Joan E. Roughgarden, Stanford University
A two-tier approach for modeling the development and evolution of social behavior is introduced. Previous work focused on how a social system emerges in developmental time from equations based on cooperative game theory. The new work presented here shows how cooperative games can be embedded in a between-generation evolutionary genetic model. The approach envisions cooperative behavioral games as nested within non-cooperative evolutionary games. The principal result is that the social system as a whole evolves as an ESS irrespective of whether individual behavioral strategies are an ESS. This result avoids the need to assume, contrary to fact, that specific behaviors, such as acting as a hawk or dove, are genetically determined. Instead, the payoff regime associated with different actions is what is genetically determined, and reflects basic trade-offs associated with morphology. The social system then emerges from how animals optimally play with one another given their morphologies as they live through their life cycle. Specifically, the entries in the payoff matrix are the fitness accumulation rates, and animals play to maximize their instantaneous fitness accumulation rates. The state variables that develop through this repeated playing are time allocations to the various strategies in the game. Then the fitness accumulated by players at the end of the generation are tabulated and entered as selection coefficients associated with genotypes specifying various payoff matrices and tactics of play. The between-generation time scale then leads to the payoff matrix itself evolving, culminating in an evolutionarily stable payoff matrix. The newly derived equations to carry out this modeling program will be presented and illustrated with examples. This modeling is a component of the social selection project to develop alternatives to sexual-selection theory.