I build computer models. The name of his talk is "Models come alive...and generate bogus data." I don't care what models do, as long as they do something interesting.
There are models in the PC and models in your head -- your imagination is a critical co-processor.
Science is about observing reality and making a model. In games you do the reverse -- take a model and generate an elaborate world.
In games today, players are getting really good at sniffing out the size of the possibility-space -- in five minutes of play, I can tell you how linear a game is. To get the complexity players demand, you need algorithms -- emergence.
We used to model with calculus -- an equation that tells you where jupiter should be. In simulations, you use little dumb automata that create stories.
Games have topologies -- Myst is linear, chess is branching, Doom is branching but collapses at the end of every level.
Dynamics are how topologies change over time -- growth (size or number), networks (nodes or connection), propagation (diffusion through networks).
Some info propagates out of band (software updates). Some propagates secretly (cheat codes). The propagation curve for cheats is really interesting: hardcore players discover and discard the info quickly, while casual players cherish it and propagate is more slowly.
In online games players randomly meet each other. But if online worlds had a mechanism to automatically introduce new players to other players they'll likely enjoy, they'll get a better experience. Using analysis of who hangs out with whom in gamespace gives us the data to make this happen.
Paradigms -- Theories aren't reality: quantum mechanics and relativity are mutually incompatible. One is good at explaining a macro scale, the other is good for a micro scale, but neither describes a duck (laughter).
Paradigms reflect the times -- futurism suggests atomic cars in the era of the A-Bomb.
System dynamics spun out of the idea of daisy chaining simple electronic components to make complex devices, from oil refineries and other paradigmatic systems of the time. It was used to model everything.
The predictions of system dynamics didn't account for chaos theory and so small errors led to bad conclusions -- system dynamics predicted global collapse in 1987.
After system dynamics came cellular automata. I've spent a good chunk of my life studying these. This is the basis of SimCity -- 3D cellular automata modelling crime, pollution, etc. This is married to system dynamics used to set macro-policies.
Chaos theory says that modelling is limited. We can't predict with any accuracy the outcomes of complex, chaotic systems. Chaotic systems are poised between order and disorder. It set out to be a general theory of complexity, but didn't generate very good tools for understanding complexity.
The systems that are interested are complex, adaptive systems. They don't just have inputs and outputs -- they've got mechanisms for adapting and learning between input and output.
Adaptive landscapes came out of evolutionary theory, which viewed evolution as a hill-climbing alogirthm.
The Sims and other games are designed as hill-climbing landscapes where the most success is achieved by picking a middle between Material and Social axes. We're mapping how people play the game, how they seek out success.
Network theory is the flavor-du-jour. Once you have an average of one connection per node, you'll have a completely connected system. The WWW has a graph width of about 19 -- humans on Earth have about 5.7 (6 degrees of separation).
The world is made of a scale-free network, were a few highly connected hubs reduce the distance between the nodes dramatically -- this is where a power-law distribution comes in. A few highly linked pages link everything. Food-webs are scale-free, so are social systems. So are terrorist networks.