Massively Multiplayer Macbeth:
Interactive Narrative After Hamlet On The Holodeck
Kevin Chen
2011
In 2009, a decade after the publication of the last Sierra-era point-and-click adventure game [1], playwright Richard Lovejoy and off-Broadway theater company Sneaky Snake Productions produced a play called Adventure Quest. Inspired heavily by the works of Roberta Williams, Adventure Quest drags its feckless protagonist—a nameless hero who can only walk stage left, stage right, forward, and backward—kicking and screaming through a Campbellian hero’s journey. A mysterious disembodied voice, speaking mostly in simple subject-verb pairs, coerces him into running fetch errands for shopkeepers, using bizarre combinations of mundane objects to get himself out of danger, and navigating his way through labyrinths of death traps by trial and error. The hero unsuccessfully attempts to eat, sleep, and drink; falls in love with a nameless peasant girl who responds to everything he says with the sentence "You must help us, brave warrior!"; and fumbles awkwardly for the correct semantics to ask a cultist for directions to the castle. Trapped in the rigid universe of this point-and-click adventure world, the hero grows increasingly despondent and begins to question the importance of his quest (to rescue a princess he's never met from an evil sorceress), his motivation for saving the princess, and the boundaries of the world itself. He rebels in a number of ways, sitting motionless by the peasant girl’s side for many hours, feigning meaningful friendships with the echolalic townsfolk, attempting suicide, and finally murdering the rest of the cast in a fit of rage—only to see his actions undone by the falling curtain, followed by the ominous intonation “Would you like to restore your game?” Tormented by the absence of control over his own fate, and realizing he will cease to exist if his tale is finished, he goes to the sorceress's castle to ask the sorceress for counsel—and accidentally kills her. The curtain drops as the princess embraces the hero, who falls to the ground, screaming, as a tinny General MIDI ending fanfare plays in the distance.
Adventure Quest is striking because it is not merely a theatrical homage to a video game—it is a critique of video games’ capacity for storytelling. Described in summary, Adventure Quest's setting, the generic heroic-fantasy kingdom of Perilton, sounds lively and bustling—merchants hawking their wares, children playing marbles in the streets, coy prostitutes flirting with passersby. But the merchants only sell one item, the children are forever playing the same game of marbles, and the prostitutes merely repeat the same coquettish solicitation over and over. Even the carpenter, the closest thing the game-world of Adventure Quest has to an intelligent agent, who chops down wood and builds fences, endlessly chops down the same tree and rebuilds the same fence.
It is astonishing that though all these characters are played by human actors, the rigid, repetitive movements, one-dimensional motivations, limited capacity for dialogue, and general lack of self-awareness immediately identify these characters to audiences as artificial intelligences. Trapped among this country of cardboard-cutout automatons, the hero—who audiences quickly identify as a human player character, with his unique capacity for genuine reflection, remorse, joy, anger, sadness, epiphany, frustration, hope, despair, and regret—is beset by a terrible existential loneliness. It is grandly ironic that the limitations on the hero’s ability to interact with the world, which ultimately suffocate him, were produced by those games' very attempts to make their own worlds come alive.
Adventure Quest's Perilton is an interesting pastiche of late-'90s attempts to so closely follow the narrative framework in Janet Murray's groundbreaking 1997 book Hamlet on the Holodeck that they entirely miss the point. Murray identifies three unique elements of interactive storytelling: immersion, agency, and transformation. On paper, Adventure Quest is immersive, since the Hero can interact with objects in Perilton as if they were objects in the real world; it demonstrates agency, since nothing happens in the world without the Hero's actions; it is transformative, since the player assumes the role of a twenty-year-old Robin Hood figure.
But the game within the play actually fails all three measures: the world is about as immersive as an animatronic theme park ride, as characters do nothing except repeat the same actions mechanically; the Hero does not have true agency, since he is only permitted to perform actions relevant to the save-the-princess narrative the game designers intended (and not abandon his quest to woo the peasant girl like the hero desires); and the Hero still acts like what is implied to be his petulant, shiftless twelve-year-old player, who is given no incentive to act heroic.
In employing a number of Murray's strategies to achieve immersion, agency, and transformation--namely, populating virtual worlds with interactive objects and quirky AI characters, implementing a deterministic event-driven narrative model with multiple outcomes, and assigning roles to the player character--Adventure Quest's game-within-a-play feels close enough to narrative that the Hero is willing to treat it as one, yet frustrates him with a lack of meaningful creative control. These are engineering solutions to a creative problem, and they produce not a holodeck, in which the player's will is projected into a virtual space, but a prison--a virtual space that coercively constrains the player's will. A terrifying existential prison, no less, in which the uncanny valley effect of the game's many near-human AI characters makes them ominous rather than humanlike, and the game's sole human character--the Hero--is left to dwell on the uniqueness of his humanity in Shakespearean soliloquy. Adventure Quest is, perhaps ironically, the tale of Hamlet told on a holodeck.
Adventure Quest would not be nearly as funny, or as sad, if it did not exemplify exactly what the legacy of Hamlet of the Holodeck on commercial video game narrative has been. In the past decade, great strides have been made in the technology Murray proposes for the construction of a unique, compelling interactive storytelling medium. The past ten years have seen titles like Fallout 2, Metal Gear Solid, Grand Theft Auto 3, Half-Life, Dwarf Fortress, and Braid dramatically change the way stories are told through video games. [2] And yet, the vision of an actual holodeck-like experience--a virtual lucid dream, unique to each player, in which players live out deeply personal adventures in virtual space--has all but fallen by the wayside. This paper will describe those technological innovations and conclude with a proposed framework for using them to continue the pursuit of that vision.
Intelligent Agent-Driven Characters
Research on intelligent agents has continued apace since Hamlet on the Holodeck's publication, but while many of the innovations have been technically groundbreaking, most of them center around new uses for the technology rather than more humanlike behavior. Significantly, Colloquis's chatterbot SmarterChild learns new responses from previous conversations with interactors, maximizing the probability of serendipitously interesting responses. Also, in 2010, Knifeandfork Productions ran a heavily Murray-inspired digital performance piece called The Wrench in which Tino, an intelligent agent based on a character of the same name from Primo Levi's novel The Monkey's Wrench, would periodically over the course of two weeks have text message conversations with visitors to the Brick Theater in New York City about his personal experiences as a 1950s construction worker. No intelligent agent has yet passed the Turing Test.
Intelligent agent driven narrative has fallen out of favor in recent times. As the general public grows more familiar with the common eccentricities of AI behaviors, such as spontaneously changing the subject or repeating interactor input in the form of a question, it becomes increasingly difficult for an AI to pass itself off as human. Worse, commercial attempts to employ intelligent agents to perform useful activities has given them a reputation for being obtuse and tiresome. When Microsoft introduced Clippy the talking paperclip as an interactive agent-driven "desk assistant" in the Microsoft Office 97 suite of productivity software, users revolted. [3] The perpetually clueless Clippy emerged at every opportunity to guess what the user was doing and guide him or her through mundane activities such as printing or composing a letter, unhelpfully suggesting stock templates for letters that few people used or attempting to draw users through a wizard interface more cumbersome that simply performing the task manually. Consequently, contemporary audiences immediately recognize intelligent agent behavior, and find it difficult to suspend their disbelief.
Not only are intelligent agents not convincing as human characters, they are just close enough that humans find them unnerving due to the uncanny valley effect. Consider the ominousness of the townspeople in Adventure Quest (mentioned in the introduction to this paper). None of the activities--sweeping a dirty floor, hawking wares in the town square, watching a door at a castle--would be particularly unusual for human characters in their role. Their single-minded obsession with the behaviors they were programmed to do, however, makes them grotesque gargoyles of everyday tedium rather than relatable characters, driven by their defining roles rather than individual motivations. Madeleine d'Engle eloquently describes the terror of this sort of endlessly repeating, single-minded "zombie" behavior in her children's novel A Wrinkle In Time:
Below them the town was laid out in harsh angular patterns.... In front of all the houses children were playing. Some were skipping rope, some were bouncing balls. Meg felt vaguely that something was wrong with their play. It seemed exactly like children playing around any housing development at home, and yet there was something different about it. She looked at Calvin, and saw that he, too, was puzzled.
"Look!" Charles Wallace said suddenly. "They're skipping and bouncing in rhythm! Everyone's doing it at exactly the same moment!"
This was so. As the skipping rope hit the pavement, so did the ball. As the rope curved over the head of the jumping child, the child with the ball caught the ball. Down came the ropes. Down came the balls. Over and over again. Up. Down. All in rhythm. All identical. Like the houses. Like the paths. Like the flowers.[4]
There is another, more fundamental problem with intelligent agents as story characters, one even more difficult to overcome than their unconvincingness as human beings. Great classics of any literary medium, from Hamlet to Citizen Kane to Maus, are celebrated for their resonance with universal, common human experiences. In the late nineties, when Hamlet on the Holodeck was first published, there had been considerable recent research by Tobias and Campbell that suggested that such human experiences, evoked in all narrative, could be succinctly categorized and broken down into archetypes; Murray then extrapolated that these archetypes were now well enough understood that they could be strung together by a computer to generate random stories. As Murray herself notes [5], however, these stories are by definition constrained to the most literal interpretations of archetype, and are therefore quite formulaic and rarely compelling. More thorough research on archetype, Turing Test-passing AI, and sophisticated computer-guided storytelling, as well as the guidance of an artful human author, could have conceivably created a believable if predictable archetypal character intelligent agent in the ensuing years.
Such research did not occur.
It turns out that while Campbellian hero archetypes are universal, it is not the archetypes themselves that make stories resonant and compelling. What audiences hunger for is the interpretation of these archetypes, their emergence in surprising ways. This is particularly true in the irony-obsessed, postmodern contemporary era, in which entire television series (Community comes to mind) are built around subverting audience expectations of well-worn stock characters, Campbellian heroes, and plot tropes. Murray hoped that a skillful human author would be able to provide this magic touch by developing intelligent agent characters with particularly unique or interesting personalities, like psychologist Kenneth Colby's Parry or Neurotic Woman AI personas [6], but the fact remains that an intelligent agent with a limited set of responses and behaviors will, to a human accustomed to the unpredictability of other human personalities, are too predictable for humans to have any sort of comfortably meaningful interaction. The Hero's failure to form meaningful friendships with the townsfolk in Adventure Quest highlights this "character gap" perfectly--he spends a good two minutes asking a farmer about his job, his family, and his life goals, only to have the farmer respond by repeatedly giving him directions to the castle.
What the mid-'90s interest in believable AI characters overlooked in their quest to have games tell compelling stories was that games were already producing compelling stories--about games. It is telling that the most celebrated stories about gaming experiences from the '90s revolve not around stories told by AI characters, but stories told by players interacting with the characters in unusual ways. Usenet archives tell grand tales of thieves in Ultima Online luring pursuing victims into friendly towns, where bands of AI guardsmen would slaughter the bewildered pursuers, interpreting the pursuers' drawn weapons as intent to murder.[7] Public beta testing of Ragnarok Online saw mischievous players dropping hundreds of individual arrows in the town square of Prontera, the game's most populous city, crashing the game server as thousands of other players overloaded the system trying to pick them up. Boatmurdered, the tale of a collaborative Dwarf Fortress session menaced by excessively aggressive AI elephants [8], is perhaps better known than Dwarf Fortress itself. A shrub brush in the Star Wars Galaxies beta, accidentally placed ten feet above the ground by the game's random terrain generator, inspired an infamous ingame cult. All of these stories were denounced as being contrary to the spirit of the game by the games' creators, yet they turned out to be far more compelling than the predictable hero's journey and survival experiences the games were intended to offer.
Even without unintended AI behavior, gamers are known to produce astounding stories based on their experiences with games. The infamous article "Bow, Ni**er," highlighted by gaming journalist Kieron Gillen as an inspiration for the New Games Journalism movement [10], describes how a session of Jedi Knight II multiplayer deathmatch mode, typically a plotless, murderous free-for-all, became a dramatic contest between good and evil:
You see what this has become? It's not just a trivial game to be played in an idle moment, this is a genuine battle of good versus evil. It has nothing to do with Star Wars or Jedi Knights or any of the fluff that surrounds the game's mechanics. I played by the 'rules' and he didn't, that makes me the 'good' guy and him the 'baddie', but this is real, in the sense that there's no telling who's going to win out here. There's no script or plot to determine the eventual triumph of the good guy (that's me, five health), there's no 'natural order' of a fictional universe or any question of an apocryphal ultimate 'balance'. There's just me and him, light and dark, in a genuine contest between the two.
And there it is. I don't even know what it was. Some chance slash or poke in all of the rolling and jumping around and his lifeless avatar, with all his racist stabs and underhand duplicity, goes tumbling to the floor vanquished by the guy who even in the face of all of that, played by the ‘rules'. Only one health point remains but I win.
I'm a fucking hero. A real one. [9]
These stories are fascinating in that they do not follow the mold of conventional literary forms like the novel or the play. They more resemble the baseball story, a narrative heavy on plot and light on character, set within the game's ludic system in an unusual configuration, building credibility through a progression of increasingly implausible feats achieved by an ordinary yet heroic human protagonist, starting from the mundane and working up to the extraordinary, and ultimately celebrating the protagonist's exceptional skill, cleverness, or ingenuity. In this sense the video game story more closely resembles an epic poem than a play. Sherry Turkle notes in The Second Self that video game players treat games as a "perfect mirror," as an abstract reflection of their own personal conflicts [11]; in this light, such stories about video games can be understood as a contemporary iteration of some primal narrative impulse. Are we misguided in seeking a Hamlet on the holodeck when what may be emerging is a sonnet?
The telling of these stories are not always merely the effect of the natural tendency of human beings to craft narrative of their experiences. Some games actively encourage the crafting of gameplay narrative without producing it on their own. Forums for A Tale in the Desert, a massively multiplayer online game series notable for its absence of non-player characters and its emphasis on social objectives rather than exploration and combat, occasionally erupt into acrimonious land ownership disputes that, without any intentional roleplaying on the part of the participants, involve in-character personal attacks and offers of reparation in the form of flax fields and mining rights [12]. rec.games.roguelike.nethack, a Usenet forum devoted to the 1987 hack-and-slash dungeoneering adventure NetHack, has a long-running category called YASD ("Yet Another Stupid Death"), in which players relate, in slapstick style, the tragicomic rise and fall of yet another promising adventurer, with the absurd, improbable means of death (typically one of NetHack's many obscure instant death traps, as well as the capriciousness of the random number generator that determines each player's fortune) being the inevitable punchline. After twenty years, this format has yet to grow stale.
Given the astounding quantity and quality of narrative constructed around gaming experiences, it is somewhat questionable as to why so much attention in the field of interactive narrative is devoted to the cumbersome, difficult task of producing intelligent agents that do the same thing.
Multi-User Storytelling
Murray devotes an entire section of the ninth chapter of Hamlet on the Holodeck, "Digital TV and the Emerging Formats of Cyberdrama" (266-271), to multi-user storytelling. MUDs, or multi-user dungeons, were the pinnacle of interactive collaborative narrative at the time. As such, it is not surprising that her projections of the future of collaborative cyberdrama are an extrapolation of that medium. She predicts the emergence of a hypothetical science fiction television series, "Jerusalem Six" (a pastiche of Babylon 5), in which audiences can explore the setting of the show as a persistent virtual world in between episodes, holding conversations with minor characters, discussing the events of the show with other audience members, and filling in details of each episode left out by the limitations of a half-hour episodic format. Jerusalem 6 is, in essence, a television show and a MUD connected by a common universe.
Impressively, this vision actually came to fruition in ensuing years. The Matrix Online and Star Wars: Galaxies are notable but not alone in their attempts to derive a persistent virtual world from well-loved settings from television and film, giving audiences an opportunity to explore characters and subplots outside of the scope of the works from which they were derived. Perhaps closer to the spirit of Murray's vision, however, is the explosion of audience-produced derivative work that has since become a staple of Internet fan culture. Starting with the fanfiction boom on Usenet in the early 1990s, in which fans of shows like Star Trek: The Next Generation and The Simpsons wrote their own episodes, and dovetailing with the slew of comic books and novels licensed by Lucasfilm to explore the backstories of minor characters from Star Wars, fans have been creating an increasingly sophisticated, impressive body of literature, art, film, and interactive media based on television and film. One example particularly close to Murray's intent is World Resolution, a community on the collaborative blogging site LiveJournal in which participants collectively write an apocalyptic story set in the universe of Japanese studio CLAMP's animated series X/1999. The setting, characters, and plot are all original; the only thing it derives from X/1999 is the complex system of archetypal roles, inter-character relationships, and fated plot events that drives the original story. (For example, one character will be a savior figure called the Kamui, and the other will be his or her fated nemesis, the Gemini, and they must confront each other in order for the plot to move forward--but how and where and when the challenge will occur is anybody's guess. [13]) The result is a permanent single work of collaborative storytelling, more open than traditional tabletop or live-action roleplaying in that any reader can experience it through any character's eyes. It is a surprisingly funny, poignant, powerful, tightly interconnected story about fate and destiny, told from a rhizomatic multiplicity of perspectives (in the vein of philosopher Gilles Deleuze (cite)) united by a single overarching narrative, in which the main protagonist is whichever one a reader likes best, and author and reader and character are all one and the same.
More ambitiously, cartoonist Andrew Hussie's webcomic imprint MS Paint Adventures has been moving beyond the idea of reader as character and into the territory of reader as creator. MS Paint Adventures stories, though unmistakably webcomics, are told within the framing device of a Zork-like interactive fiction game.[14] Readers submit commands to Hussie through a suggestion box, and Hussie selects his favorites and use them to guide the action of subsequent panels. The result is an interactive fiction game in which the entire readership acts as one collective interactor who drives the events of the story, and Hussie acts as its interpreter, the possibilities of the narrative limited only by Hussie's imagination and artistic ability. In a typical MSPA story, neither Hussie nor the audience knows exactly how the story will turn out--they are dependent on one another to bring the story to completion.
In his book Twisty Little Passages Nicholas Montfort characterizes interactive fiction as a conversation between interactor and author through the medium; MS Paint Adventures eliminates the middleman entirely and turns the work into the product, not the medium, of a direct interactor-author conversation. Hussie's most recent MS Paint Adventures work, Homestuck, eliminates the suggestion box mechanic to allow himself greater control over the narrative's direction, but he compensates by using audience-produced music, art, and programming in occasional short animations and game-like interactive segments in between his usual webcomic panels, allowing audiences to participate in the creation of the world of the story as it is being told. The result is a transmedia epic of unprecedented scale, consisting of seven acts, five thousand single-panel comic strips, three short video games, about a dozen animated shorts, five albums of mostly unused soundtrack music, and a comic within the comic. As of writing it has been produced on a daily basis for two years and is still being updated.
Sandbox Environments
In the early 2000s many game studios answered Murray's call for "constructivist virtual environments" (147-152) by creating virtual worlds rich in interactive objects--books, telephones, pool tables, trash cans, and the like that could be used much like their real-life counterparts. 3D Realms's 1996 title Duke Nukem 3D first set the precedent for interactive object-rich environments, populating its cityscape-like levels with interactive pay phones, escalators, pool tables, toilets, and even stripper characters. Interacting with these objects had no significant effect on the game's gameplay or narrative progression, but the ability to interact with them transformed Duke's levels--which, like most other first person shooters at the time, were merely obstacle courses demarcated by sequences of color-coded keys and doors--into pieces of immersive, believable cityscapes, hinting at the existence of the rest of a virtual Los Angeles beyond the boundaries of their jagged skyscraper walls. Many games would follow in Duke's example, notably Remedy Entertainment's Max Payne (for which 3D Realms assisted as consultants), in which every virtual object in the game was created from photographs of objects in real-life New York, and Valve's Half-Life 2, which introduced a "gravity gun" that allowed players to weaponize toilets, soda cans, chests of drawers, or any other object in the environment, giving these world-establishing objects an additional purpose as gameplay resources. With its 2001 title Grand Theft Auto 3, Rockstar Games took the concept of the object-rich environment one step further and created the single player "sandbox"--a large, highly interactive simulated ecosystem populated by intelligent agents, in which a lone player is free to roam and interfere with the simulation in whatever way he or she desires. Many imitators soon followed, including Crackdown, Just Cause, and Streets of L.A.
Many of these games have bold narrative aspirations, and as Murray predicted, populating these games with these objects produced an unprecedented level of immersion in their respective virtual worlds.
Strikingly, however, the actual stories told by these games are typically quite conventional, and are not interactive in the slightest. Duke Nukem 3D, for all its much-celebrated one-liners and irreverent sense of humor, ultimately has no more sophisticated a narrative than Space Invaders--more framing device than plot, its story is strictly about a one-dimensional hero single-handedly thwarting an alien invasion through violence, facing no conflict other than the thousands of rounds of exotic ammunition fired in his direction. Max Payne is, by design, a story built entirely around the conventions of 1980s neo-noir cinema--one told almost entirely through the use of comic book cutscenes occasionally interrupted by the game itself. Grand Theft Auto 3's tale of criminal temptation and betrayal is so disassociated from the choices players make within its immersive sandbox world that the narrative almost feels irrelevant. (The player can murder every member of a particular gang, for example, and still be hired by that gang in the next mission as an enforcer if the story requires it.) None of these games tell an interactive story; they tell non-interactive stories that interrupt a game. Among mainstream titles, only Half-Life 2 is successful at creating the illusion that the player has any narrative agency--and that is only because so many of the decisions the player is forced to make are false dilemmas. (Climb down through this sewer grate, or run into a hail of bullets?)
A few game designers have taken a crack at deliberately designing open, interactive worlds around true player-driven narrative agency. Conceptually, these games can be categorized as "ant farms"--they are designed around the idea that if one drops a large number of players into a highly interactive sandbox-like environment and allows them to do whatever they wish, the players will create, drive, and guide narratives on their own.
The first and possibly best known example is Ultima Online, one of the first graphical massively multiplayer worlds, in which nearly all objects are interactive but in which gameplay, player roles and objectives, laws of conduct, and any other such considerations were left entirely to the will of a community of players [15]. A more structured approach emerged with the BYOND System, a narrative engine created with the intent of being one of the "artistic tools" Murray suggested technologists create (citation), in which a large group of interactors are assigned characters with roles, duties, and vague objectives, situated in an object-rich environment, but can fulfill them in any way they see fit.
The most serious problem with this approach is that it overestimates the willingness of players to accept the narrative roles assigned to them in a game where achievement is measured entirely in terms of the ludic system. Ultima Online and BYOND System games are games as well as narratives, and games, as Espen Aarseth and Gonzalo Frasca have noted (http://www.ludology.org/articles/ludology.htm citation), are ludic systems as well as stories. Contemporary game players, for which the conventions of ludic systems are not new, are accustomed to breaking narrative assumptions to exploit such systems for the sake of character achievement. This is perhaps best exemplified by the practice of "boosting", where players on opposing teams in Modern Warfare 2 enter gentlemen's agreements to take turns killing each other without resistance in order to quickly gain in-game experience points, with which they can acquire new weapons and equipment.
Whereas players may be attracted to UO and BYOND games by the promise of creating narrative, the lack of concrete objectives quickly turns the focus of UO and BYOND sessions to their underlying systems, making gameplay prevalently an exercise in repeatedly performing repetitive tasks with no narrative meaning or direction, such as hunting animals or forging barrel rims. players predictably respond to this tedium the same way people respond to tedium in real life: through mischief. UO quickly degenerated into a hellscape of endless banditry punctuated by the occasional totalitarian fiefdom, with an economy so wracked by inflation that a few players secured monopolies on common in-game resources and random murder and theft so rampant that player paranoia all but quashed any serious attempt at role-playing or continuous storytelling. BYOND games, too, have suffered a similar fate. The most popular BYOND title, Space Station 13, a meticulously crafted, open-ended homage to science fiction horror cinema that guides players through one of several plots typical to the genre, also typically degenerates into random acts of murder, obliterating any pretense of drama or suspense produced by player characters being picked off one by one by corporate traitors or shapeshifting alien xenomorphs, and making most experienced players too afraid of each other to attempt meaningful inter-character relationships.
Bay 12 Games's Dwarf Fortress takes the reverse approach, randomly generating an incredibly detailed virtual world populated entirely by randomly generated intelligent agent characters, each of which generates its own personal narrative by interfacing with the simulation in attempts to fulfill roles assigned by the player. (A dwarf, for example, who is generated afraid of water but fond of the color purple, if told by the player to go fishing, will wander tentatively along the riverbank, looking for easy catches in shallow water, occasionally succumbing to the temptation to pick violets at the water's edge. This is remarkably close to the direction Murray anticipates for "goal-based critters." 226-233.) In this manner Dwarf Fortress provides emergent, occasionally compelling narrative, over which the player has true agency.
Beyond AI: The Human Holodeck
A common thread in contemporary approaches to interactive media is the use of human intelligence to overcome the limitations of current artificial intelligence technology. Artificial intelligence is notoriously terrible at being spontaneous, interesting, unpredictable, relatable, or subject to personal development over time--all qualities at which human beings excel. Where Grand Theft Auto 3 is populated by a thousand clones of the same five schizophrenic pedestrian, who shrugs off a hit from a car but turns violent at being bumped on the sidewalk, the cities of Ragnarok Online are loud with haggling merchants, flirting lovers, sulking mercenaries, and cantankerous priest buses[16]. Where Jedi Knight II offers the most hackneyed imaginable struggle against a Sith Lord with no personality, its multiplayer mode is used to settle real-life scores between good and evil.
Such dramatic dichotomies do not represent a failure of artificial intelligence to live up to Murray's aspirations of a truly literary cyberdramatic medium, but a promising alternative. Who is better suited to giving form to a human being's flights of fancy than a human being? Computer scientists are well trained in designing systems around using the best possible tools for the job, utilizing the strengths of some software components to overcome the weaknesses of others. It makes sense, then, that if the holy grail of a truly literary form of cyberdrama, capable of bearing Shakespearean genius, is to be achieved, that the parts of the form computers do best--virtual world representation, networking, pseudo-randomness, brute force computation, branching paths, the rapid generation of values for many random variables--should be left to computers, whereas the parts that humans do best--creativity, character building, suspense, theme, context, universal resonance--should be left to humans.
The holodeck in Star Trek may have been driven by an artificial intelligence, but the stories told about the holodeck in Star Trek episodes were written by human beings about human characters. Would a human-powered holodeck not be the most reliable way to create open-ended, evocative, compelling cyberdrama? Human storytelling gave us the timeless characters, witty, original prose, and exploration of broad historical and personal themes that made Hamlet an immortal work of theater; surely it is also capable of producing a Hamlet on the holodeck. My approach seeks to leverage the advantages of human over artificial intelligence in the following ways.
In his book Cybertext: Perspectives on Ergodic Literature Espen Aarseth argues that treating games as narrative misses the point that they are also games. Insofar that efforts to combine games and narrative by concatenating them--narrative-breaking cutscenes and text boxes being well-reviled examples--have generally been poorly received, he is right. However, as games, like any technology, are a social construction, games are not exclusive from narrative media if treated as such. The key to creating compelling interactive narrative in the vein of the Star Trek holodeck is to create a medium that is not defined purely as a ludic system, as Aarseth defines games, nor purely as a narrative medium, as Murray defines it, nor as something outside but partially inclusive of both concepts, as per Frasca's ludological approach. The proposed cyberdrama medium would conceptually be a mutual subset of both narrative and ludus--the intersection, not concatenation, of the two.
Thus, the system should fit a definition of a video game similar to that proposed by veteran designer Chris Crawford in Chris Crawford On Game Design: it should have rules, an objective, one or more players, and a means of achieving the objective. It should also fit a broad definition of narrative, which includes a plot, one or more characters, character conflict, and the potential for narrative tension and exploration of theme. [17] For narrative to be interactive in the mold of the Star Trek holodeck, it should, as per Murray's criteria, immerse the player in the virtual world, provide the player with true agency over narrative development and outcome, and transform the player into an agent of the story.
I propose the concept of cooperative improvisational storytelling as a framework for developing this system. The general idea is that a team of players will cooperatively compete against the computer, with the act of storytelling itself being a gameplay element.
In one example of a potential cooperative improvisational storytelling system, I borrow from the unlikely example of role-playing therapy sessions. Suppose we have two human players, each who represents a team of characters. Both teams will have the same objective--to reconcile a pair of lovers, for example, or to escape a haunted house alive. Each character will have a name randomly selected from a list by the computer, as well as a list of personality quirks randomly selected from another list, a positive relationship with a character on the same team (such as friend, co-worker, or lover), and a negative relationship with a character on the opposite team (such as rival, business competitor, or creditor). Each turn, each player will select a character from his or her team and improvise a conversation with the selected character from the other team. (This would be the only permissible interaction between characters.) The characters would then be dropped into a scene in a virtual room, which the players could interact with in context-specfic ways (e.g. a banana could be eaten, peeled, or thrown, and a balloon could be held, rubbed, or popped).
Unlike traditional role-playing, where player actions have immediate consequences, this hypothetical game would allow players little direct control over the plot. Plot events, chosen by a weighted Bayesian algorithm from a pool of genre-appropriate beginnings, middles, and ends previously created by other users, would be generated by a non-character AI between scenes. The setting for each character would populate the game environment with a contextually relevant set of algorithmically selected objects; for example, a scene in a film noir game might have the characters sitting in a diner, with one character holding a smoking pistol and the other a wallet-sized photo of a character from the previous scene, with thirty empty coffee cups on the table. Players would instinctively construct a context for their situation--constructing narrative out of random data is part of human nature--and role-play around it.
At the end of each scene, the computer would ask each player questions about the other player's character in the scene. The answers to these questions would deliberately bias the otherwise random selection of the next plot event. For example, in the previously described scene, the computer may ask each player, "Does your character trust the other player's character?" If both players answered no, the next scene could begin with the next two characters each hiding an item from each other. If one player answered yes and the other no, the next scene could begin with the trusting character's corpse lying on the floor. If both players answered yes, the next scene could begin with both new characters sitting in a burning building, staring at a briefcase full of marked bills.
In narrative terms, the players would control the characters, while the computer would control the setting and plot. In role-playing terms, this is a game where the players are the characters and the AI is the gamemaster. The goal of a live role-playing therapy system is to overcome a personal demon--alcoholism, domestic abuse, aggression--that tends to steer real-world conflicts into unfavorable outcomes; in this kind of game the demon would be the computer's capricious control of plot.
The potential of such a system lies in the fact that players have no inkling what is going to happen, yet both drive the story and have an incentive to form meaningful, complex relationships between characters. As a product of the game's ludic system, the players would naturally create an emergent, unpredictable story that neither relies on cumbersome AI for character behavior nor a competition through a ludic system irrelevant to the narrative. It would be a game of social skills, based on character relationships, not individual characterization or plot--role manipulation rather than role playing. The fun of the game would lie not in being surprised by the plot, but in anticipating and working around plot tropes and developing interesting character relationships to subvert them. There would be an inevitable element of tragedy in these stories, which could be used to great effect--players would have ultimate freedom to make their characters say or do whatever they wanted, but little control over the consequences. It would also be an interesting exploration of the nature of a game as a literal extension of the self.
In being forced to rapidly role-play different characters in different roles over a short period of time, players would also invariably develop meta-characters around their storytelling styles. One set of possible endings could be more favorable to one player's characters whereas the other could be favorable to the other, encouraging players to compete against each other to achieve the most favorable outcome, implicitly imbuing the stories they produce with a powerful underlying tension, and creating the illusion of powerful cosmic forces guiding the characters' fates. Ludic elements could reward skillful players with a small amount of plot influence, such as the introduction of a particular object into the next scene or a plot device eliminating an opponent's character from the narrative. Two players wishing to see different endings may even find themselves in a game of the prisoner's dilemma, trying to anticipate each other's actions in order to achieve the ending they prefer. Experienced play could result in the development of strategies around certain objects or plot events known to have a dramatic effect on the outcome, with players engaged in a deep, complex tactical battle over control of these elements, like Greek gods fighting over the fate of their champions. Or they could cooperate against a particularly nasty plot AI, which would be configured to overwhelm players' characters with negative emotional events and difficult moral dilemma in a test of the players' moral integrity.
This is but one way the cooperative storytelling paradigm could be realized. An alternative could be a virtual world version of the parlor game Exquisite Corpse, in which two players would take turns adding objects and characters to a virtual room, and a third player would wander the final scene, constructing a context and an objective, and attempting to achieve it. Another could be a storytelling adaptation of the tower defense genre, in which a team of players must help a pair of characters protect their marriage from a soap-operaesque litany of personal tragedies, or a Faust figure from an endless march of moral temptations, with a Greek chorus of online observers voting on which player is winning. The idea would be to combine the essence of personal reflection and achievement Turkle observed in early arcade gamers [18] with the excitement of self-identified conflict Murray observes in fighting games, while also opening up endless possibilities for narrative catharsis. Such games would transform the overcoming of personal challenges, a great pleasure of narrative rarely seen in games, into dramatic contests of character.
Slings, Arrows, and the End of Outrageous Fortune
Murray wrote Hamlet on the Holodeck in 1997, a time when the game industry was enjoying explosive growth, innovation in data media like the CD-ROM and the Internet were blowing away the storage limitations that had constrained previous experiments in interactive media, and narrative-heavy games like Final Fantasy VII and Myst were dominating sales charts. Her research, and that of Espen Aarseth, drove intense academic interest in hypertext and interactive storytelling, encouraged by experiments from studios whose teams grew from tens to hundreds as they strove to produce million-dollar narrative masterpieces on the scale of Hollywood blockbusters. (cite) This trend was not financially sustainable, however, as the price of video games stayed relatively stable while development costs rocketed sky-high, and unfortunately the commercial gaming industry now associates narrative-heavy games with tremendous risk and expense. The end result is a new paradigm in which games are becoming more purely ludic, with scrolling text boxes, animated cut scenes, and other '90s experiments in bringing interactive storytelling to games now cemented into convention rather than being explored to their next iteration. With the game industry losing interest in defining their games as cyberdrama, the task of developing the medium outside of games (or within the specialized niche of "art games") has fallen to tiny, increasingly obscure circle of artists, writers, designers, and academics, often completely oblivious to each other's work. Soon there will be a generation of professional game designers who do not even know what a holodeck is.
And yet, Heavy Rain, the spiritual sequel to Indigo Prophecy, a game built upon Murray's multiform narrative system, was released just last year in 2010. The adventure gaming genre has seen a revival, with the recent releases of Sam and Max and Penny Arcade Adventures. Fandom thrives, inventing all sorts of collaborative literary media that academia is just starting to find names for The quest for meaningful, compelling, truly interactive cyberdrama, intended or unintended, still continues.
After all, after thirteen years, years that have seen many of Murray's proposed innovations come to fruition, the end of the television franchise that introduced the concept of the holodeck, and the gradual obsolescence of the "cyber-" prefix by which Murray defines cyberdrama, interactive media still awaits its Hamlet.
Perhaps the dream was too ambitious. Perhaps our understanding of semiotics is insufficient, our deconstruction of the nature of media too incomplete, for us to build what we enjoy best from traditional media from the ground up. Perhaps we still have yet to find the Shakespeare of the Internet age. Or perhaps, in the most powerful piece of computing technology ever harnessed by man, we have, all along, had everything we needed to project the incomprehensible magnitude of human experience into the virtual world.
The holodeck still awaits.
Notes
Bibliography
Adventure Quest. By Richard Lovejoy. Dir. Ivanna Cullinan. Perf. Brick Theater. Brooklyn, NY. 24 July 2009.
“Boatmurdered.” Let’s Play Archive, 14 Apr. 2007. Web. 24 Dec. 2010. <http://lparchive.org/Dwarf-Fortress-Boatmurdered/>
Crawford, Chris. Chris Crawford On Game Design. 2003. Indianpolis, IN: New Riders Pub, 2003. 5-9.
“Criticism of Ultima Online.” Wikipedia, The Free Encyclopedia. Wikimedia Foundation, Inc. 14 January 2011. Web. 31 January 2011. <http://en.wikipedia.org/wiki/Criticism_of_Ultima_Online#Abuse_of_guard_zones>
Dwarf Fortress. Bay 12 Games. Vers. 0.31.18. 16 Nov. 2010. <http://www.bay12games.com/dwarves/>
Barthes, Roland. The Pleasure of the Text. Trans. Richard Miller. New York, NY: Hill and Wang, 1975.
“Bow, Nigger.” Always Black, n.d.. Web. 26 Dec. 2010. <http://www.alwaysblack.com/blackbox/bownigger.html>
Frasca, Gonzalo. “Ludology and Narratology: Similitude and Differences Between (Video)games and Narrative.” Ludology.org., n.d. Web. 28 December 2010. <http://www.ludology.org/articles/ludology.htm>
Gillen, Kieron. “The New Games Journalism.” Always Black, n.d. Web. 26 Dec 2010. <http://alwaysblack.com/blackbox/ngj.html>
Hussie, Andrew. Homestuck, 2009-2011. Web. 31 January 2011. <http://www.mspaintadventures.com/?s=6 >
Joseph, Paul. “Ultima Online: Justice in a Virtual World.” Picturing Justice, 1998. Web. 31 January 2011. < http://usf.usfca.edu/pj/articles/Ultima.htm>
L'Engle, Madeleine. A Wrinkle in Time. New York, NY: Yearling Books, 1973. 117.
Montefort, Nick. Twisty Little Passages: An Approach to Interactive Fiction. Cambridge, MA: MIT Press, 2005.
Murray, Janet. Hamlet on the Holodeck: The Future of Narrative in Cyberspace. New York, NY: The Free Press, 1997.
"The Reverse Truth." A Tale In The Desert Forums. talos. 2009 Feb. 17. Web. 2011 Jan. 30. <http://www.atitd.org/forum/viewtopic.php?f=8&t=493>
Space Station 13. BYOND, n.d. Vers. 13. 22 Dec. 2010. <http://www.byond.com/games/Exadv1/SpaceStation13>
Swartz, Luke. “Why People Hate the Paperclip: Labels, Appearance, Behavior, and Social Responses to User Interface Agents.” Stanford Computer Science, n.d. Web. 31 January 2011. <http://xenon.stanford.edu/~lswartz/paperclip/>
Turkle, Sherry. The Second Self: Computers and the Human Spirit. Cambridge, MA: MIT Press, 2005.
Wiki Resolution, n.d. Web. 30 Jan 2011. <http://editthis.info/wiki_resolution/Main_Page>
The Wrench. By Brian House, Sue Huang, Nathan Phillips, et al. Perf. Brick Theater. Brooklyn, NY. 24 July 2009.
World Resolution, n.d. Web. 31 January 2010. <http://community.livejournal.com/worldresolution/profile>