Interactions, Model Mechanisms and Behavioral Attractors in Complex Social Systems

In social modeling, a computational environment runs a model that represents the world. The states the model explores (its behavioral attractor) are typically fewer than its description suggests. The mapping between model and attractor depends not only on its parameters (exploring variants of the world) and its conventions (imposed by the computing environment), but also its mechanisms (components of the model representing selected dimensions of the world). This paper equates mechanisms with sets of coupled interaction classes, thus connecting the relative richness of possible choices of agent behaviors to the size of the state space sampled by computational procedures. We illustrate the impact of different mechanisms on the attractor with a specific simulation platform, SCAMP. In our case, in general, the more mechanisms one implements, the smaller the attractor, but with unexpected twists. We discuss the implications of the richness of the corresponding repertoire of interactions available to agents during simulation for the apparent combinatorial explosion of future possible states in agent collectives. We finally observe how some of these twists appear to correspond with the existence of constraints, hinting at underlying conservation laws in silico and ideally in real systems these intend to portray.