Thursday, April 17, 2008

I for one welcome our new story overlords

From the drowning of suspected witches in pre-Reformation Europe to Star Trek's Kobayashi Maru scenario, no-win situations have been seen as a way to test one's character. When faced with disastrous loss, how does one behave? As Star Trek's performance test makes obvious, it's all so much more useful when this is done in a simulator, and not in real life, when lives are on the line. That's the kind of setup that Swartout et al describe in their paper, "Toward the Holodeck: Integrating Graphics, Sound, Character and Story."

They have set up the Mission Rehearsal Exercise Project for Army testing, creating an in-country scenario for new officers to experience dilemmas and associated problem solving. Along the way, they've faced some of the most challenging problems in interactive narrative. In solving it, they come up with a story engine they call StoryNet, and develop a hybrid approach to modeling virtual humans. For the least involved characters, their actions are scripted, while more integral characters have AI reasoning components -- the most central of which use additional emotional simulation.

StoryNet is probably the most interesting part of their overall solutions, the rest being important but necessary technical compromises due to modern limitations. Regardless of how good our technological base is, we still face a fundamental question of how stories are generated in an interactive context. The paper admits that early attempts had little capacity for getting off-track and must have essentially been an exercise in getting the individual being trained to think through the proper steps. As valuable as that might be, the bigger question of having your StoryNet respond to ad-hoc player/trainee decisions is still very open.

The MRE StoryNet solution is a hybrid one. It allows free-play within certain nodes of activity, while certain "hook" actions lead down links to other nodes -- essentially a (well-?)disguised choose your own adventure. While Swartout et al explicitly reject that sort of obvious mechanism, I wonder how many attempts a single individual could make before the "seams" started showing. That's not to fully criticize what they've accomplished however -- I'm sure it meets the specific needs they've been funded for. However, it leaves open the question of how much work it would take to change to a different scenario. It would probably require a fundamental overall including all the assets and reauthoring of the story line and its component nodes.

This remains the open question we face. How do you accomplish meaningful narrative, while still allowing elegant changes to the scenario and sophisticated interaction? The MRE is one approach, and there are many more...

Wednesday, April 16, 2008

Pervasiveness and the Magical Circle

In the world of computer games, a player usually knows exactly when they enter the game world. The click of a certain icon, the animated splash screen, the opening theme music. However, in the real world, this has not always been the case. When Roger Caillois first coined the terms and ludus and paidia to distinguish between formal game events and playful exchanges in 1958, he was getting at the sort of difference I'm thinking about. The liminal experience of crossing the threshold of non-game to game has been called the "magic circle" experience, and Nieuwdorp's paper "The Pervasive Interface: Tracing the Magic Circle" is all about how new digital experiences are crossing that threshold in new (and yet old) ways.

Nieuwdorp's emphasis in the interface experience, citing new and freeing innovations in technology that are taking games from the world of the stationary PC into the everyday world in the form of mobile gaming, augmented reality, and similar ludic experiences. He sees the innovation coming because the "geographical setting of a pervasive game is shared with an already existing environment with laws and conventions of its own." His example of "The Go Game" -- played in an urban environment by receiving text messages via cell phone -- demonstrates how everyday city streets can be turned into game elements by the introduction of a game setting and status. He turns to Johan Huizinga's definition of the magical circle as a "temporary world within the ordinary world, dedicated to the performance of an act apart." In this way, ordinary objects are transformed by a game interface (no matter how metaphorical the transformation in the case of The Go Game) into ludic elements.

He cites Goffman's categories to find three axes of definition for such objects:
1. rules of irrelevance (game pieces as playable items not intrinsic value)
2. transformation rules (patterns that denote how much influence an item has within the game)
3. realized resources (availability of material and moves)

While debates as to the theoretical presence of a game experience may seem a little too far removed from practice to matter, they serve the important goal of informing designers. In order to design new games and experiences, one must know the rules of said design, and Nieuwdorp's paper helps formalize the sort of experience that is becoming mainstream and in fact almost defining of a new type of game: at once casual and formal, present and abstract. In case that needs defining I refer to game-like experiences he cites, such as alternate reality games like "I Love Bees" -- we participate in the "game" but fumble to know the rules of participation while its true impact on us is partially formed by its very ordinary-ness and normalness as a regular-seeming website within our web browser.

Autonomous Characters and Animation

Bill Tomlinson's paper concerns itself with one of the fundamental problems facing the portrayal of believable characters within a computer game: the automation of their interactive animation. If a character is to cause us to emphasize, then it should behave in humanistic ways. Essentially, this is the second of the two main problems encountered by the creation of agents such as the movie S1m0ne's synonymous AI movie star (the first problem is behavioural automation). Indeed, this paper makes the connections back to Hollywood animation, where linear procedures (whether traditional or digital) allow animators to breath life into their characters with a mimicry of human activity.

In a computer game, where dynamic behaviour is desirable -- that is, the automatic performance of activities by a character -- the next question becomes, how will that behaviour be portrayed? Tomlinson locates these differences along 5 axes: intelligence, emotional expressiveness, navigation collision avoidance, transitions, and multi-character interaction. In order, these require animators to specify suites of behaviour to deal with visible decision making, feelings, the intersection of two characters while moving, the smooth look of animations in between specific states, and those complicated sets of actions such as a fist-fight or a romantic scene.

In each of these cases, Tomlinson mentions recent attempts to solve these problems, but his greatest success seems to be a thorough revealing of the problematic areas of dynamic character animation.

Interdisciplinary Analysis

When we want to build computer models for any kind of human behaviour, whether automated or not, we must turn to realistic source material as a basis for our work. All too often, bastions within the ivory tower ensure that people with expert knowledge don't collaborate in ways that combine their respective disciplines. Seif El-Nasr, coordinator of the SIAT GameLab (now EMIEE), writes convincingly from a multi-disciplinary point of view, combining a formal computer science education with film and theatre training.

Her paper on "Interaction, Narrative, and Drama" starts by examining various forms of filmic flow. Flow, of course, is that amazingly concise term that encapsulates that state of being absorbed at peak capacity in a task. Coined by Csikszentmihalyi in 1990, it has been found in
film-land by Boorstin, who finds the "voyeuristic eye," the "vicarious eye," and the "visceral eye" describe the differing sense of engagement a viewer can have. Seif El-Nasr moves on to the debate by game developers over the sense of fun and engagement felt by the player of a video game. Developing this kind of theory is crucial to inform game designers in their efforts. Seif El-Nasr identifies a lack of theory and game design behind the "vicarious eye" type of pleasure, that "feeling of empathy through an understanding of characters’ emotions and choices." This could be because allowing the user to be surprised or learn along side a character, or cause them to respond emotionally to a character's actions (the voyeuristic and visceral eye, respectively) are more easily facilitated by computer games.

This is just one example of how Seif El-Nasr bridges filmic theory and game design theory. Other examples include Stanislavski's acting and directing theories. From this theoretic background, she proposes several dramatic techniques that can be used in video games to connect the player and the characters. These include: the ticking clock, character tactics, character goals. Useful definition of story engine methodology and agent models are presented.

While this is not a complete review of this paper, it covers its highlights, especially one of its most notable highlights, that of a comprehensive interdisciplinary approach to the work. That combination provides a legitimate real world set of experiences that ground the theory.


Thursday, February 7, 2008

Systemic level design for emergent play– Harvey Smith at GDC

Emergent play seems very powerful to me. Something about the verisimilitude of characters imbued with goals left alone to achieve them really speaks to me. I think it's a sense of "Hamlet on the Holodeck" -- emphasis on the holodeck. However, our best contenporary example of that is the Sims, which forms micronarratives but no coherent story. Even from a ludic stand-point, I doubt such a free range of actions makes sense for much gameplay. So clearly, constraints are key.

Smith's talk is about that kind of balance, especially necessary for our imprecise mastery of artificial life. He diagrams a method that can employ modern best practices in software engineering to great convenience and effect. Things that are the same should behave the same by default, then get any special modifiers (environment, mood, motiviation, etc) added on. This is especially handy for inanimate objects. This is essentially Crawford's ratio of interactive excellence from The Art of Interactive Design (85), where the number of accessible states over the number of conceivable states should approach 1 for best results. Avoiding depicting doors that can't open (Prince of Persia, I'm looking at you!), and otherwise usable items that are actually just part of the background is a great start at reducing those conceivable states. Especially when some ledges are graspable, others are walkable, and others just painted onto walls for effect (and instant death when you try to jump onto them). And why can the Prince pick up a fallen foe's saber, but not a bow?

The challenge and need for balance comes from narrative considerations. To avoid "toy-like" environments (Sims etc) the player needs to be led/driven/coerced through some series of events/levels/whatever... because we construct narratives linearly, and if the designer cares what narrative we put together, then she cares what order we see these events/levels/whatever in. Even if we withdraw to a purely ludic point of view, challenging gameplay often calls for a change in states, increases of difficulty, or introductions of novel items. The challenge is -- how much of a special case are these?

From Smith's POV (and I too remember Deus Ex fondly), besides the software engineering benefits which I can affirm, his principles give direct benefits to the player. A sense of immersion in a consistent world. This allows for plan-making, understanding, and achievements. On the other hand, you still have those "one-ofs" that are required (for some definition of required) and stand alone as de facto special cases anyways, getting none of those benefits. And emergence, while creating powerful "second order consequences," is too unreliable to count on for important narrative events.

I think we need to find some formal constraint language for controlling emergence.

Wednesday, February 6, 2008

Prince of Persia: Rival Swords

So I'm watching Pat play Super Mario Galaxy on the Wii. A couple of Mario's new moves really surprised me. I say surprised because they were actually fairly familiar. One move allows Mario to jump against a wall, hang on, then jump higher against the opposing wall, and so on, all the way up a "chimney" (close walls). When he's close to the edge of a narrow walkway, he hangs off the edge. I think Mario would feel right at home in the Prince of Persia universe, except for his awkward tendency of trying to jump on his enemies heads rather than slash at them with a saber. Oh yeah, there's the same awkward third-person camera angles you get to fiddle with.

I was struck by an obvious but still enjoyable sense of nostalgia playing Prince of Persia. It evoked distinct memories of being back in grade 6 at middle school playing the original (game, perhaps not version -- I don't know if it was a port) Prince. I remember how the Prince skidded to a halt when you stopped running, how he whipped out his saber when you had to fight, the 8-bit spikes at the bottom of pits and all the rest. In that sense this Prince hits home in just the right way. Lots of new toys, but enough of the original to pull in original players. I haven't played the others in this new series, so I can't really say anything to series fatigue, or how well the overall story line plays out.

The most applicable reading from last week would be Jenkins' 'Narrative Architecture.' I admire how he tries to carve that pragmatic middle ground where the narrative aspects of games are allowed to be relevant and designers are allowed to try improve their presence in games. Prince is a game that's had quite a lot of environment design done, in order to create a real-seeming backdrop. However, since this is a puzzler, much of this is static, as opposed to, say, Assassin's Creed's more life-like crowds of native speakers. Nonetheless, the sense of exploring these ruins in a land and time far away is a good one to nurture, as it directly strengthens the believability of the puzzles and traps that await you.

Your NPC foes seemed somehow lackluster. That's likely a result of playing on easy, and I kept wondering silly things like why the archers didn't have backup swords anyways. On the other hand, when I say lackluster I refer to more than their fighting prowess. Their presence in various places is definitely just as scene-based as the traps etc (ie you have to accept that as a trope of the game, it doesn't make sense except as a predefined problem needing a solution). They also seemed strangely deaf, and even if you messed up the stealth kill of one guy, his partner would still be oblivious. Strange, but since I like to notice AI behavior in games, it seemed noteworthy.

It also seems worth noting that Prince is a game that doesn't want emergent behaviour from its NPCs. While some games with systematic level design can afford to have its NPCs act on their motivations and move around levels, Prince absolutely requires this almost-cinematic level of precision to a well-accomplished scene. Having all the guards decide to meet in one room to defend the castle would bypass that entirely, as well as probably trigger most of the traps. :)
Oh narrative ambitions. What will happen to you when we give computer characters goals and amibitions?

Wednesday, January 30, 2008

Assassin's Creed

This continues in the tradition of 'stealth' games popularized by the Thief series, and then later by Splinter Cell, and perhaps Deus Ex if you squint your eyes a little enough. This type of game seems to focus on providing a detailed world simulation, often including localized emergent interactions. It would be interesting to do a more detailed comparison to see if these are a sort of response to "unrealistic" first-person-shooters, since they focus on providing dramatic consequences for straying outside normal behaviours. This does stand in opposition to games where out-and-out gunplay is what the non-player-characters expect and respond in kind, whether intentionally (or with an agenda) or not.

Assassin's Creed, in my short experience with the game, stands out as having great graphics, a very immersive sense of spatiality, but lacking a little in the overall story-line. That may well not matter to most gamers, especially those for whom "the play's the thing." To them, only the repetition in certain quests (eg gathering redundant flags) may damper their enthusiasm. Running and climbing and jumping and kill and hiding like a medieval assassin covers a multitude of sins, as they (don't really, I guess) say.

So from a game-play perspective, I don't have too much to criticize. I especially like the "social stealth" aspects with the blatant mapping of low profile to socially acceptable actions and high profile to socially unacceptable actions. While the initial introduction of these features seemed a little cheesy, these game-play mechanisms served to emphasize the important elements of the game -- staying "hidden." I think making it explicit what people will tolerate made for an interesting design territory. If anything, I think they failed to fully mine this fertile ground. The protagonist's actions and methods could have been questioned and problematic. Instead, efficacy remains your by-word, and instead this seeming spiritual man seems to have no qualms about his actions, something either a ordinary Christian or Muslim of his time might have (and apparently the protagonist descends from a mother who is one and a father who is the other). While I'm truly delving into "you should source this properly land", it appears medieval assassins belonged to an offshoot Muslim sect persecuted as infidels by other Muslims, and whose main quarrels were with them. Nonetheless the Crusaders played a big part in defining them, so I suppose they can be the "big bads" of this game.

Oh well, enough of that. From this game, I derive the idea that explicitly linking your character's maneuvers to narrative goals seems like open design territory. Also, making up a ridiculous and pointless metanarrative that's approximately one micrometer removed from "it's all a dream" is dumb, even if upper management ordered it late in the design process and it wasn't really your fault and hey we've all been there amirite?