January 20, 2008
Malcolm Ryan, a frequent commenter on here who several of us recently heard undertaking to analyze a page of Peter Rabbit, has apparently been circulating the beginning of a manifesto pertaining to interactive narrative and drama systems. Most people in this area think of Dogz when they hear “dogmatic,” but there’s also the obedience to doctrine that the Dogme 95 filmmakers have pledged and the Dogma 2001 vows Ernest Adams proposed, following the lead of those filmmakers.
Grand Text Auto has managed to obtain the first vow from this still-secret document:
As a signatory to the 2007 Narrative AI Manifesto, I do hereby solemnly vow that:
1) I will no longer assign events a numeric “tension” value and plot a story as an Aristotelean “dramatic arc”. I recognise that real stories contain situations and devices that are more semantically complex than can be represented by a sequence of numbers that rises and falls.
What else might be in store? We have some ideas …
?) A programming language, subsystem, or component may not be used in the development of the work if its name is an acronym.
?) Knights, elves, dwarves, dragons and the like are the only creatures that may be represented. I recognize that the portrayal of couples undergoing a marital crisis and other ordinary people in domestic situations is alienating to the largest group of people interested in computer-simulated characters.
?) Interaction must be completely based around verbs. Only pure verbs may be supported. Verbs which are also nouns will not be permitted.
?) Development of some sound, music, and art assets is allowed, but all developers must use identical consumer laptops and sit together at the same folding banquet table.
Do you have other ideas?
January 21st, 2008 at 1:25 am
How about: I will not resort to dynamic player modeling to mitigate my game’s design flaws. Ie. Measuring what the player is doing and adjusting the game to it can easily become a crutch for not giving them an interesting & well balanced set of things to do in the first place.
January 21st, 2008 at 4:58 pm
I should explain this ‘manifesto’ a little further. As I was saying in the email discussion, without wanting to denigrate any work that has come before, I think it is time to name a couple of ideas which have reached the limits of their usefulness to encourage us all to move on and try more ambitious things.
The “dramatic arc” idea is a good example. I mean no offence to Michael and Andrew — quite the opposite, their work has been ground-breaking — but as a result I think many of the more CS-minded among us have latched onto the idea of representing drama as a number that goes up and down. Such a representation is very attractive to programmers, and the temptation is to hone it into a sparkling jewel, but I think ultimately this will be a dead end.
Of course, the one who is flying should never listen to the one on the ground who says it is impossible. I admit that I have a big mouth for someone with no results of his own. My ‘manifesto’ is largely tongue-in-cheek, but if we can stomach some critical self-reflection, this might be a useful exercise.
January 21st, 2008 at 8:57 pm
I hate to get serious, even for a moment, but I feel like I should mention a difference between movies, video games, and AI-driven interactive narrative. In the first two cases, there are many examples produced by huge industrialized teams with immense budgets. Dogme 95 and Adams’s version of it for video games were both reactions to this. In the last case, there are almost no examples at all, and they few that there are have been provided by individuals and small bands of outsiders with little funding. To me, the problem is not that tension is parameterized; it’s that tension is almost always not parameterized because almost no ones creates AI-driven interactive narrative.
January 22nd, 2008 at 1:40 am
Malcolm, this is an interesting provocation!
I love the Dogme films and the spirit behind them: stripping away spectacle and overwrought production methods to focus on the core story and actors’ performances. (Lars von Trier is one of my favorite writer/directors.)
I appreciate the idea of a Dogme approach to building interactive stories, and am sympathetic towards it. We’ve been known to consider manifestos, lay out lists and agendas for the goal of advancing the state of the art of interactive stories, and bemoan overproduction.
Thinking about this idea, I realized there is already an interactive story presentation format that is Dogme-like in form: text! Stories presented in text (perhaps with minimal graphics) should allow authors to focus their energies on the story itself. Text-based interactive fiction, Masq, Storytron, are all Dogme-like in that way; authors in these formats spend little or no time on presentation spectacle.
But Malcolm, I think you intend something else: you state you want move past ideas that have reached the limits of their usefulness, in order to push interactive stories further. (That doesn’t seem like a Dogme thing to me exactly, but anyhow.)
While I’m happy to promote new techniques, I’m not ready to say what techniques have lived beyond their usefulness. For example, I still think Eliza is (not was, but still is) among the most powerful interactive character/story experiences yet built, and I could imagine using variations of its techniques in certain situations. Or, while I would never want to build a fixed branching story like a Choose-Your-Own-Adventure, it’d be kind of cool to make a system that can generate Choose-Your-Own-Adventures, and so wouldn’t really want to completely disavow fixed branching.
More seriously though, I can’t see the concept of dramatic arcs losing its potential usefulness. I certainly don’t think interactive stories must strictly follow an Aristotelian arc, but I would never disavow that as a structural guideline or as an operationalized technique.
(Please note, I agree that drama can’t be represented by a single number that goes up and down. For example In Facade the drama is really represented by a collection of heterogeneous objects: a large pool of story behaviors (beats, beatgoals and mix-ins) that each have preconditions and effects on shared state; this shared state contains several variables, including an current overall tension value. The tension value is merely one form of communication between beats; it doesn’t serve as a complete model of the the drama. Really, it’s the sum total of these parts (the behaviors themselves, and the shared state they are modifying) that comprises our representation of the drama. What this means is, it’s hard to distill the structure of Facade out into a single technique. We’ve attempted to describe the various mechanisms we’ve employed in this paper (pdf). Each mechanism is relatively simple, such as representing story tension as a single value, but each is just one part of a larger machine.)
Anyhow, back to the larger point about manifestos and Dogme’s. I suppose I’m not as enthusiastic about this idea as I might have been a few years ago, because at this moment, for myself, I feel I’ve written these thoughts out and expressed enough of opinions on the topic (hopefully at least to the mild amusement of a few patient readers). Right now I’m happy to just continue building new systems, and enjoy reading the manifestos of others.
But if I must, I’d contribute this:
I will not make a linear story that contains ironic statements and meta-commentary about how such a story lacks agency.