November 9, 2004

Crawford Tells All

by Andrew Stern · , 9:11 am

Chris Crawford’s long awaited book on interactive story, in which he finally reveals his secrets and hard-won lessons learned from slaving for years to build the Erasmatron, is now out and available from New Riders. (For more on the Erasmatron, go here and scroll down for essays.)

Chris considers it “the most important book I have ever written”, and describes his goals for the book on the back-cover blurb:

[He] resolves misleading dilemmas, such as the feckless debate over plot versus interactivity, and then offers detailed descriptions of technologies for implementing interactive storytelling. Herein lies the meat of the book. Instead of vague, hand-waving wishlists, Crawford gives workable solutions. Instead of intellectually pretentious gobbledygook, Crawford explains in plain English what works and what doesn’t. This is the real thing.

I have quickly read a pre-release version of the book, and can say it goes further than any other book to date on the true requirements for actually building more deeply interactive, generative stories — not just design goals, but specific paths to solutions. I plan to write up a thorough review once I find some free time.

(n.b. I’m reminded of my disdain for the oxymoronic term “interactive storytelling” — which Chris himself agrees is a poor term…)

10 Responses to “Crawford Tells All”


  1. Erik C Says:

    andrew, your disdain comment is an amazing thread, you should warn how many threads it refers to!

    A few suggestions:
    in a game or virtual heritage environment (which is what I research) agency is not actually the component but _perceived_ agency. You can keep calling it agency but really the fact it is an illusion (but the 2d squirrel really _does_ like me!) does not mean it is not important, for we as pattern matchers assume agency to anything we can, God, inkblots, good looking cars, images of newsreaders.
    Once I sat in the front row of a cinema and this guy (drugs drinks or some other condition) started having a conversation with characters on the screen-he honestly thought they were real. Fascinating, well to me not the others in the cinema.

    Dialogue-it does not mean assymmetry, it means in ancient Greek parlance two people trying to reach a conclusion (Socrates via Plato). Gadamer wrote a wonderful definition of democracy/dialogue: entering a conversation with the tacit acknowledgement one could have one’s mind changed.
    Anything less is not a true dialogue.

    Story telling: story making? story sharing? story playing? It depends on the audience. The Kalevala is a story remembered, a play is a story interpreted and the nuances affected, theatre games are stories directed by the audience. All are slowly remade over time (nobody notices we have modified/semi-standardised Shakespeare’s spelling when he himself/themselves kept changing it).
    Making stories and (spatial narrative in 3D games) is more fun because we are continually surprised by possibly meaningful forks in the narrative path and intrigued by its possible psychological effect on the audience/gamer:
    Immanuel Kant nearly stumbled on this with his talk of aesthetic ideas, and Pynchon had an idea in the crying of lot 49.

    Sharing meaning: when you can produce this in a virtual environment (game) I think it is hermeneutic. So far we have it in 2D (a href=”http://www.fragment.nl/mirror/various/Towell_et_al.1997.Presence_in_MUDs.htm”>MUDS etc) but I don’t think we have it in 3D due to lack of personalisation or truly unique identification (ok so I can build a house in ActiveWorlds that looks like any other house in ActiveWorlds).

    Sorry regards length, the thread re-awoke what I should be writing about, and reminds me what I should have said: Chris Crawford really came close a long time ago to the idea of interactivity so if his new book is even better it is a must-read IMHO.
    s

  2. andrew Says:

    andrew, your disdain comment is an amazing thread, you should warn how many threads it refers to!

    well, our posts and comments tend to link back to older posts and comments, which link further… it’s a tangled web we’re weaving on GTxA I suppose.

    in a game or virtual heritage environment (which is what Iresearch) agency is not actually the component but _perceived_ agency. You can keep calling it agency but really the fact it is an illusion (but the 2d squirrel really _does_ like me!) does not mean it is not important, for we as pattern matchers assume agency to anything we can, God, inkblots, good looking cars, images of newsreaders.

    Interactive art/entertainment experiences that purport to give players agency almost always offer a blend of real agency and the illusion of agency. By real agency, I mean the actions that the player takes truly alter the internal state of the simulated world, and that change is clearly expressed back to the player. That is, real effects, short-term (local) and/or long-term (global). By the illusion of agency, we mean when people “read in” effects they’re having on the world, either purely serendipitously, or because the system gives enough (fake) cues that people are fooled — the Eliza effect. As you say, this happens a lot in people’s lives in general.

    But Erik, faking it can only go so far. Interactive art/entertainment will stagnate in its current premature stage forever unless we make progress to build much more real agency into the experiences. This is at the core of what Chris and other advocates of agency, including me, are trying to say.

    I’m leaning towards getting more purist about this all — building experiences that rely less and less on the Eliza effect, and are more truly generative, offering more real agency.

    This issue parallels the ludology vs. narratology debate (a dangerous link!) — in pure ludology, everything is fully simulated, which should allow for the most player agency. When narrative is represented shallowly (e.g. as canned text, images, videos, etc.) and not generated from a live simulation, it diminishes player agency. The more of an interactive experience that is shallowly represented and not simulated, the less agency that is offered. (It’s interesting, I don’t recall proponents of ludology arguing in these pro-agency terms exactly, but I think it’s an effective angle to take.)

    Note this is by no means a suggestion to abandon narrative! Instead, it’s a push to make simulations that generate narrative.

    This also doesn’t mean solving Artificial Intelligence, or taking a strictly cognitive science approach to simulating human minds! It does however mean building systems that are heavily process-oriented and have rich knowledge representation. They will be minds, but not necessarily mimicking human minds.

    Chris Crawford really came close a long time ago to the idea of interactivity so if his new book is even better it is a must-read IMHO.

    Yes, I think his new book is a must-read, however it gets more focused and technical than his previous ones, nor is it as well-written. It’s probably not going to be as generally useful for people as his previous books. But it’s his most ambitious. At a minimum, it should illuminate for people the issues involved in creating simulations that generate narrative, and help people realize there is no silver bullet to cracking this nut. Interactive story is truly a Big Hard Problem.

  3. Erik C Says:

    Thank you. If I may ask, (for it seems you want to go all the way with Interactivity) do your scripted characters (I am sure you have a better term) have a previous history? Do they have tendencies, baggage etc? For real life interaction is never 100%. I also wonder, can they forget or mishear words? I like imperfections ;)
    Overall, are you saying that scripted characters cannot just afford believable story elements-they have to actively create it? A

  4. andrew Says:

    Erik, yes, I think the goals of many of those creating interactive fiction and drama do ultimately include creating virtual characters that have their own personal backstory and knowledge, their own particular tendencies, baggage, imperfections — all those things that go into creating fully fleshed-out fictional characters. (Of course, any decent interactive experience with characters, whether computer game, text-based IF, what have you, will have fleshed-out characters with backstory, but that content is typically “baked in” — represented shallowly in large, coarse pieces of content — preventing the system from being able to reason about it, to generate variations and extensions of it, to offer much agency with respect to it. That’s the problem.)

    Chris talks in detail about these things in his book. Also, here’s a (slightly stale) list of links to various activity in advancing the state of the art for interactive characters and story, in both academia and industry.

    Speaking of imperfections, this reminds me to link this post and discussion from last May, that goes further to describe what I’m talking about here, that suggests virtual characters should have irrational wants.

    In our almost-completed interactive drama project (we’re in a final debugging and polishing phase now), we have implemented some of these ideas, at least in simple ways. The characters have a lot of backstory and baggage, they have tendencies, they certainly mishear words!, both from accidental technical limitations and from author-intended slanted interpretations of what the player is saying and doing. To get more info about Facade I encourage you to check out the papers listed on the website, particularly the GDC 03 paper. There’s much, much further to go in this direction, of course; Facade achieves barely more than a minimal level of the important personality features we’re talking about here.

  5. Marie-Laure Says:

    I have read Chris Crawford’s book very attentively, and it leaves me very puzzled as to what he means with interactive story. What I found in the book were some advices for constructing a story-generating program, something in the style of Meehan’s Talespin of Bringsjord’s Brutus, maybe with more generative power (though the book does not give a single example of a story produced by the system). But what, I’d like to ask, makes the stories that come out of the Erasmatron INTERACTIVE ?

  6. michael Says:

    Marie-Laure, I’m actually using Crawford’s book in a class I’m teaching right now on Interactive Narrative. I have a partially finished review of his book that I plan on finishing this week and posting to the blog. But the short answer to what I like about his book is the focus on procedural narrative – that, in order to create deeply interactive narrative, you actually need to procedurally model the narrative (characters, plot structure, etc.) in order to generate responses to interaction, that playback of pre-written narrative “chunks” (playback of assets) isn’t enough. The position generates useful discussion both for the game-centric (often technical) students who want to “add narrative” to existing game genres, and for the humanities students who want to “interactivize” traditional narrative forms (or call everything “narrative”).

  7. Marie-Laure Says:

    Michael,
    I can see that deeply interactive narratives (i.e. stories generated in real time rather than from canned data) will require of the computer to respond in narratively coherent ways to the user’s actions, and maybe Crawford’s view is that beore we can do that we must teach the computer to think narratively. So in this sense the programming strategies described in his book could be viewed as a prerequisite for the creation of interactive narratives. Still, I think that the Erasmatron, when completed, would be a storytelling machine, not an engine for interactive stories.
    Looking forward to your review. Are you familiar with Andrew Glassner’s Interactive Storytelling ? It’s my favorite of the 4 books recently published on the topic. (The others are: Carolyn Handler Miller’s Digital Storytelling, and Mark Meadows’ Pause and Effect.)

  8. michael Says:

    I agree that it is unclear whether the Erasmatron itself is an effective approach for creating interative stories. A while ago I played the King Arthur demo and Shattertown Sky and didn’t find either particularly compelling. Part of this for me is the discrete event model – the player takes action by selecting discrete choices from relatively small menus, making it feel to me as if I can have little effect on the action. But one could imagine interposing an abstraction layer between the presentation layer (what the player actually experiences and the continuous actions they take) and the Erasmatron, similar to the natural language understanding system (which includes physical pattern recognizers that map physical actions into discourse acts) in Facade.

  9. andrew Says:

    Cory Ondrejka, one of the creators of Second Life and a Terra Nova contributor (and who a while ago wrote an excellent essay I’m planning on reacting to in a blog post), has reviewed Crawford’s new book.

  10. joris Says:

    I wrote an “artsie” review of Crawford’s book. I liked the book but find it lacks a certain something. The concept of mathematical representation is very appealing but it falls short when needed to explain an artistic dimension of representation. If anybody is interested you can find it here.

Powered by WordPress