If we are to understand the interactions of a large number of agents, we must first be able to describe the capabilities of individual agents.