Simulating Cooperative Behaviors in Dynamic Networks

Abstract

Agent-based social simulation uses agent systems to study social behaviors and phenomena. A difficulty in producing social simulations lies in the problem of modeling the emergence of social norms. Although empirical evidence has provided insight into how human relationships are organized, the way in which those relationships are used to produce cooperative behavior where each agent only seeks to maximize its own utility is not well defined. This paper proposes a new rule called the Highest Rewarding Neighborhood HRN for social interactions. The HRN rule allows agents to remain selfish and be able to break relationships in order to maximize their utility. Our experiment shows that when agents are able to break unrewarding relationships that a Pareto-optimum strategy arises as the social norm. In addition, the authors conclude the rate and amount of Pareto-optimum strategy that arises is dependent on the network structure when the networks are dynamic, and the rate is independent of the network structure when the networks are static.