The Evolution of Cooperation In N-Person Public Goods Game
Under Different Social Enviornments
Jung-Kyoo Choi
jungk@econs.umass.edu


     We analyze the evolution of cooperation in N-person social dilemmas. Agents are assigned into groups and play a repeated N-person public goods game. Games are played under different settings. Agents either interact with a random group (global) or the same group (local). Furthermore agents either update their strategies based on global scale or on their neighbors’ strategies. Therefore we have four different combinations, each of which is an analogy of different social environment.

    Global Interaction    Local Interaction
    Global Selection    Market    Long-term Relationship(e.g. firms)
    Local Selection    Peer Group Updating    Segregated Small Village

    A genetic algorithm is used to adapt the strategies with a selection mechanism that involves either global or local agents.  With random groupings, defection is the norm with some occasional, short-lived outbreaks of cooperation. These outbreaks are enhanced when selection is local. When agents stay within the same group cooperative outcomes are much more likely, with relatively high levels of sustained cooperation in many of the resulting groups when selection is global. This effect disappears when the selection is local.
    Localizing agents’ interactions is a way to sustain cooperation by generating group benefit, whereas group benefit is only transitory in random grouping. However this benefit can be sustained only when local interaction is paired with global selection. We found that “group selection process” plays an important role in this dynamics. Localized interactions and global selection allow group selection process to come to exist.
     We also found that genetic drift occurs in strategy (gene) space and plays a crucial role. It greatly enlarges the effect of mutation and helps cooperative strategy spreading over the population (even if it is dominated strategy). The role of genetic drift is discussed in detail in this paper.