A multi-agent computational model of the evolution
in children's playgroup formation

    Current social science models are inadequate for predicting playgroup formation in pre-school children.  Most descriptions of this process derive from post-hoc explanations.  Physical, temporal, and psychological factors contributing to the formation of these playgroups are currently unknown but the dynamics of this evolution appear to have similar characteristics as other structures in social organizations (e.g., agent actions appear to be rule-based).  Last year we presented an equation based modeling platform that addressed how young children between 3-6 years of age, brought together in a university child development laboratory, evolve into play partner clusterings over the course of the year.  We have rewritten the model as a multi-threaded ABM, where each child (i.e., agent) possesses rankings on 10 attributes obtained via observational and rating scale methods (e.g., aggression, prosocial, physical attractiveness).  Other factors that we hypothesize to influence inter-agent play propensity is preference for a same sex vs. other sex play partner and recency of play (as a proxy for memory); to represent developmental change, each factor allows modification of its putative influence.  To assess fit between the simulated and realized data, we examine localized differences using cell-by-cell comparisons within the matrix, and more generally, matrix sub-region signal-to-noise values, and entropy estimates for assessing the system dynamics.  We show how the model approximates the emergence of play propensities as function of differences in attribute ranking as influenced by gender, memory, and age among children using data collapsed into 5 periods over the school year.

William A. Griffin, Carol Martin, Richard Fabes, Laura Hanish, Mary Anders, Stacie Leonard, Melissa Herzog


William Griffin
Arizona State University
Dept. of Family and Human Development
http://www.public.asu.edu/~atwag/
william.griffin@asu.edu