An Agent-Based Modeling Approach to Iterated Prisoner's Dilemma on Networks

Abstract

Prisoner’s dilemma is a framework in game theory to explain why rational individuals would choose to not cooperate even at their own expense. In 1984, Axelrod’s computer tournament and its simple yet profound results brought iterated prisoner’s dilemma (IPD) to the view of the public and a myriad of research domains. Sociology, political science, economics, anthropology, evolutionary biology and computer science all used this framework to explore the cooperative behavior in their discipline and continued pushing forward the study of IPD to this day. One particular thriving branch is evolutionary dynamics, where the interaction and adaptation between multiple strategies are simulated to illustrate their robustness.

Nevertheless, the majority of studies done assumed a complete graph or grid-based topology for the strategies to inhabit. Admittedly, they have their own merits in revealing general patterns, yet neither is suitable to model actual social structures especially those based on the Internet, where spatial contiguity does not entail interaction. This study examines the evolution of strategies on different network models, including scale-free network, small-world network, and social circles. Due to their characteristic topology, the evolution on these networks exhibits distinct features comparing to existing literature. This research develops an agent-based model to generate different networks and simulate the evolution. The research proposes future studies and the model developed serves as a general-purpose model open for inquiries in the field of IPD.