AskTog: Magic and Software Design

[Tognazzini, Bruce "Tog." (1993), "Principles, Techniques, and Ethics of Stage Magic and Their Application to Human Interface Design," Proceedings of INTERCHI, 1993 (Amsterdam, The Netherlands, April 24-29, 1993). ACM, New York, pp 355-362.]

Software designers are struggling to generate more effective illusions for purposes of communicating to their users the design model of their applications (Heckel, Laurel, Norman, 1983, 1986, 1988, Rubinstein & Hersh, Shneiderman, 1992). At the same time, they are confronting serious issues of ethics: when does an attempt to create an empowering illusion become trickery, when does an attempt at anthropomorphism become cheap fraud? (Laurel, Shneiderman, 1988, Tognazzini, 1992.)

Our profession has primarily drawn its lessons from psychology, computer science, and graphic design. While these have supplied much valuable material, we are still busily re-creating organized knowledge in our field that Laurel points out has been well understood for thousands of years.

Perhaps no field other than magic is tied so closely to the field of graphical interface design: The people working at Xerox PARC in the 1960’s and early 1970’s were aware of the principles of theatrical magic when creating the first graphical interfaces, to the extent that David Smith named the interface itself the "user illusion" (Kay). We are designing interfaces for an interface system based on magic, yet there is almost nothing written about it in our literature. (An exception is a single page by Heckel.) Magicians have been struggling with the principles, techniques, and ethics of illusion for at least 5000 years (Burger). There’s a lot we can learn from them.

This chapter is not exhaustive. On the contrary, it barely scratches the surface. My goals are to introduce other software professionals to the teachings of this parallel profession of magic and to excite interest in what I believe to be a powerful set of tools for software designers. I have been an amateur magician for as long as I’ve been a professional software designer and have applied the principles, techniques and ethics of magic routinely in my design work. I have found them to be invaluable.

It’s hard to read through a book on the principles of magic without glancing at the cover periodically to make sure it isn’t a book on software design. These books clearly delineate the basic principles and techniques that support graphical user interfaces:

Both software designers and magicians create virtual realities. We bring ours alive on computer displays; magicians bring theirs alive on the stage. We capture our "performances" in code so they can continue to occur long after their writing; magicians traditionally appear live. We depend on our knowledge of the "mechanics" of computer technology, the aesthetics of graphic design, and the science of psychology. Magicians depend on their knowledge of the "mechanics" of their tricks, the aesthetics of showmanship, and the science of psychology.

Amateur magicians start out struggling with the mechanics of magic, and most never get beyond working with rigged gimmicks, like marked cards and "magic" milk pitchers. Competent magicians can work fluidly with apparatus, so that the spectators are unaware that the card seemingly dealt from the top of the deck came from the bottom, or that the rabbit that seemingly arrived from out of nowhere had been residing in the magician’s pocket, munching peacefully on carrots. Magicians must be competent at the mechanics of their craft, yet such competency does not make someone a magician, any more than knowledge of a rapid-prototyping system makes someone a software designer. It sets the stage. It makes everything else possible. (Of course, we software designers have a distinct advantage over our magician friends: we can depend on programmer colleagues–confederates–to do our technical trickery for us.)

One interesting difference exists between the mechanics of magic and of software design: while we are struggling to hide from our users the vast complexity of our electronic "reality," magicians are attempting to hide how trivially simple their tricks are: we’ve all been told that magicians never reveal their secrets because it is "a magician’s sacred obligation...to keep the secrets of his brotherhood," (Jillette & Teller) but there is another, more practical, reason: people get very angry when they discover they were so completely taken in by such simple ruses, and they invariably turn on the magician.

Showmanship seems like an unimportant aspect of software design, but an exploration of how it applies to magic reveals some unexpected results.

What is this thing we call a human interface? Lawyers will tell you it’s the "look and feel" of the software, but when you press them they will be at a loss to explain what that means. (They will also bill you $250 dollars for their failure.) Don Norman calls it the "System Image," the physical embodiment of the designer’s "Design Model" (Norman, 1983, 1986, 1988). Rubinstein and Hersh call it an "external myth." Ted Nelson calls it "virtuality." As mentioned earlier, Alan Kay and his cohorts at Xerox PARC, in deliberate reference to magic, called it the "user-illusion" (Kay). All of these words are descriptive, yet all are abstract and, therefore, somewhat elusive.

Magicians work to produce illusions, but they don’t call their stage presentation an illusion, they call it an act. That’s a good, down-to-earth term you can get your hands around. It comes equipped with expectations: we know that an act should inform, excite, and entertain us. If it doesn’t, we know what to do about it, which might or might not involve computing trajectories for rotten fruit, depending on the poverty of the performance.

Look at the most famous contemporary magicians–Doug Henning, Paul Daniels, Penn & Teller, Siegfried & Roy, David Copperfield–all are consummate showmen. Indeed, an examination of David Copperfield’s tricks shows them all to be rather old and prosaic, but he performs them in such a theatrical style and on such a grand scale, that we are enthralled. Most magicians, with a suitable trap door, can perform Servais Le Roy’s turn-of-the-century "Asrah" illusion, making a comely young person disappear in thin air, but David Copperfield has made a 100 ton steam locomotive disappear into thin air. Most impressive.

Many principles of magic showmanship are directly applicable to software design. Here are just a few from Fitzkee (1943):

I submit that showmanship has an important place in computer software. It certainly was a conscious component of the Lisa and the Macintosh. As one example, we didn’t adopt the trash can for file elimination just because it fulfilled the requirement of a "real-world" metaphor. We chose it because it seemed "neat," and we kept it because, when other people saw it, they loved it.

Showmanship does not mean, to use Ted Nelson’s term, "adding ketchup" (Nelson, 1991). It implies the application of a deep understanding of human nature to the task of making software seem vital, involving, and fun. Negroponte (1991): "I remember scholarly papers (circa 1972) that argued against color in the displays. They were filled with tables, control groups, and the like, proving that color could be confusing. Today, I hear of similar studies that speech and natural language are not appropriate channels of communication between people and computers....

"I certainly understand that I don’t want the pilot of a 747 to land an airplane by singing songs or a doctor to perform open heart surgery by humming into his apparati. But I can’t fathom any cause to avoid richness in communication, especially as computers become ubiquitous and are part of the quality of our daily lives."

Magicians learn showmanship the hard way, by standing on a stage, receiving instant and often painful feedback from a live and lively audience. One way people can learn how to inject showmanship into software designs is through a similar, if less painful mechanism: by dragging prototypes to customer sites, computer stores, club meetings–anywhere one can find an audience.

When setting out to design Apple Presents... Apple, the first in-box microcomputer tutorial back in 1979, Dave Eisenberg and I forwent the goal of teaching everything about the computer in favor of the more attainable goal of teaching a few fundamentals in such a way that new users would become interested and confident enough to want to learn more on their own.

The application started with an "attract mode" that gently cajoled the reluctant user to "just press the Return key." It continued with a building set of "success experiences," until after approximately five minutes, even the most fearful users were usually zipping away inside the application, feeling fully in command. During its design, we tried the application on more than 300 people, making it more lively and interactive with every pass. Within six months of its release, the Apple independent dealers identified it as their most valuable sales tool: they reported it made prospective owners love the computer and that the sale was then easy.

Not everyone will immediately agree that showmanship in spreadsheets is as important as showmanship on a Los Vegas stage, but consider this reality: software must be bought before it can be used (except in India, China, and Italy). Showmanship does not imply the injection of irrelevant frills and fancies. Showmanship is the gentle seduction of the users, leading them to accept, believe in, and feel in control of the illusory world we have built for them.

"The art of illusion is at least 95% applied psychology.... When [modern conjurers] use more than one part of trickery to nine parts psychology, they cannot hope to create the maximum impression."– Nelms.

The act is the entity in magic. The mechanical devices and techniques are there solely to support the act, showmanship is there to enliven the act, but psychology makes the act "work."

Actually, there are two simultaneous acts performed in magic: the one the magician actually does–the magician’s reality–and the one the spectators perceive–the spectators’ reality: The magician’s reality consists of all the sleights of hand and manipulation of gimmicked devices that make up the prosaic reality of magic. The spectators’ reality, given a sufficiently competent magician, is entirely different: an alternate reality in which the normal laws of nature are repeatedly defied, a reality where the magician, as well as his or her tricks, appear supernatural.

On the surface, this defiance would seem in direct contradiction to software design, where we more often engage in making our "supernatural" machines appear natural. However, at a deeper level, both camps spend their time doing the same thing: making people believe one thing is going on when quite another is really taking place: the Macintosh has no actual, physical trash can, and no amount of rummaging around inside with a screwdriver and a pair of wire-cutters will reveal one.

This dichotomy, this concept of two simultaneous acts–the real and the perceived–is both key to magic and the most important link to our own work in design. In the course of working with hundreds of people assigned to design software during my career, most of them programmers with no formal knowledge in human-computer interaction, the most critical concept I have found missing has been the dichotomy of the programmer’s "reality" and the user’s "reality." (I have put "reality" in quotes in both places, because what the programmer sees as "real" in the computer is just the system image provided by the systems engineer or hardware engineer: our computers consist of realities within realities within realities.)

Magicians’ spectators, with the exception of young children, are adversaries, there for the specific purpose of finding the magician out. Magicians often reflect their adversarial relationship in their vocabulary: "The deception the magician seeks to accomplish is an attack upon the spectator’s mind. Specifically, it is an attack upon his understanding."–Fitzkee (1945).

Because of the natural suspicion of their spectators, magicians have had to develop the psychology of the illusion to a high level. If we apply their techniques with our users, who are not suspicious (unless we insist on "burning" them a few times), we surely will achieve a believable result. Let’s look at a few key techniques magicians use to generate their simultaneous, but alternate, reality:

Fitzkee identifies six techniques for causing misdirection: simulation, dissimulation, ruse, disguise, monotony, and maneuver. For purposes of this paper, let us examine just the first two:

Simulation and dissimulation are both important magic techniques: the magician who wishes to simulate a coin disappearing into thin air from his or her right hand had better be proficient enough at dissimulation that spectators don’t notice the coin actually sliding into a pocket from the magician’s left hand. However, dissimulation is there to cover-up, not to startle and amuse. For example, in the "Asrah" illusion, the assistant slides through a trap door, but appears to still be lying on the platform. Dissimulation. Dull. It is when we as spectators believe we see the assistant, shrouded in silk, slowly rising into the air with no visible means of support, following the beckoning of the magician’s wand, that we become interested. Then, when the magician suddenly whips the shroud away, to reveal nothing but empty space, we are left dumbfounded and amazed. Simulation.

Amateur designers are far more likely to cover up a reality than alter it, but so are professionals who are deprived of sufficient resources to "do the right thing." In the early days of Lisa and Macintosh, we at Apple were given all the resources we needed, and we used simulation widely and effectively. Now our entire industry seems bent on power at the expense of everything else, and programs and interface systems are showing the symptoms of dissimulation. For example, Apple’s confusing System 7 feature, Publish and Subscribe, is essentially raw technology, with just the lightest sugar coating (dissimulation).

In contrast, human interface and graphic designers designed Apple’s QuickTime video animation tool from the ground up as a simulation responsive to the needs of its users. They repeatedly tried out the software on real users under real conditions, to see whether the design illusion was realistic, productive, and responsive. It ended up being all three.

Magicians continually stress that illusions don’t work without attention to detail. Fitzkee (1945):

Magicians talk about the "delicacy" of the illusion: a bit of light escaping from what is supposed to be a dark box or the errant corner of the assistant’s dress protruding from the trap door destroys forever all hopes of maintaining the illusion. Fitzkee (1945): "The performer should be particularly careful that his handling of all of his properties, in every respect, is in keeping with what they are purported to be, at all times. If they are handled as if they are what they seem to be, this contributes to convincingness and conviction.... Naturalness is the most powerful weapon at the disposal of the magician when he seeks to deceive."

The spectator doesn’t have to know details of the deception to know deception occurred, thereby destroying the illusion. In fact, the spectator can be dead wrong in his "explanation," and still the illusion evaporates: Fitzkee tells the story of a time when he was filming Howard Thurston performing "The Levitation of the Princess Karnac," in which the woman playing the Princess is actually levitated, without benefit of shroud. Fitzkee, meanwhile, was up in the balcony of the auditorium, filming the procedure, when, "several spectators heard the camera operating. They immediately connected the camera with the levitation. I could see them nudge their companions, call their attention to the camera, then point to the activity on the stage. From the way they relaxed and settled back into their seats, I am positive they felt they had solved the mystery" (Fitzkee, 1945).

America’s Disneyland and Disney World are studies in attention-to-detail. They are masterful illusions that make people believe in the unbelievable. The illusion does not rest on fooling a single sense. Sight, sound, smell, taste, and touch all receive a flood of concordant information.

Many other amusement parks feature theme rides and costumed employees, but, as you walk around, you are confronted with the seams. In the land of make-believe, you see a Chevrolet pick-up truck not quite hidden behind the magic carrousel. While dashing through the Old West, you can’t avoid overhearing your stage coach driver talking loudly with the man riding shotgun about his date with Allison at the disco last night.

The better the illusion, the more shocking is any flaw. At EuroDisney, as I boarded the train, I saw a young male member of the "cast" (as Disney employees are called) riding the rails during his break. The fact that he and every stationmaster at each stop saw fit to chit-chat about their personal lives was surprising enough. What was truly amazing was that he had unbuttoned the top of his eighteenth century French costume to reveal a most 20th century American undershirt beneath. The entire illusion that so many people had invested so much time and energy building up was, at least for the moment, destroyed.

No cast members in the United States would dare walk around with their costumes unbuttoned, and, even if they did, no other cast members would be foolish enough to engage them in idle chit-chat in front of the Guests (Koenig). It would be unthinkable. This lack of attention to detail may at least help to explain why EuroDisney is having such a struggle to survive.

Maintaining an illusion on a computer requires the same level of commitment and fervor historically displayed by the Disney company. And it can be as easily damaged. The Star, Lisa, and early Macintosh displayed finely crafted illusions. Today, our system illusions are rent by inconsistencies, program crashes, and unfathomable conceptual models.

Most of us have seen–or even been involved in–software projects where attention to detail was slight or nonexistent, resulting in software with unclear System Images, flashing redraws, unresponsive feedback, or, even worse, a dangerous lack of forgiveness. Such breeches result in software that is confusing, awkward to use, and even frightening–not exactly our normal design goals.

Once an objection, or even a suspicion, arises in the mind of the spectator, the trick is finished. Magicians put great effort into making sure their illusions develop in such a way that the spectator is not even given to questioning why a certain action is taking place. They do this by providing a motive within the spectators’ reality for every action taken within the magician’s reality. For example, in the "Asrah Levitation," the assistant will not actually be present during much of the performance, having slipped through the trap door. The device that takes the place of the assistant, the device that floats up into the air, is made of metal, rather than flesh and blood. It is therefore given to a certain unnatural stillness. If the spectators slowly became aware of that stillness, they might become suspicious, so to avoid their potential objection, the assistant is first seen to be hypnotized or forced to drink some poisonous elixir, thus seemingly rendering her as still as the apparatus that will replace her. The magician has overcome the objection before it arises.

Hormuz was an early program written in the PILOT language. It featured a character that would "speak" with children through the screen and keyboard. The character introduced himself as an ancient Arabian, then urged the child to "come close to the fire, for the light is weak." He then went on, after asking for the child’s name, to request the child’s gender: "My eyes grow dim with age: are you a little boy or a little girl" (words approximate). Having established the character as old, having established that it was dark out, the designer is then able to ask what would otherwise be an insulting question of a child. The designer has overcome the objection before it arises.

Magicians live in both the world of their mechanical tricks and the illusory world they are creating for their spectators, but they "believe" in the spectator’s world: "All of the most successful showman-conjurers agree that you must believe in your own magic; you cannot hope to convince an audience unless you first convince yourself."–Nelms.

In my experience, programmers face an ongoing struggle to believe in their own illusions. They continually want to slip back into the comfort of their mechanical world down below. In the early days of software, the System Image directly reflected every convolution and limitation of the structure of the program, the programming language, and the operating system. Today’s best visual interface systems are designed before the underlying systems are built, so that, while we continue to design for the expected capabilities of our systems, we need no longer design to a pre-determined structure. Instead, we design our System Images to reflect the structure of the Design Model (Norman, 1986, 1988, Tognazzini, 1992). Later, when the programmers lay out the software, they tend toward a structure in general conformance with that of our System Image, neatly reversing the historical sequence.

Programmers who have not made the transition to the design model’s illusion are easy to spot: they meet any attempt on the designer’s part to create a new and interesting design model with, "Yes, but that’s not the way it really works." The last thing a magician wants is for his spectator’s model of the act to bear any relationship to "the way it really works."

While we don’t share the magician’s pressing need to hide this "reality," we often gain advantage by offering the user a well-constructed illusion. For example, ISDN, the new all-digital telephone system, can complete a call connection in a few milliseconds, instead of today’s 5 to 10 seconds, exchange data in brief, high-speed bursts, then log-off, terminating billing charges, after an additional few milliseconds. By being faithful to the realities of ISDN, we could offer our users a much faster log-on procedure, but consider how much more we could accomplish by separating the illusion of the interface from the realities of the hardware: With "instant" log-on, high-speed communication, and short-duration transmission, we could create the illusion that the user is always connected. Any time the system saw the user perform a task that requires transmission, such as dropping an addressed document in an out-basket, the system could sign on, send the document, and sign off, all without the user’s conscious awareness. As far as the system is concerned, it is saving every precious penny of the user’s money. As far as the user is concerned, he or she is "connected" 24 hours a day, seven days a week.

The business of constructing a new metaphor, a new conceptual model is one of spinning a tightly-woven illusion. For example, Starfire proposes that there be only one primary holder of information, the container. The instances of this container would range from the folders of today to the document of tomorrow. Any container can change its form to become any other container. So a document can become a folder and a folder can become a document.

If this interface illusion is taken literally, it requires the software engineer to create a single object so malleable that it can be molded like silly putty into a dozen different forms. Given the powers of tomorrow’s object environments, that might prove to be a possible solution to projecting the illusion, but suppose it isn’t. Does that mean that the illusion must be changed to reflect the "realities" of the system? No.

Consider that, in fact, we end up with exactly a dozen different hard-wired containers. Not one of these containers is capable of being anything but what it is–a folder, a scrapbook, a notebook, an active paper document. We then build utilities that can do conversions for us: feed the utility with the name of an active paper document and it will produce a notebook from it.

We could, within the illusion of the interface, require the user to open the utility, enter the name of the source container, enter the name of the destination container, and press the "process now" button. But we can just as easily have the user select the source document, command, "change to notebook," then carry out the conversion, and replace the old object with the new, seamlessly and invisibly.

From the programmer’s point of view, what has just happened is that the data from one container has been transferred to another, and the old container has been deleted. From the user’s point of view, however, there is only a single object, and that object has just changed form.

Magicians use two techniques to offset the actual time a trick (the essential working of the apparatus) takes place from the time of the spectators think it takes place: Anticipation, where the magician does the trick early, before spectators begin looking for it, and Premature Consumption, where the magician does the trick late, after spectators assume it has already occurred.

I once saw a magician on television, seated before a small, 1 inch thick table, performing one of the oldest magic tricks known: Cups and Balls, described by the Roman philosopher, Seneca, in the first century AD. The spectators attempt to guess under which of three cups (or, in this magician’s case, stainless steel pans) the ball lies, while the magician makes sure they cannot. The magician began with the usual sponge ball, about the size of a golf ball, which, when squeezed into the palm, becomes the size of a pea and is rather easily inserted beneath any pan desired. After the people around the table had, predictably, failed to identify the correct pan on a couple of occasions, the ball began to grow. And grow. And grow. By the end of the trick, the man tilted the pan forward, then lifted it away to reveal a bowling ball beneath. (The bowling ball, while no wider than the pan, was a great deal taller than the pan was deep.)

The table was not gimmicked in any way. The man had, from all appearances, not lifted any heavy objects. Rather, he had been very busy talking with his spectators when the bowling ball suddenly appeared.

By chance, I had videotaped the performance and was able to study it in detail. The magician had used a technique I call counterpoint: for the entire course of the act, his trickery had preceded its revelation by around 5 seconds: as he would engage in a vociferous explanation of why the spectators had failed on the last round, gesticulating with his hands for emphasis, he was actually placing the load for the next round. The choreography was superlative: like the rich counterpoint of Bach or Händel, the notes of his own reality blended perfectly with the reality he was presenting his spectators. All of that man existed in the reality of his spectators–his intellect, his emotion, his gestures, his words, his apparent actions–except some tiny part of his mind, like some little demon, which went quietly about the task of preparing the next trick. It was a beauty to behold, and, try as I might, even examining the tape frame-by-frame, I could not catch him loading that bowling ball behind the pan.

"Pipelining"–drawing screens, gathering data before needed–is a form of anticipation. Printer buffering is a form of premature consumption. These techniques illustrate that the timing of the user’s illusion need not track the reality of the operating system or hardware.

One caveat: illusion is sometimes shattered on our computers when something goes wrong: telling the user that "the document has been successfully sent to the printer" when the document has in fact only been spooled to the computer’s internal print buffer would seem like a good idea, but not when a difficulty arises with the print buffer software and the user ends up dragging a properly-functioning 100 pound laser printer into the shop for repair. We need to consider the entirety of the user’s reality, and that consists of both the expected and unexpected.

Houdini, the great escape artist, was famous for his Milk-Can Escape in which he was squeezed, half-naked and hand-cuffed, into a three-foot tall vessel brimming with water. The top of the can was securely fastened from the outside with six padlocks, and then the apparatus was curtained off from the spectators. They grew increasingly nervous as they watched the minutes tick by as he attempted to make his escape. By the time some ten minutes had passed, the tension would build to an explosive level, not at all relieved by the mounting panic seen in the faces of Houdini's helpers. Finally, the helpers would be able to stand it no longer and would tear aside the curtain to smash open the milk can. But there beside the can would sit Houdini, nattily dressed in a suit and tie, calmly reading a newspaper (Henning).

The magician Blackstone told the story of one occasion when Houdini's manager could hear the rustle of Houdini's paper within a couple minute's of the curtain's closing. The manager whispered to Houdini to keep the noise down, to which Houdini replied, "tell the band to play louder." (Henning, as told to him by Blackstone, Jr.)

Houdini’s fondness for reading extended to magazines, too: after his famous escape from inside a locked safe behind a screen, which took mere seconds, he spent the next fifteen minutes back stage, idly reading a magazine, after which he dotted himself with water and burst forth, looking properly sweaty and exhausted (Daws & Setterington).

Houdini understood the importance of not making a task look too easy; the designers at Fairchild did not: Fairchild produced one of the first home video game machines called Channel F, back in the mid 1970’s. It featured a first-rate tic-tac-toe game with a tragic flaw: regardless of how long the player took to plot his or her next move, the computer would respond within one-half second with its next move. Combined with the machine's skill at choosing the best possible move, this fast reaction left the user feeling puny and inadequate.

Researchers from Robert B. Miller on have been studying the psychological effects of response time on the user (Bergman, et. al., Grossberg, et. al., Henning, Miller, Shneiderman, 1992). What the study of magic offers us is a different perspective on the subject: we are looking beyond efficiency and accuracy; we are looking at the effectiveness of the "act" and its "big picture" impact on the user.

The throwing of the I Ching is an ancient ritual involving the repeated casting of yarrow-stalks in a meditative atmosphere. The process of a single prediction can easily be stretched to the better part of an hour, and no prediction is to be repeated on the same day. I will offer no speculation on the accuracy of the prediction (although the system was designed with enough ambiguity that the diviner can "tailor" the results to the question and circumstances at hand). What I will call attention to is the immediate beneficial effect on the subject, who is being honored with a great deal of the diviner’s time and attention in a warm, spiritually-comforting atmosphere. It is the epitome of "quality time."

One of the earliest programs on the Apple II was an automated I Ching caster. It could electronically "cast the stalks" in less than 1 second. You could ask one question and get 60 completely different answers in less than 1 minute! It may be safe to say the designer was unstudied in the ways of the Tao.

By applying the magic technique of time-stretching, along with a healthy dash of showmanship, we could write a quite different I Ching program. It would ask the user many pertinent and perhaps not-so pertinent questions, "cast the stalks" in a way that was both visually interesting and very time-consuming, deliver results in a quiet, dignified, poetic way, and refuse to answer the same question again on the same day. It would likely be no more accurate in its predictions than its faster cousin; it would surely be less productive, taking hundreds of times longer to use; but it would be far more responsive to the real (if imagined) needs of the person using it.

The designers of Zen gardens stretch time to generate an illusion of space. Those paths of rough-shaped stepping stones that require such balance and concentration are there to slow you down, to create the illusion that you are covering vast distance when, in fact, you are traversing but a few feet. This is a valued illusion in a country where land is scarce and gardens, tiny.

Magicians purposefully disconnect the illusion of speed from its reality to aid their tricks. We often do it by accident, and to our detriment. The real speed of a computer and the user’s perception of that speed often end up quite different. For example, workstations are typically far faster than the Macintosh, but the experience of using the Macintosh would often leave one with the impression that the Macintosh is faster. Why? Because the Macintosh designers made sure that the most important object in the user’s space, the mouse pointer, would have zero latency, would never lag behind the user’s true position. They did this through a combination of very fast software and frequent position updating, so none but the most rapid mouse movements result in a series of separated pointer images. (Most Macintosh users will claim a break-up never occurs, having never noticed it.)

Latency becomes far more troublesome as perceptual bandwidth increases. Virtual reality systems today suffer from significant latency problems, giving them a "mushy" feel and leaving many users disoriented, dizzy, and confused. VR latency does not arise from a dearth of design; the hardware just isn’t yet up to the task. During this decade, latency will begin to disappear, and public acceptance of VR systems will be on the rise.

Responses that can be made instantly should be made instantly. When a response will take some time, the user must receive instantaneous intermediary feedback (<.5 seconds), then be kept interested and entertained to reduce their subjective perception of how long they’ve been waiting.

In the 1930s, a major high-rise office building was opened with what soon proved to be too few elevators. After a series of engineering firms were brought in, each confirming that there was no way to either speed up the existing elevators or add additional ones, the building owners, in despair, brought in an interior designer. The designer recommended that huge floor-to-ceiling mirrors be installed between each pair of elevators, and the complaints disappeared. People now had something to do while waiting: either gazing at their own magnificent image or peering secretively at others with little fear of being caught.

The engineers had concentrated on reducing objective time; the designer concentrated on reducing subjective time. Reducing subjective time works.

I propose that there is a "threshold of believability," a point at which careful design and meticulous attention to detail have been sufficient to arouse in the spectator or user a belief that the illusion is real. The exact point will vary by person and even by mood, so we must exceed it sufficiently to ensure believability. Disneyland and Disney World are above the threshold of believability; county fairs are not. Lucas’s Star Wars was above the threshold; Attack of the Killer Tomatoes was not. Penn and Teller cutting a live snake in two, then restoring it, on Saturday Night Live was above the threshold; Uncle Charlie’s tired, old card tricks at Christmas were not.

The original Star, Lisa, and the early Macintosh all exceeded the threshold. Some of the graphical user interfaces appearing now, with their underlying dependency on dissimulation and their lack of consistency, are falling short. If users cannot trust the system, if they are occasionally but violently thrust into the programmer’s reality, they can not, will not, and should not believe in the world we are making for them.

One need sacrifice no power in building a believable illusion. For years, Macintosh programmers were able to recover from a crashed application by typing arcane incantations in their debugger, while regular users were left with no method of recovery at all. With the advent of System 7, anyone could press Command-Option-Escape to achieve the same result. Recovery is now not only achievable by every end-user, but is easier for the programmers.

Illusions, to be believed, must not only look, but act "real." For example, consider today’s data bases that have the visual appearance of a book (a popular metaphor in HyperCard and its offspring). Most of these show the same number of pages in the illusion whether there is one page of information or a thousand. The illusion is not real. It may be a lot more work to program for conditional graphics, but the thickness of a book communicates important information. Denying users this natural way of gauging volume, we then end up supplying the same information in yet another object or dialog the user must battle to do their work.

Stage magicians have been impersonating "real" magicians (such as Merlin) for a long, long time. They’ve had plenty of time to experiment with the ethics of stage magic and have come up with workable solutions. I present these solutions because I find them applicable to the ongoing discussion of impersonation of people by computers (anthropomorphism).

Shneiderman (1988) argues strongly against having the computer personify a human, although he does suggest that young children might be exposed to a cartoon character for the sake of visual appeal. Laurel agrees in part, but has expanded his horizon: "...I have argued, not for the personification of the computer, but for its invisibility.... The representation of agents or characters is a different idea altogether than the notion of ‘personified’ computers." She goes on to argue the benefits of visible characters in the interface.

The mainstream of magic is in essential agreement with Brenda Laurel’s view: "A magician is an actor playing the part of a magician."–Fitzkee (1945). Mainstream magicians find it important that spectators do not leave the theater under the impression that the magic performed or that the magician who performed it is supernatural. This is such an important issue to them that when people claim supernatural powers, magicians flock to expose them: Houdini spent much of the latter part of his life both seeking out a true spiritualist and utterly destroying all the fakes that lay in his path (Gibson & Young). Uri Geller, who claims to bend spoons and start broken watches through spiritual intervention, has had his own personal exposer, in the person of James (The Amazing) Randi, who has taken great pains over the past 15 or 20 years, to duplicate any "spiritual" trick Geller attempts with plain old-fashioned magic techniques, much to Geller’s discomfort.

A notable exceptions to the mainstream view of magician-characterization lies within the offshoot of mainstream magic called, "Magick." Harkening back to the dark days of ancient sorcery, these conjurers carry out acts of bizarre ritual, like decapitation and disembowelment, with accompanying blood and entrails, after which they reassemble their victims and restore them to life. Cameron: "The average magician has long since given up dread–what he requires is a sound Gothic revival!... The silks-from-a-box man rants and raves about the ethics of claiming supernatural powers and insists that the public should be protected from these charlatans and psychic impostors. Rubbish and triple balderdash! Either he is working MAGIC or he is presenting mechanical and manipulative mumbo jumbo–he cannot have it both ways! The usual argument is that the spectator should believe in the magician’s powers during his act... but regain reality at the conclusion of the performance. Stuff and fadding nonsense! Either you are a magician with magical powers or you are not. It is as basic as that!

My observations of the purveyors of "magick" are, that for all their bluster, they are as committed to not leaving their spectators deceived as mainstream magicians. Perhaps the difference between these conjurers and outright frauds is exposed in this quote from Matthew Field, from the forward of Burger’s book on magick: "Magicians get angry at [Uri] Geller because he is, we believe, a magician who does not admit to being one. That anger might be better directed at the fact that Geller uses his magic to promote only himself, leaving those who believe in him, and our world, no better off (and perhaps a bit worse than that)."

Like mainstream magicians, magick conjurers imbue their characters with supernatural power, not themselves. Whether they make an announcement at the end of the show to the effect that, "no, we were just kidding," does not seem important to me. (It does to Laurel, though, who suggests we should make such an announcement before our software performances) What both factions have in common is that there is a character involved, and the audience understands there is a character involved.

Applying the "super-anthropomorphism" ethics of magic to software, I see it calling for Laurel’s split between the invisible computer, on one hand, and the robust, visible character on the other. This fulfills the same requirement of honesty: the magician is not supernatural; the character he plays is. The computer is not capable of human intelligence and warmth; the character we create is. People will not end up feeling deceived and used when they discover, as they must ultimately, that the computer is nothing but a very fast idiot.

Building demos for unreleased products can be ethical or unethical, depending on how the demo is presented. A magician who forces a seven of diamonds within the confines of a magic act is acting ethically: He and his audience have a tacit understanding that he will be carrying out sleight-of-hand. The same magician who later that night forces the same card while playing a game of poker is a cheat.

One early computer manufacturer kept showing a demo for a word processor for three years. Not only did the word processor never actually make it to market, but it demonstrated features, such as smooth scrolling, that, under normal (non-demo) circumstances, were not achievable on the target hardware. The result was that a lot of people held off buying for a long time, to the detriment of their customers and competitors.

So-called docudramas on TV strive to make people believe they have witnessed an event, when what they have actually witnessed is a carefully-crafted illusion.

In the case of video prototypes, we must assure spectators, within the context of the prototype, that the system being shown is not real. Starfire not only does not exist, but much of it couldn't exist for a good ten years, so we wrapped the demo of our proposed technology around an obviously fictional story, a story firmly set a decade into the future. We worked hard to make the technology seem so real that people would feel they had touched it themselves, but we also took pains not to pass it off as anything but a vision.

Computer simulations can take on such a sense of reality that spectators are unable to convince themselves they aren’t real. In San Francisco recently, prosecutors replayed their version of the events surrounding a fatal shooting by means of computer simulation–a rigged demo of the first order. They claimed they chose the medium strictly to communicate their version of the events more clearly to the jury, but obviously their intent was to implant in the jury's minds the illusion that the jury members had personally eye-witnessed the crime.

The technique is designed to attack the human subconscious directly, bypassing all the information filters with which evolution has provided us. Orson Wells and the Mercury Theater proved on Halloween, in 1939, when they presented a fictional "War of the Worlds" and terrified half the United States that there are no disclaimers big enough when the sense of reality and potential harm grow large. "War of the Worlds" was peppered with disclaimers, but nobody heard them.

If, having finished reading this chapter, you feel strangely unsatisfied, I have accomplished my aim: 5000 years of magic cannot be compressed into a few pages. We have much to learn from the best of the master magicians, and several important resources for doing so. First, Fitzkee’s trilogy and Nelms single volume (all currently in print), when taken together, lay out the organized knowledge of illusion-making clearly and precisely. Reading just these four books will give you a firm foundation. Second, while perhaps you were not left so strangely unsatisfied by the paper that you are ready to become a full-time apprentice to Penn & Teller, you may at least want to become a more active spectator by studying the performances of them and the other master magicians mentioned in this paper. Finally, there is no substitute for direct experience: learning to perform just one or two tricks well will instill many of the techniques and principles I have discussed at the level not achievable by either reading or watching.

I have found the experience of learning just a touch of magic to be most enlightening and useful. I urge you to consider it.

One closing word about magic stores: avoid the fancy one out at the mall, run by a dull young person who doesn’t know The French Drop from Kellar’s Blue Room Illusion. The store you want is old and dingy and has a huge plastic fly in the window. And don’t be afraid, as Penn Jillette says, "...the seedy side of magic stores is the only good part. I love the fact that the tricks themselves are a rip-off, a lie within a lie. My friend Elliot thinks that if he had been ripped off in a magic store as a child, he might not have been ripped off by getting his master’s degree. A sleazeball with sponge bunnies might have saved him a few years and several thousand dollars."