, would probably have rethought those aspects of his plan concerning the release of carnivorous dinosaurs.)
Four of the survivors have gathered in the control room after managing to restore the power, but must still restart the various computer systems. They have discovered that the computer control extends down to door locks, which are consequently not working and have suddenly become the number one priority due to the velociraptors trying to break in.
Our interface user is Lex, a teenage visitor, being given an advance tour of the park before its official opening. The others are Dr Grant, paleontologist; Dr Sattler, paleobotanist; and Lex’s younger brother Tim, dinosaur enthusiast. As a self -described computer hacker Lex is easily the best person qualified to work with the computers as everyone else in the room only has expertise in subjects more than sixty-six million years old.
The computers were all rebooted when the power came back on but the programs that control Jurassic Park did not automatically restart. Dr. Sattler spent a moment in front of the computer with Lex, but all she seemed to do is deactivate the screen saver. It’s up to Lex to find and start whatever program runs the security systems for the control room.
Backworlding aside: Unix-savvy viewers might be wondering why these control programs, since they are critical to the park functionality, don’t automatically start when the computer is rebooted. I hazard that perhaps normally they would, but Nedry turned this off to ensure that no-one could undo his sabotage before he got back.
The file system of the computer is rendered as a tree, with directory names (/usr in the image above) shown as text labels, the contents of each directory shown as LEGO-like blocks, and lines linking directories to subdirectories.
Most of the information is drawn on a flat two-dimensional plane. The third dimension is used to present information about the number of, and perhaps sizes, of the files in each directory. Note in the image above that the different directories below the foremost park block have different sized heights and areas.
Rendering this plane in perspective, rather than as a conventional 2D window, means that areas closest to the viewpoint can be seen in detail, but there is still some information given about the directories further away. In the image above, the subdirectory of park on the right is clearly smaller than the others, even though we can’t make out the actual name, and also has a number of larger subdirectories.
Up close we can see that each file can have its own icon on top, presumably representing the type of file.
The viewpoint stays at a constant height above the ground plane. Moving around is done with the mouse, using it as a game-style directional controller when the mouse button is held down rather than as an absolute pointing device. It is almost “walking” rather than “flying” but there is a slight banking effect when Lex changes direction.
Here Lex has browsed through the hierarchy and discovered a promising file. She selects it, but we don’t see how, and a spotlight or sunbeam indicates the selection.
This is the last of the 3D interactions. The 3D file browser is just a file browser, not an entire operating system or virtual environment, so opening a file or program will open a new interface.
Tagged: 3D, 3D rendering, blue, cathode ray tube, color, comparison, constant movement, control room, cyan, desk, direct manipulation, disambiguation, finger press, flight control, flying, green, icon, interaction design, light, lighting, map, missing information, motion cue, navigating, pink, point to select, projection rays, selection, sense making, stress, up is more
When Lex runs this program (again, we don’t see how) it is in fact the security system controller for the visitor centre, including the control room. This has a conventional 2D GUI interface and she immediately switches everything on.
Success! Well, it would be if the control room did not also have very large windows which are apparently not velociraptor-proof. But the subsequent events, and interfaces, are not our concern.
Analysis
This isn’t a report card, since those are given to complete films or properties, not individual interfaces. But we can ask the same questions.
How believable is the interface?
In this case, very believable. The 3D file browser seen in the film is a real program that was shipped with the computers used in the film. It was created by the manufacturer Silicon Graphics as a demonstration of 3D capabilities, not as an effect just for this film.
How well does the interface inform the narrative of the story?
It supports the narrative, but isn’t essential — there’s plenty of drama and tension due to the velociraptors at the door, and the scene would probably still work if the camera only showed Lex, not the interface. The major contribution of using the 3D file browser is to keep the technology of Jurassic Park seemingly a little more advanced than normal for the time. Apart from dinosaurs, both the book and the film try not to introduce obviously science fictional elements. A 2D file browser (they did exist for Unix computers at the time, including the SGI computers shown in the film) would have been recognisable but boring. The 3D file browser looks advanced while still being understandable.
How well does the interface equip the characters to achieve their goals?
The most interesting question, to which the answer is that it works very well. One problem, visible in the film, is that because the labels are rendered on the 2D ground plane, users have to navigate close to a file or a folder to read its name. Rotating the names to vertical and to always face the user (“billboarding”) would have made them recognisable from further away.
Both at the time of the film and today some computer people will argue that Lex can’t be a real computer hacker because she doesn’t use the command line interface. Graphical user interfaces are considered demeaning. I disagree.
Lex is in a situation familiar to many system administrators, having to restore computer functionality after an unexpected power loss. (Although the velociraptors at the door are a little more hostile than your typical user demanding to know when the system will be back up.) Earlier in the film we saw Ray Arnold, one of the technical staff, trying to restore the system and he was using the command line interface.
So why does Lex use the 3D file browser? Because, unlike Ray Arnold, she doesn’t know which programs to run. Rebooting the computers is not enough. The various programs that control Jurassic Park are all custom pieces of software developed by Nedry, and nothing we’ve seen indicates that he would have been considerate enough to write a user guide or reference manual or even descriptive file names. Everyone who might have known which programs do what is either dead or off the island.
Lex needs an interface that lets her quickly search through hundreds or even thousands of files without being able to specify precise search criteria. For a problem involving recognition, “you’ll know it when you see it”, a graphical user interface is superior to a command line.
Film making challenge: diegetic computers
Writing for SciFiInterfaces can be quite educational. Chris asked me to write about the “diegetic” aspects of rendering 3D graphics in film, and I agreed to do so without actually knowing what that meant. Fortunately for me it isn’t complicated. Diegetic images or sounds belong to what we see in the scene itself, for instance characters and their dialog or hearing the music a violinist who is on-screen is playing; while non-diegetic are those that are clearly artefacts of watching a film, such as subtitles, voice overs, or the creepy violin music that is playing as a character explores a haunted house—we don’t imagine there is some violinist in there with them.
So, SciFiinterfaces.com focuses on the diegetic computer interfaces used by characters within the film or TV show itself. We’ve just been discussing the 3D file browser in Jurassic Park. Which, since it was a real interactive program, just meant pointing a camera at the actor and the computer screen, right?
It’s not that easy. Our human eyes and brain do an enormous amount of interpolation and interpretation of what we actually see. There’s the persistence of vision effect that allows us to watch a film in a cinema and see it as fluid motion, even though for a significant percentage of the time we’re actually looking at a blank wall while the projector shutter is closed. Cameras, whether film or digital, take discrete snapshots and are not so easily fooled, leading to various odd effects. One example that’s been known since the early days of filmmaking is that at certain speeds spoked wheels can appear to be rotating far more slowly than expected, or even to be rotating backwards.
Jurassic Park was made in the days when television sets and computer monitors used Cathode Ray Tube (CRT) technology. A CRT cannot display an entire frame at once, instead starting at the top left and drawing pixels line by line (“scan lines”) to the bottom. Just as the top line of pixels fades out, the new frame begins. At 50 or 60 frames a second we see only continuous moving images thanks to our persistence of vision; but a camera, usually running at 24 frames a second, will capture a dark line moving slowly down the screen and the images themselves will flicker. This was a common sight in TV news reports and sometimes in films of the time, when computer monitors were in the background. Here’s a shot from the 1995 film The Net where the new frames have been half-drawn:
One technique that avoids this is to film the computer interface in isolation and composite the graphics into the footage afterwards. This is very easy in the 21st century with all digital processing but Jurassic Park was made in the days of optical compositing, which is more expensive and limits the number of images that can be combined before losing picture quality.
So to shoot CRT monitors with their graphics live, the camera shutter opening must be synchronised to the start of each frame. In TV studios and film sets this is done with genlocking, connecting all the monitors and cameras via cables to a single electronic timing signal. This was apparently the technique used in Jurassic Park, with impressive results. In one control room scene the camera pans across at least eight different monitors, and none of them are flickering.