Past attempts to describe the behaviour of crowds involved treating them as fluid bodies and modelling them as such using, for example, Maxwell-Boltzmann distributions or the Navier-Stokes equation. Describing the behaviour of individual agents however, is more intuitive than describing that of a crowd; this is known as microscopic, as opposed to macroscopic, modelling and can provide a better understanding of naturally occuring phenomena.
By defining how individual agents interact with those around them and assigning targets, complex, and often realistic, behaviours are said to emerge. When two groups of people move towards each other, individuals naturally follow those in front of them to avoid bumping into others. Steady lanes form in the flow of people and their movement becomes ordered.
Local interactions between agents and their environment, whilst unknown to them, result in a global, self-organising system. This kind of artificial intelligence resembles real conditions observed at street crossings and other high density situations.
Modifying or adding new behaviour to agents is an easy task with immediate, visible results on the behaviour of the system.