Friday, October 15: 3:00pm

Strategic Adaptation in Spatial Evolutionary Games

Steve Scott, PhD Student
Department of Computational Social Science
George Mason University

Evolutionary Game Theory (EGT) has been used in the natural sciences to study population dynamics and in the social sciences for the study of group behaviors. Evolutionary game theory is based on the concept that the proportional distribution of the various types in the population will converge to a long-term steady state based on the greater reproductive success of individuals having higher fitness in the population. Spatial evolutionary game theory uses the classical evolutionary game theory concepts to govern population dynamics, but uses spatial constraints to limit interactions between agents based on their location.

Classical EGT models make certain simplifying assumptions, such as complete random mixing among members of the population, infinite population sizes, and no mutation. This study relaxes these assumptions and investigates a combination of two extensions to classical EGT – the use of spatially explicit representation to localize agent interaction, and several methods for modifying agent strategy preferences over time. The model considers the effect of synchronous versus asynchronous strategy updating, the use of pure strategy versus mixed strategy profiles, the effect of neighborhood size on strategic adaptation, and the effect of local versus neighborhood influences on strategic adaptation in a Particle Swarm Optimization (PSO) environment.

A spatial agent-based model is used in which a set of stationary agents interact with nearby agents in a repeated-play Prisoner’s Dilemma game in order to explore the tension between cooperative and competitive strategy adaptation over time. Agents play the game with their neighbors, and decide whether to pursue a cooperative or competitive strategy based on a comparison of their own payoff with the payoffs of their neighbors, adopting the strategy that yields the best outcome. The population strategy preferences change over time as the strategy preferences of many individual neighborhoods emerge.

The modeling results indicate that the population strategy preference between cooperation and competition in populations using mixed strategies is markedly different from populations using only pure strategies. In addition, the results show that population preferences for cooperation and competition are sensitive to neighborhood size. Finally, when comparing the effects of local influence versus neighborhood influence in a PSO environment, the effect of the neighborhood strategy choice appears to be much more significant than the effect of the local strategy choice or the interaction between the local and neighborhood effects. The results are of interest in the study of the strategic adaptation of population preferences over time, especially in spatially constrained environments.