When Frito is driving Joe and Rita away from the cops, Joe happens to gesture with his hand above the car window, where a vending machine he happens to be passing spots the tattoo. Within seconds two harsh beeps sound in the car and a voice says, “You are harboring a fugitive named NOT SURE. Please, pull over and wait for the police to incarcerate your passenger.”
Frito’s car begins slowing down, and the dashboard screen shows a picture of Not Sure’s ID card and big red text zooming in a loop reading “PULL OVER”
The car interface has a column of buttons down the left reading:
At the bottom is a square of icons: car, radiation, person, and the fourth is obscured by something in the foreground. Across the bottom is Frito’s car ID “FRITO’S F’N CAR” which appears to be a label for a system status of “EVERYTHING’S A-OK, BRO”, a button labeled CHECK INGN [sic], another labeled LOUDER, and a big green circle reading GO.
But the car doesn’t wait for him to pull over. With some tiny beeps it slows to a stop by itself. Frito says, “It turned off my battery!” Moments after they flee the car, it is converged upon by a ring of police officers with weapons loaded (including a rocket launcher pointed backward.)Continue reading →
The initial invasion of Klendathu is disastrous, and our hero Rico suffers a massive penetration wound in combat, with an Arachnid digging its massive, thorn-like pincer straight through his thigh.
Rico is (spoiler alert) mistakenly reported as deceased. (There’s perhaps some argument for outfitting soldiers with networked biometrics so this sort of mistake can’t be made, but that’s another post.)
After returning to dock, Ibanez hears reports about the military disaster, and sees a death roll scrolling by on a large wall display. Three columns of off-white names tick along, surname first, with an initialism indicating whether the soldier was killed, wounded, or missing in action. At the very top three legends summarize key information, WOUNDED IN ACTION 2,548; KILLED IN ACTION 205,515; and MISSING IN ACTION 105,753. Largest of all is the KLENDATHU CASUALTIES: 308,563. (I know, the math doesn’t add up. It’s possible I misread the blurry numbers.) But the screen could use some more deliberate graphic design. Continue reading →
After Ibanez explains that the new course she plotted for the Rodger Young (without oversight, explicit approval, or notification to superiors) is “more efficient this way,” Barcalow walks to the navigator’s chair, presses a few buttons, and the computer responds with a blinking-red Big Text Label reading “COURSE OPTIMAL” and a spinning graphic of two intersecting grids.
Yep, that’s enough for a screed, one addressed first to sci-fi writers.
A plea to sci-fi screenwriters: Change your mental model
Think about this for a minute. In the Starship Troopers universe, Barcalow can press a button to ask the computer to run some function to determine if a course is good (I’ll discuss “good” vs. “optimal” below). But if it could do that, why would it wait for the navigator to ask it after each and every possible course? Computers are built for this kind of repetition. It should not wait to be asked. It should just do it. This interaction raises the difference between two mental models of interacting with a computer: the Stoic Guru and the Active Academy.
Stoic Guru vs. Active Academy
This movie was written when computation cycles may have seemed to be a scarce resource. (Around 1997 only IBM could afford a computer and program combination to outthink Kasparov.) Even if computation cycles were scarce, navigating the ship safely would be the second most important non-combat function it could possibly do, losing out only to safekeeping its inhabitants. So I can’t see an excuse for the stoic-guru-on-the-hill model of interaction here. In this model, the guru speaks great truth, but only when asked a direct question. Otherwise it sits silently, contemplating whatever it is gurus contemplate, stoically. Computers might have started that way in the early part of the last century, but there’s no reason they should work that way today, much less by the time we’re battling space bugs between galaxies.
A better model for thinking about interaction with these kinds of problems is as an active academy, where a group of learned professors is continually working on difficult questions. For a new problem—like “which of the infinite number of possible courses from point A to point B is optimal?”—they would first discuss it among themselves and provide an educated guess with caveats, and continue to work on the problem afterward, continuously, contacting the querant when they found a better answer or when new information came in that changed the answer. (As a metaphor for agentive technologies, the active academy has some conceptual problems, but it’s good enough for purposes of this article.)
Consider this model as you write scenes. Nowadays computation is rarely a scarce resource in your audience’s lives. Most processors are bored, sitting idly and not living up to their full potential. Pretending computation is scarce breaks believability. If ebay can continuously keep looking on my behalf for a great deal on a Ted Baker shirt, the ship’s computer can keep looking for optimal courses on the mission’s behalf.
In this particular scene, the stoic guru has for some reason neglected up to this point to provide a crucial piece of information, and that is the optimal path. Why was it holding this information back if it knew it? How does it know that now? “Well,” I imagine Barcalow saying as he slaps the side of the monitor, “Why didn’t you tell me that the first time I asked you to navigate?” I suspect that, if it had been written with the active academy in mind, it would not end up in the stupid COURSE OPTIMAL zone.
Optimal vs. more optimal than
Part of the believability problem of this particular case may come from the word “optimal,” since that word implies the best out of all possible choices.
But if it’s a stoic guru, it wouldn’t know from optimal. It would just know what you’d asked it or provided it in the past. It would only know relative optimalness amongst the set of courses it had access to. If this system worked that way, the screen text should read something like “34% more optimal than previous course” or “Most optimal of supplied courses.” Either text could show some fuigetry that conveys a comparison of compared parameters below the Big Text Label. But of course the text conveys how embarrassingly limited this would be for a computer. It shouldn’t wait for supplied courses.
If it’s an active academy model, this scene would work differently. It would have either shown him optimal long ago, or show him that it’s still working on the problem and that Ibanez’ is the “Most optimal found.” Neither is entirely satisfying for purposes of the story.
How could this scene gone?
We need a quick beat here to show that in fact, Ibanez is not just some cocky upstart. She really knows what’s up. An appeal to authority is a quick way to do it, but then you have to provide some reason the authority—in this case the computer—hasn’t provided that answer already.
A bigger problem than Starship Troopers
This is a perennial problem for sci-fi, and one that’s becoming more pressing as technology gets more and more powerful. Heroes need to be heroic. But how can they be heroic if computers can and do heroic things for them? What’s the hero doing? Being a heroic babysitter to a vastly powerful force? This will ultimately culminate once we get to the questions raised in Her about actual artificial intelligence.
Fortunately the navigator is not a full-blown artificial intelligence. It’s something less than A.I., and that’s an agentive interface, which gives us our answer. Agentive algorithms can only process what they know, and Ibanez could have been working with an algorithm that the computer didn’t know about. She’s just wrapped up school, so maybe it’s something she developed or co-developed there:
Barcalow turns to the nav computer and sees a label: “Custom Course: 34% more efficient than models.”
Um…OK…How did you find a better course than the computer could?
My grad project nailed the formula for gravity assist through trinary star systems. It hasn’t been published yet.
BAM. She sounds like a badass and the computer doesn’t sound like a character in a cheap sitcom.
So, writers, hopefully that model will help you not make the mistake of penning your computers to be stoic gurus. Next up, we’ll discuss this same short scene with more of a focus on interaction designers.
Carl, a young psychic, has an application at home to practice and hone his mental powers. It’s not named in the film, so I’m going to call it DuoMento. We see DuoMento in use when Carl uses it to try and help Johnny find if he has any latent psyhic talent. (Spoiler alert: It doesn’t work.)
DuoMento challenges its users with blind matching tests. For it, the “thought projector” (Carl) sits in a chair at a desk with a keyboard and a desktop monitor before him. The “thought receiver” (Johnny) sits in a chair facing the thought projector, unable to see either the desktop monitor or the large, wall-mounted screen behind him, which duplicates the image from the desktop monitor. To the receiver’s right hand is a small elevated panel of around 20 white push buttons.
For the test, two Hoyle playing cards appear on the screen side-by-side, face down. Carl presses a key on his keyboard, and one card flips over to reveal its face. Carl concentrates on the face-up card, attempting to project the identity of the card to Johnny. Johnny tries his best to receive the thought. It’s intense.
When Johnny feels he has an answer, he says, “I see…Ace of Spades,” and reaches forward and presses a button on the elevated panel. In response, the hidden card flips over as the ace of spades. An overlay appears on top of the two cards indicating if it was a match. Lacking any psychic abilities, Johnny gets a big label reading “NO MATCH,” accompanied by a buzzer sound. Carl resets it to a new card with three clicks on his keyboard.
Not very efficient
Why does it take Carl three clicks to reset the cards? You’d think on such a routine task it would be as simple as pressing [space bar]. Maybe you want to prevent accidental activation, but still that’s a key with a modifer, like shift+[space bar]. Best would be if Carl was also a telekinetic. Then he could just mentally push a switch and get some of that practice in. If that switch offered variable resistance it could increase with each…but I digress since he’s just a telepath.
A semi-questionable display
I get why there’s a side-by-side pair of cards. People are much better at these sorts of comparison tasks when objects are side-by-side. But ultimately, it conveys the wrong thing. Having a face down card that flips over implies that that face-down card is the one that Johnny’s trying to guess. But it’s not. The one that’s already turned over is the one he’s trying to guess. Better would be a graphic that implies he’s filling in the blank.
Better still are two separate screens: One for the projector with a single card displayed, and a second for the receiver with this same graphic prompting him to guess. This would require a little different setup when shooting the scene, with over-the-shoulder shots for each showing the different screen. But audiences are sophisticated enough to get that now. Different screens can show different things.
At first it seems like Johnny’s input panel is insufficient for the task. After all, there are 52 cards in a standard deck of cards and only 20 buttons. But having a set of 13 keys for the card ranks and 4 for the suit is easy enough, reduces the number of keys, and might even let him answer only the part he’s confident in if the image hasn’t quite come through.
Does it help test for “sensitivity”?
Psychic powers are real in the world of Starship Troopers, so we’re going not going to question that. Instead the question at hand will be: Is this the best test for psychic sensitivity?
I do wonder that having a lit screen gives the receiver a reflection in the projector’s eyes to detect, even if unconsciously. An eagle-eyed receiver might be able to spot a color, or the difference between a face card and a number card. Better would be some way for the projector to cover his eyes while reading the subject, and dim that screen afterward.
The risk of false positives
More importantly, such a test would want to eliminate the chance that the receiver guessed correctly by chance. The more constrained and familiar the range of options, the more likely they are to get a false positive, which wouldn’t help anything except confidence, and even that would be false. I get that when designing skills-building interfaces, you want to start easy and get progressively more challenging. But it makes more sense to constrain the concepts being projected to things that are more concrete and progress to greater abstraction or more nuance. Start with “fire,” perhaps, and advance to “flicker” or “warmth.” For such thoughts, a video cue of a word randomly selected from that pool of concepts would make the most sense. And for cinematic directness (Starship Troopers was nothing if not direct) you should overlay the word onto the video cue as well.
The next design challenge then becomes how does the receiver provide to the system what, if anything, they’re receiving. Since the concepts would be open-ended, you need a language-input mechanism: ANSI keyboard for typing, or voice recognition.
Additionally, I’d add a brain-reading interface that was able to read his brain as he was attempting to receive. Then it could detect for the right state of mind, e.g. an alpha state, as well as areas of the brain that are being activated. Cinematically you could show a brain map, indicating the brain state in a range, the areas of the brain being activated. Having the map on hand for Johnny would let him know to relax and get into a receptive state. If Carl had the same map he could help prompt him.
In a movie you’d probably also want a crude image feed being “read” from Johnny’s thoughts. It might charmingly be some dumb, non-fire things, like scenes from his last jump ball game, Carmen’s face and cleavage, and to Carl’s shame, a recollection of the public humilation suffered recently at his hand.
But if this interface (and telepathy) was real, you wouldn’t want to show that to Johnny, as it might cause distracting feedback loops, and you wouldn’t want to show it to Carl less he betray when Johnny is getting close, and encourage Johnny’s zeroing in on the concept through subtle social cues instead of the desired psychic ones. Since it’s not real, let’s comp it up next more cinematically.
While preparing for his night cycle, Wall-E is standing at the back of his transport/home. On the back drop door of the transport, he is cleaning out his collection cooler. In the middle of this ritual, an alert sounds from his external speakers. Concerned by the sound, Wall-E looks up to see a dust storm approaching. After seeing this, he hurries to finish cleaning his cooler and seal the door of the transport.
A Well Practiced Design
The Dust Storm Alert appears to override Wall-E’s main window into the world: his eyes. This is done to warn him of a very serious event that could damage him or permanently shut him down. What is interesting is that he doesn’t appear to register a visual response first. Instead, we first hear the audio alert, then Wall-E’s eye-view shows the visual alert afterward.
Given the order of the two parts of the alert, the audible part was considered the most important piece of information by Wall-E’s designers. It comes first, is unidirectional as well as loud enough for everyone to hear, and is followed by more explicit information.
Equal Opportunity Alerts
By having the audible alert first, all Wall-E units, other robots, and people in the area would be alerted of a major event. Then, the Wall-E units would be given the additional information like range and direction that they need to act. Either because of training or pre-programmed instructions, Wall-E’s vision does not actually tell him what the alert is for, or what action he should take to be safe. This could also be similar to tornado sirens, where each individual is expected to know where they are and what the safest nearby location is.
For humans interacting alongside Wall-E units each person should have their own heads-up display, likely similar to a Google-glass device. When a Wall-E unit gets a dust storm alert, the human could then receive a sympathetic alert and guidance to the nearest safe area. Combined with regular training and storm drills, people in the wastelands of Earth would then know exactly what to do.
Why Not Network It?
Whether by luck or proper programming, the alert is triggered with just enough time for Wall-E to get back to his shelter before the worst of the storm hits. Given that the alert didn’t trigger until Wall-E was able to see the dust cloud for himself, this feels like very short notice. Too short notice. A good improvement to the system would be a connection up to a weather satellite in orbit, or a weather broadcast in the city. This would allow him to be pre-warned and take shelter well before any of the storm hits, protecting him and his solar collectors.
Other than this, the alert system is effective. It warns Wall-E of the approaching storm in time to act, and it also warns everyone in the local vicinity of the same issue. While the alert doesn’t inform everyone of what is happening, at least one actor (Wall-E) knows what it means and knows how to react. As with any storm warning system, having a connection that can provide forecasts of potentially dangerous weather would be a huge plus.
After her escape from the nucleolab, Leeloo ends up on a thin ledge of a building, unsure where to go or what to do. As a police car hovers nearby, the officers use an onboard computer to try and match her identity against their database. One officer taps a few keys into an unseen keyboard, her photograph is taken, and the results displays in about 8 seconds. Not surprisingly, it fails to find a match, and the user is told so with an unambiguous, red NO FILE banner across the screen.
This interface flies by very quickly, so it’s not meant to be read screen by screen. Still, the wireframes present a clear illustration of what the system doing, and what the results are.
The system shouldn’t just provide dead ends like this, though. Any such system has to account for human faces changing over the time since the last capture: aging, plastic surgery, makeup, and disfiguring accidents, to name a few. Since Leeloo isn’t inhuman, it could provide some results of “closest matches,” perhaps with a confidence percentage alongside individual results. Even if the confidence number was very low, that output would help the officers understand it was an issue with the subject, and not an issue of an incomplete database or weak algorithm.
One subtle element is that we don’t see or hear the officer telling the system where the perp is, or pointing a camera. He doesn’t even have to identify her face. It automatically finds her in the camera few, identifies her face, and starts scanning. The sliding green lines tell the officer what it’s finding, giving him confidence in its process, and offering an opportunity to intervene if it’s getting things wrong.
The taxi has a screen on the passenger’s side dashboard that faces the driver. This display does two things. First, it warns the driver when the taxi is about to be attacked. Secondly, it helps him navigate the complexities of New York circa 2163.
After Korben decides to help Leeloo escape the police, they send a squadron of cop cars to apprehend them. And by apprehend I mean blow to smithereens. The moment Korben’s taxi is in sights, they don’t try to detain or disable the vehicle, but to blast it to bits with bullets and more bullets. It seems this is a common enough thing to have happen that Korben’s on-board computer can detect it in advance and provide a big, flashing, noisemaking warning to this effect.
In many cases I object to the Big Label, but not here. In fact, for such a life-threatening issue, more of the taxi’s interface should highlight the seriousness. My life’s in danger? Go full red alert, car. Change the lights to crimson. Dim non-essential things. You’ve got an “automatic” button there. Does that include evasive maneuvering? If so, make that thing opt-out rather than opt-in. Help a brother out.
At other times during the chase scene, Korben can glance at the screen to see a wireframe of the local surroundings. This interface has a lot of problems.
1. This would work much, much more safely and efficiently for Korben if it was a heads-up display on the windshield. Let’s shrink that feedback loop. Every time a driver glances down he risks a crash and in this case, Korben risks the entire world. If HUD tech isn’t a part of the diegesis, audio cues might be some small help that don’t require him to take his eyes of the “road.”
2. How does the wireframe style help? It’s future-y of course, but it adds a lot of noise to what he’s got to process. He doesn’t need to understand tesselations of surfaces. He needs to understand the shapes and velocities of things around him so he can lose the tail.
(Exercise for the reader: Provide a solid diegetic explanation for why this screen appeared in the film flipped horizontally.)
3. There’s some missing information. If the onboard computer can do some real-time calculations and make a recommendation on the best next step, why not do it? We see above that the police have the same information that Korben does. So even better might be information on what the tail is likely to do so Korben can do the opposite. Or maneuvers that Korben can execute that the cop car can’t. If it’s possible to show places he should definitely not go, like dead ends or right into the path, say, of a firing squad of police cars, that would be useful to know, too.
4. What are those icons in the lower right meant to do? They’re not suggestions as they appear after Korben performs his maneuvers, and sometimes appear along with warnings instead of maneuvers.
Even if they are suggestions, what are they directions to? His original destination? He didn’t have one. Some new destination? When did he provide it? Simple, goal-aware directions to safety? Whatever the information, these icons add a lot cognitive weight and visual work. Surely there’s some more direct way to provide cues, like being superimposed on the 3D so he can see the information rather than read and interpret it.
If they’re something else other than suggestions, they’re just noise. In a pursuit scenario, you’d want to strip that stuff out of the interface.
5. What is that color gradient on the left meant to tell him? All the walls in this corridor are 350…what? The screen shot above hints that it represents simple height from the ground, but the 2D map has these colors as well, and height cues wouldn’t make sense there. If it is height, this information might help Korben quickly build a 3D mental map of the information he’s seeing. But using arbitrary colors forces him to remember what each color means. Better would be to use something with a natural order to it like the visible spectrum or black-body spectrum. Or, since people already have lots of experience with monocular distance cues and lighting from above, maybe a simple rendering as if the shapes were sunlit would be fastest to process. Taking advantage of any of these perceptual faculties would let him build a 3D model quickly so he can focus on what he’s going to do with the information.
Side note: Density might actually make a great deal more sense to the readout, knowing that Korben has a penchant for ramming his taxi through things. If this was the information being conveyed, varying degrees of transparency might have served him better to know what he can smash through safely, and even what to expect on the other side.
6. Having the 2D map helps a bit to understand the current level of the city from a top-down view. Having it be small in the upper right is a sound placement, since that’s a less-important subset of the information he really needs. It has some color coding but as mentioned above it doesn’t seem to relate to what’s colored in the 3D portion, which could make for an interpretation disaster. In any case, Korben shouldn’t have to read this information in the tiny map. It’s a mode, a distraction. While he’s navigating the alleys and tunnels of the city, he’s thinking in a kind of 3D node-graph. Respect that kind of thinking with a HUD that puts information on the “edges” of the graph, i.e., the holes in the surfaces around him that he’s looking at. That’s his locus of attention. That’s where he’s thinking. Augment that.
So, you know…bad
Fortunately, given that the interface has so many problems, Korben only really glances at this once during the chase, and that’s at the warning sound. But if the younger Korben was meant to use this at all, there’s a lot of work to make this useful rather than dangerous.
Later in the scene General Staedert orders a “thermonucleatic imaging.” The planet swallows it up. Then Staedert orders an “upfront loading of a 120-ZR missile” and in response to the order, the planet takes a preparatory defensive stance, armoring up like a pillbug. The scanner screens reflect this with a monitoring display.
In contrast to the prior screen for the Gravity (?) Scan, these screens make some sense. They show:
A moving pattern on the surface of a sphere slowing down
clear Big Label indications when those variables hit an important threshold, which is in this case 0
A summary assessment, “ZERO SURFACE ACTIVITY”
A key on the left identifying what the colors and patterns mean
Some sciency scatter plots on the right
The majority of these would directly help someone monitoring the planet for its key variables.
Though these are useful, it would be even more useful if the system would help track these variables not just when they hit a threshold, but how they are trending. Waveforms like the type used in medical monitoring of the “MOVEMENT LOCK,” “DYNAMIC FLOW,” and “DATA S C A T” might help the operator see a bit into the future rather than respond after the fact.
The first bit of human technology we see belongs to the Federation of Territories, as a spaceship engages the planet-sized object that is the Ultimate Evil. The interfaces are the screen-based systems that bridge crew use to scan the object and report back to General Staedert so he can make tactical decisions.
We see very few input mechanisms and very little interaction with the system. The screen includes a large image on the right hand side of the display and smaller detailed bits of information on the left. Inputs include
Rows of backlit modal pushbuttons adjacent to red LEDs
The operator of the first of these screens touches one of the pushbuttons to no results. He then scrolls the trackball downward, which scrolls the green text in the middle-left part of the screen as the graphics in the main section resolve from wireframes to photographic renderings of three stars, three planets, and the evil planet in the foreground, in blue.
The main challenge with the system is what the heck is being visualized? Professor Pacoli says in the beginning of the film that, “When the three planets are in eclipse, the black hole, like a door, is open.” This must refer to an unusual, trinary star system. But if that’s the case, the perspective is all wrong on screen.
Plus, the main sphere in the foreground is the evil planet, but it is resolved to a blue-tinted circle before the evil planet actually appears. So is it a measure of gravity and event horizons of the “black hole?” Then why are the others photo-real?
Where is the big red gas giant planet that the ship is currently orbiting? And where is the ship? As we know from racing game interfaces and first-person shooters, having an avatar representation of yourself is useful for orientation, and that’s missing.
And finally, why does the operator need to memorize what “Code 487” is? That places a burden on his memory that would be better used for other, more human-value things. This is something of a throw-away interface, meant only to show the high-tech nature of the Federated Territories and for an alternate view for the movie’s editor to show, but even still it presents a lot of problems.
Shaw sits at this device speaking instructions aloud as she peers through a microscope. We do not see if the instructions are being manually handled by Ford, or whether the system is responding to her voice input. Ford issues the command “compare it to the gene sample,” the nearby screen displays DNA gel electrophoresis results for the exploding alien sample and a human sample. When Ford says, “overlay,” the results slide on top of each other. A few beats after some screen text and a computerized voice informs them that the system is PROCESSING, (repeated twice) it confirms a DNA MATCH with other screen text read by the same computerized voice.
When Halloway visits Shaw in her quarters, she uses a small, translucent glass cuboid to show him the comparison. To activate it, she drags a finger quickly across the long, wide surface. That surface illuminates with the data from the genetic tester, including the animation. The emerald green colors of the original have been replaced by cyan, the red has been replaced by magenta, and some of the contextualizing GUI has been omitted, but it is otherwise the same graphic. Other than this activation gesture, no other interactivity is seen with this device.
There’s a bit of a mismatch between the gesture she uses for input and the output on the screen. She swipes, but the information fades up. It would be a tighter mapping for Shaw if a swipe on its surface resulted in the information’s sliding in at the same speed, or at least faded up as if she was operating a brightness control. If the fade up was the best transition narratively, another gesture such as a tap might be a better fit for input. Still, the iOS standard for unlocking is to swipe right, so this decision might have been made on the basis of the audience’s familiarity with that interaction.