Hierarchy Formation in Agent Based Models


    The emergence and dynamic evolution of hierarchies of various kinds are not adequately described by existing models.  One of the promises of computational social science is that it could provide a means through which such phenomena can be better understood.  In this talk I will provide examples of models in which such emergence can be observed and describe a framework for using computational experiments to build a theoretical understanding of these phenomena.  That is, the goal of this research is not just interesting models, but a rigorous means for making inferences about social systems using such models.
There are various sorts of social hierarchies that are potentially of interest, including those based on economic relationships, personal bonds, or communication.  But most central to the future of complexity science is a better understanding of the formation of adaptive hierarchies.  Examples of this phenomenon appear in biology in connection with the appearance of eukaryotes, multi-celled animals, and in speciation.  Similar questions of how adaptive groups emerge from the action of individuals occur in social science across a range of scales, from small tribal groupings to the formation of nation states.
    In this talk I will exhibit simple models that exhibit hierarchy formation.  While interesting in their own right, the focus of this talk is demonstrating an approach to creating social science theory through modeling experiments.


Steve Bankes
Evolving Logic
RAND Graduate School
www.evolvinglogic.com
bankes@evolvinglogic.com