An agent-based model for evolution of cooperation with proactive information gathering

This thesis proposes a new model to investigate the impact of proactive information gathering upon the evolution of cooperation among self-interested agents in a multiagent system. It builds upon an existing game-theoretical model of spatially distributed mobile agent population with energy-based individual life cycle, in which individuals keep playing one-shot Prisoner's Dilemma games in neighbourhood encounters. Using proactively gathered information about past behaviour of others, advanced agents in the new model can dynamically adjust their strategies towards different types of opponents. Simulation experiments establish some patterns of how this ability impacts the evolution of cooperation in the presence of varying levels of environmental adversity. The adequacy of the model is demonstrated through a specific design involving two types of advanced agents, one oriented towards cooperation, the other towards defection. The results show that cooperation prevails in a substantially larger area of parameter space than in the basic model without information gathering.