The last day of AGDC was an excellent day, with two talks in particular that led to a good deal of spirited, academic discussion about storytelling and a third lecture that demonstrated some very slick next-gen controllers that could have a significant impact in the future on game design and interface.
The first talk of the day was given by Andrew Stern, he of Façade fame, although he did not focus specifically on the accomplishments of that project. Instead, his talk, provocatively titled “Linearity is Hell: Achieving Truly Dynamic Stories in Games,” explored the possibility of truly dynamic storytelling in games and how a system like that might be designed. Stern did acknowledge that this was more of a theoretical talk and that he has no claim to a solution for this; rather, he was hoping to express his understanding of what such a system might entail.
He began by describing the well-known problem of combinatorial explosiveness that typically characterizes systems that try to accomodate branching storylines, and proceeded to offer that one solution to this would be “story generativity”, or a system that generates story nodes on the fly (akin to “procedural storytelling”), which would be capable of providing players with the Big Three game elements we seem to most covet: freedom, well-formed story, and agency; whereas most games these days are typically able to provide at most two of those three.
Without going into too much detail here, he related this type of system most closely with improvisational theater, and the Oz Project in particular, where improv actors received direction through headsets for a single independent participant. The idea for a dynamic storytelling game, generally speaking, is to accomplish something similar through the creation of behaviors for NPCs which help direct and guide them through a narrative scene, encompassing such things as motivations, goals, dynamics, and dialog.
Interestingly, Stern briefly discussed what he called a “calculus” for NPC dynamics, which comprised defining different narrative states to track, and then using various approaches to compute the narrative state; for instance, with a narrative state of romantic interest, the idea is to use a calculus (fuzzy, algebraic, statistical, etc.) to evaluate things like flirtations, responses to NPC advances, tone and language of the player, and so on. It immediately brough to mind some of the approaches embraced by Chris Crawford, specifically the use of mathematics and algorithms for different aspects of character appraisal and storytelling, so perhaps there are some similarities there.
He did acknowledge that the type of system he described would be more adept at generating sequences rather than sentences, and in that respect the system would probably not be generative enough to be a true end product. But to me it seemed that the real challenge is precisely in going from structure to presentation; that is, while the mechanics of creating and manipulating the components of story, including character motivations and behaviors in response to (or in lieu of) player actions, do appear to have features amenable to mathematical or algorithmic control, it’s the process of taking the product of those calculations and algorithms and working them into a coherent narrative where the true challenges lie ahead. And I would propose that this final step, the delivery of the story, is where much of the art of storytelling exists — stories can easily be broken down to reveal their components and structure, but the presentation of that structure is dependent on the skill of the storyteller. Can that ever be done procedurally? In a way that is artistic and moving?
I’m reminded again of Crawford, this time of the product of his work on the Storytron, after playing a bit with the online demo of Balance of Power: 21st Century. There’s much I could write about the system, and it’s clear a massive amount of work has gone into the modeling and implementation of relationships and character interactions — certainly a significant accomplishment. I don’t know Crawford’s intentions for the system beyond BoP:21K, or whether future storygames (like the one in progress by his colleague, Laura Mixon) will look or play differently, so I could be off base; right now, though, the presentation has the look and feel of a basic story structure and its components, with none of the crucial dressing of language and style that an author provides. That is to say, the storygame is plenty of substance without any style. Does this work? Perhaps for others, not so much for me. Could it be more? I’m not sure.
The discussion that followed the talk was spirited and entertaining. Points raised included how a system like this could incorporate certain issues with language, writing style, and humor, which are notoriously difficult to reproduce procedurally. Others noted that the system he described sounds like something better suited to drama management than actual end-dialog or behavior generation, and that perhaps it matters more how the story is delivered rather than constructed. Stern argued that this type of system does not have to entirely remove the author’s voice from the output, and that he sees it as more of a 50/50 relationship. I think the system is still too theoretical to be able to envision that myself.
I guess it really comes down to the question of what we really want from the stories in our games, and which we find more influential: substance, or style? Personally, I think both play an important role, but it’s still too difficult to imagine a system that can procedurally handle both with equal skill.
It’s fair to say, though, that Stern’s talk was accepted by the speaker and audience as mostly hypothesis, with the intention of stimulating discussion and debate. And in that sense, I think the talk succeeded far better than most at the conference. Plenty of people stayed behind afterward to continue the discussion, and it’s likely to continue even longer on e-mail discussions and blogs. And it made for a very entertaining and enjoyable experience.
More to come later on the rest of Day Three as I get back in the swing of things.
Enjoyed this article? Subscribe to The Monk's Brew RSS feed.
The more I think about stuff like Storytron and other generative interactive technologies, the more I wonder if the future of generative story-telling will not depend on the current standards of how an author puts a sentence together. It seems to me that the techniques of storytelling have changed dramatically throughout human history — consider the difference between something like The Odyssey and Ulysses. If we’re looking at generative technology through the lens of current tastes, we may be doing our future Homer and Joyce a disservice.