The Emergence of Social Organization in the Prisoner's Dilemma: How Context-Preservation and Other Factors Promote Cooperation

While complex adaptive systems (CAS) theories focus primarily on phenomena such as systemic robustness against perturbation, self-organization, and on the emergence, transformation, and dissolution of organizational entities or action patterns, the metaphorical resonance of CAS work is not easily translated into careful scientific results. It can be very difficult to identify the right level at which to develop more precise theoretical generalizations with well-specified domains of applicability. And constructing experimental parameters that cleanly map to important, general constructs is usually not a simple exercise. This paper demonstrates an approach to this problem. We report results of agent-based simulation experiments in which the basic activity of the agents is to play the Iterated Prisoner's Dilemma with other agents. We systematically investigate how the emergence and maintenance of cooperation is affected by variations in three key dimensions: (1) strategy space from which the agents' strategies are selected, (2) the interaction processes that channel agents into interactions, and (3) the adaptive processes that govern the changes in agents' strategies over time. Overall, our experiments both confirm results which have been reported in the literature (e.g., that embedding agents in a two-dimensional space can lead to the emergence of cooperation), and our results demonstrate surprising results (e.g., that high levels of cooperation can arise even when agents are randomly mixing, when the agents use simple deterministic strategies and update them using a kind of evolutionary algorithm). Our results also support a generalized view of "neighborhood" where the important factor is the degree to which the interaction processes lead to context preservation, independent of any particular topology. The preservation of context, even as agents are changing their strategies, acts as a "shadow of the adaptive future," resulting in sets of agents who are highly cooperative and resistant to invasion by cheaters.