on character
Can a computer express a convincing personality? How much information, what kind of complex code, what subtle graphics might be needed in order for a computer to have character, depth, reality?
When I originally presented Hypertext With Friends -- a discussion of hypertexts in which aspects of the hypertext become colleagues, part of the experience as well as part of the text -- several scientists in the audience worried that the computational burden of creating characters would prove intolerably great. An artist, however, can create character from a few grains of information: a shared glass of orange juice, the tension in an arm, a few words of dialog. The right details, set together, can reveal more than a thousand pictures.
People see personality everywhere. To learn how to use this human tendency to advantage, we turn not to artificial intelligence but to the writer's art.
Consider, for example, Shelley Jackson's hypertext, Patchwork Girl in which the monstrous protagonist,stitched together from scattered parts of departed women, tells of her love for her creator and her journey to the New World.
My right arm has two parts: the upper belonged to Tristessa, a woman known in the ship-yards for her deadly aim with a bottle--at stray dog or man, for she let fly at the one with as little or as much cause as she trounced the other. Indeed, she claimed to see no great distinction between the two: "You're neither of you safe to pet but in captivity, and there you're naught but a middling bed warmth, a gaping maw, and the bugs you bring home in your pelt." Followed a crash and the scatter of bright fragments. Yet she never lacked for company. It was rumored she had a dog of her own, but she kept it well hid, and as for the other, if you're asking, you'd do well to duck.
Shelley Jackson cannot tell us much about each woman who contributed a part to the Patchwork Girl's body -- this is, after all, the Patchwork Girl's tale, not an anthology of her components -- but each tiny sketch adds immeasurably to the depth of the whole. A few words, and our own minds, create more than 10MB of QuickTime animation could show.
Hypertext fiction seems to abound in memorable characters who inhabit the ragged edges of their worlds. Jude, in Stuart Moulthrop's Victory Garden, seems destined never to be seen alone and never to hold the spotlight -- in her most touching moment she dons a wig to become, for a moment, a ghostly Other Woman -- and yet her after-image becomes a redemptive aura over the novel's broad canvas. The inhabitants of Deena Larsen's Marble Springs are always glimpsed in passing, in motion through half-closed shutters or in transit through town.
He remembers her with longing. He remembers watching too many movies. He remembers her, nearly naked, arms, legs, entwined, and he, too, undressed, soft flesh.
He follows her home. He walks her to the door. Movie-style kiss. Orange juice. Naked thighs.
They do not speak to each other.
Morning.
Mary-kim Arnold's miniature "Lust" achieves the same sort of shorthand in a different way. The characters are nameless. In each writing space, they are simply the man, the woman, the child, and the knife. The knife receives as much attention as any of the people. The writing spaces themselves are tiny, and the entire work is encompassed in 37 intricately woven (and recurrent) spaces. Yet this compression is amazingly rich in effect; we know almost nothing about the man in this quotation, and yet it seems we know exactly what we need.
A common mistake, in multimedia and on the Web, is to rely on archetype and stereotype, to reach for pre-drawn characters tied into popular TV shows or cloned from popular novels, hoping to build upon the accumulated information the reader already possesses. This may be exactly wrong; the hardest characters to draw may well be the characters we know too well. The discordant, revealing detail (that orange juice!) is far easier for a writer to discover (and for the reader to believe) when describing a character we know, someone we feel strongly about. Readers are more receptive to the visualize new, unfamiliar characters; catching their imagination by reaching one more time for the Brave Starship Captain or the Hard-Boiled Detective can inspire boredom, not recognition.
People are eager to perceive personality wherever a hint of personality might exist. Two psychologists, Byron Reeves and Clifford Nass, summarize a series of fascinating experiments on the way people treat computers like real people in a a new book, The Media Equation (Cambridge University Press, 1996). Throughout history, they note, anything that spoke or acted intentionally was almost surely a person; even sophisticated users and computer programmers treat computers as if they were people in dozens of subtle ways. For example, Reeves and Nass have shown that users are reluctant to criticize a computer to its face: if a computer asks, "How am I doing?" or "Do you like me?", people will respond more positively than if a different computer (or a person) asks, "How well did that computer do? Did you like it?" We know the computer has no feelings to hurt, but we tend to be polite anyway.
Programs like Microsoft BOB, which try to apply simple, cute personalities to the computer, may fail to hold our attention because those simple personalities are difficult to sketch. On the other hand, Virgin Interactive's Catz and Dogz, software toys that place pets on the desktop of your computer, are most credible when the character's behavior is quirky, unexpected, but closely observed. That a newly-installed "kitten" will chase the mouse is clever, but that it sometimes stumbles over obstacles, or simply decides to stop and wash its ears, reminds us why real pets are more interesting than automata.
The ability to suggest much by showing little has been part of many realms of art for thousands of years: poetry, drawing, dance, fiction, and now hypertext. Learning these techniques and using them to extend hypertext, or to create ways of interacting with a computer that more usefully and deeply engage our passions, are studies that will take us many years more.
