Computational framework for modeling the dynamic evolution of large-scale multi-agent organizations

A multi-agent system model of the origins of an archaic state is developed. Agent interaction is mediated by a collection of rules. The rules are mined from a related large-scale data base using two different techniques. One technique uses decision trees while the other uses rough sets. The latter was used since the data collection techniques were associated with a certain degree of uncertainty. The generation of rough set was guided by genetic algorithms. Since the rules mediate agent interaction, the simplest rule set with the fewer rules and with fewer conditional to check will make scaling up the simulation easier to do. The results suggest that explicitly dealing with uncertainty in rule formation can produce simpler rules than ignoring that uncertainty.