GESSLER, Nick (UCLA)
DIFFERENTIAL COGNITION AND COGNITIVE LOAD IN A MULTIAGENT ARTIFICIAL CULTURE

There are very few comparisons of social simulations which try to simulate the same phenomenon by assessing the efficacy of different representations of parallelism in time and space. Simulating a parallel world of interacting agents inside a serial computer is usually accomplished by two different representations of simultaneity. The first is simple sequential or random turn-taking wherein each agent assesses the state of the stopped world, decides what it wants to do, and acts upon it. The second is a more complex and realistic parallelization wherein the world proceeds as frames stopped on a movie film. Within each frame each agent has an equal opportunity to act and any biases, collisions or conflicts entailed by the simpler method are resolved before the subsequent frame is rendered. As the time represented by each frame is cut shorter the action becomes smoother, not unlike the trajectory by which flicker is reduced by increasing the frame rate which results, ultimately, in slow motion. Disregarding the differences in power between parallel and serial computers, serial computers can be configured to represent any parallel system. Time in the simulated world proceeds, more or less realistically, by frames, but time within each frame remains unmanaged. What happens when we introduce the much more realistic constraints of distributed differential cognition not just among but within each agent? What happens when within each agent's mind lower order cognitive agents compete for influence in sensing, thinking and acting? How might we represent cognitive load inside a simulation without submitting to an infinite regress of clips of film within each frame of film, of agents within other agents, and simulations within other simulations?