Cathode Tan - Games, Media and Geek Stuff
logo design by man bytes blog

Sunday, October 15, 2006

Doc Dobb's Chris Crawford Interview

Dr. Dobb's Portal had a sitdown with Chris Crawford and chatted Storytron with him (via /.). Oddly, even though interactive fiction is one of my follies causes - I honestly don't follow Chris much and nor do I really side with either camp that seems to be for or against his standing in the gamedev circles.

This is partially because I keep drifting away from the concept of a storyworld as the saving grace of interactive literature (or more generically storytelling). But that's not to say that I don't find the idea intriguing and so generally wish well on those going after it - it's just not my cup of tea.

I do think the interviewer, Michael Swaine, gives Crawford a few passes where deeper questions would have been better ... for instance:

DDJ: SWAT is an acronym for...

CC: "Storyworld Authoring Tool." There are three programs. A storybuilder uses SWAT to create a storyworld. The second technology is the [Storytron] Engine, the third technology is Storyteller, which is the consumer program. Both Storyteller and SWAT access the Engine, Storyteller to play it, and SWAT to run Rehearsal, our testing feature.

DDJ: And Diekto?

CC: Diekto is the language.
-- Dr. Dobb's | Interactive Storytelling | September 6, 2006

Now ... one might think that would merit some follow up discussion. Something more than just "it's a language". Like, why would a storyworld need it's own language? How is it different from English ... and so on. Straying away from a natural language parser is one of the more controversial aspects of Storytron and while the interview doesn't need to critique it - it should at least be brought into the conversation ... but the interview never goes there.

Diekto is a "mini language", according to Crawford, which seems designed to state precisely what needs to be illustrated to an NPC. Now, here is Crawford in his own words (or at least his website's):

Deikto may not be the most inspiring language to write sonnets in, but the singular beauty of interactive storytelling is not in its representation - it is in the richness, depth, variety and drama of the interactions it allows.
-- STORYTRON - Interactive Storytelling

Distinctly this is where a construct like Storytron doesn't work for me. I'm not denying the complexity of natural language parser (that's why I removed it completely in Randolph Carter) ... but you can't tell another artist what's beautiful. This is the conflict I'm seeing with Storytron. Repeatedly, it's presumed to be for creating interactive works of art. That same page notes, "there are techniques by which a conversation can be "faked" using a library of prefab sentences, but these are too narrow and cumbersome for artistic purposes".

But enforcing a "mini language" isn't?

This fine line between technical complexity and artistic license is evident again when the interviewer starts talking about how Storytron is made for use by the everyman and while describing that it's not programmatic in nature and resistant to crashing, Crawford goes down this path:

CC: A great many calculations can be skipped over without undue harm. It makes the system more boring in that fewer things can happen, but it guarantees that the system doesn't crash. Oh, another major issue is data typing. The real innovation is a data type we came up with that we call a bipolar number, or B-number. Bipolar numbers are numbers that range from -1 to +1, and for the variable they describe, 0 represents average, +1 the highest conceivable value, and -1 the lowest value. The advantage of this is that it eliminates all scaling issues. When I hit upon this, it took a long time sorting out how to make it work, and it is weird. But it's especially advantageous for clearly subjective quantities like how faithless or honest someone is. That's another thing: Every variable is defined with two words representing the lower extreme and the upper extreme. People get confused thinking about zero honesty.

DDJ: Of course ordinary arithmetic doesn't work with B-numbers.

CC: We have [developed] bounded arithmetic. Mathematically I can prove [the bounded B-number operators] are analogous to their linear counterparts.
-- Dr. Dobb's | Interactive Storytelling | September 6, 2006

For one thing, and I don't want to armchair code anything or whatnot, but Crawford's "bipolar number" just seems like a way of describing a signed integer with a range from -1 to 1. Be that as it may - we're now talking pretty technical stuff here and it's clearly at the core of what goes in within a Storytron storyworld. I'm not sure how this "eliminates scaling issues" so much as "we've defined you scaling issues for you" and hopefully the description of it won't put you to sleep while trying to be artistic.

And it ends on the note that I've harped on before ... it's a singular issue I have with certain storyworld concepts:

CC: Oh yes. It'll be very useful for corporate training, military training, educational stuff. Basically, it's a social interaction simulator. In fact [it might be] better to think of it as a simulator, because the stories it generates are very different from conventional stories. They don't have plots.

DDJ: No plots?

CC: Because of the interactivity. What happens is that the player explores a dramatic universe. A storyworld.
-- Dr. Dobb's | Interactive Storytelling | September 6, 2006

And again - the interviewer simply lets this pass along. I mean, it's a plotless dramatic universe where you speak in a mini language and your emotions are defined by bipolar numbers.

How could this not be art?

Don't get me wrong - I would love for Crawford to succeed at his goals. I think they are very, very worthy goals. I just haven't seen the conversation take place which makes think that this will be an artistic solution. Course, I don't really think you can "solve" art for artists. I think it's a mistake to assume that artists either aren't or can't be technical people. Take Piet Hein as an example ... a contempory of John Nash's who both wrote poety and designed games.

tagged: ,


Anonymous said...

My alternative is to provide an AI toolset for truly distinct characters defined by algorithms rather than data permutations. Then you can build whatever game engine you want, to house those characters in an emergent space.

I do think there's mundo value in playing with SWAT and trying to produce something with it, so I think I'll devote a few hours when the Alpha gets version stability.

I think Crawford's biggest mistake is following a fourteen year waterfall based on one advanced theory, he should've been prototyping instead of building a huge monolithic engine AND dev environment over years and years, all on his own for many. Thats crazy to me, and I understand that insanity, because I have it too, so I empathize with the guy. There's a place for Storytron, but I don't know how big the niche could be. Maybe a million subscribers of five years and a few generations.

Josh said...

I think AI will be big this generation. I honestly could care less about Valve's facial animation technology when the NPC's act like stumps half the time ... and Valve has pretty good AI compared to much of the industry.

I actually hope Storytron finds its niche ... because I'd like to see ideas like this one get to some kind of fruition. So when I critique it - its mostly an explanation of a counter-approach to modes like Facade and Storytron which, admittedly, I'm not done forming yet. Poking at stuff helps to form that.

Anonymous said...

The way I understand it Facade and Storytron are applying this highly modeled systemic approach to generating stories, and you and I (and Craig Perko and Micheal Samyn) are reaching for solutions in an emergent style, something more rooted in principles of game design.

Josh said...

Yeah, that's how I see it - although part of me is also trying to return the process of writing itself to the central part of the framework. But I'm equally stymied by the need to have some kind of mechanic.

I'm not looking for a universal format or framework though - I'll be happy to have something that reads like a story but plays like a game but have the game elements be tailored specifically to the story concept itself.