Chris Hecker (EA) - Structure vs Style

Chris Hecker was asked to give one of the keynote addresses at MIGS, which he took as an opportunity to put forth a new idea he was formulating regarding the solving of hard problems – problems that arise at the intersection of interactivity, technology, and aesthetics. Generally the best way, he finds, to solve these problems is to throw at it both human style and computer structure in a way that brings out the best of both worlds, mitigating their weaknesses. Chris feels that this is something that doesn’t happen as much as we might like in game development. So while he still doesn’t know exactly how to achieve this in practice, he does have thoughts towards getting there, which is what he shared with the audience at MIGS. He was quick to make a disclaimer regarding the content of the keynote which, as opposed to a lecture he gave yesterday, really had no concrete topic, fancy demos or a provided solution. It’s still a work in progress and he plans to give the updated 2.0 version at GDC 08.

Chris’ main example for the intersection of structure versus style comes from what he considers to be the technology that has had the most profound impact on games – the texture-mapped triangle. The powerful structure of the triangle allows the computer to reason about it at a very deep level, looking at its morphology (transforming, rasterizeing, lighting, hit testing) and its topology (traversal). On the other hand, the expressive style of the texture map lets the human artist represent subtle yet rich designs through the minimal shape and linear form and the intuitive degrees of freedom (vertices, texture pixels). Put together, “these two things are far more powerful than either one separately” and you get the beautiful 3D game worlds that have so greatly advanced our industry in the recent years, evoking emotion from a virtual world that 2D cannot accomplish.

Other examples of the structure versus style breakdown included skinned mesh and bones, where the bones formed the structure that the computer could read, and the skin atop those bones could be expressed in any way by the human artist yet an arm would still remain an arm to the computer, regardless of whether it looked like a human arm or the scale-covered arm of some reptile creature. This decomposition, posits Chris, is everywhere hard interactive problems are being solved. As mentioned before, these are the problems at the intersection of technology, and aesthetics and interactivity, which is the key point regarding games. Movies, as another form of entertainment, can also benefit from structure versus style breakdowns, but they don’t have to. Games being interactive rely on this a lot more heavily.

How much so? Chris theorizes that solutions to hard interactive problems will always have a deep structure versus style decomposition. In order to enable interactivity, the computer must be “in the loop” and in order to enable emotion and aesthetics, a human must be “in the loop”. This being a topic Chris has been thinking about for a long time, he’s come up with other decompositions as well before settling on structure versus style. Others were synthesis vs. sampling, code vs. content, simulation vs. emulation, and interactivity vs. verisimilitude. You’ll note a consistent computer vs. human pattern amongst the decompositions as they are the two key elements involved.

But how best do we apply this new way of thinking to game design? What’s one problem that still remains unsolved from this perspective? The answer comes from another question that Chris asked the audience which was “what technology should have had the most profound impact on games?” In response, the red eye of HAL from 2001: A Space Odyssey slowly appeared on the screen. "I think AI is the key to making games the pre-eminent artform of the 21st century," the only problem however is that no one has yet found a structure versus style decomposition for AI. Chris has some thoughts on the issue, the basic of which is figuring out what a “sample” of Ai is. What are simple, explicit and pseudo-static? Can we define a single script of code a small piece of AI? How about a finite state machine? FSMs seem plausible, but then what’s our blend operator? How do we combine it with the style of human expression to create deep interactivity, consequence, and meaning?

So many questions and so few answers. Chris came up with a few ideas, none of which felt “right” to him. Behavior trees, a model used with FSMs to create complex behaviors from simple state machines joined together was one idea where the tree “dithers” between samples, each sample being a specific action like running, walking or shooting. Blending narrative into interactive stories was another idea, as well as the method used in The Sims to drive sims behavior by objects advertising their properties to the sims’ character AI. The most difficult problem is that AI is completely code-driven, there’s really nothing you can draw of AI for example, so what then is the “style” of code? How do you blend code?

There’s a lot about Chris’ talk that I didn’t cover here, I will admit. It’s a new idea that he’s had a chance to speak about for the first time, but which he will address once more in a few months at GDC 08. Rest assured we’ll be there to deliver more information on this stimulating idea.

Page 6