Network rewiring promotes cooperation in an aspirational learning model
We analyze a cooperative decision-making model that is based on individual aspiration levels using the framework of a public goods game in static and dynamic networks. Sensitivity to differences in payoff and dynamic aspiration levels modulate individual satisfaction and affects subsequent behavior. The collective outcome of such strategy changes depends on the efficiency with which aspiration levels are updated. Below a threshold learning efficiency, cooperators dominate despite short-term fluctuations in strategy fractions. Categorizing players based on their satisfaction level and the resulting strategy reveal periodic cycling between the different categories. We explain the distinct dynamics in the two phases in terms of differences in the dominant cyclic transitions between different categories of cooperators and defectors. Allowing even a small fraction of nodes to restructure their connections can promote cooperation across almost the entire range of values of learning efficiency. Our work reinforces the usefulness of an internal criterion for strategy updates, together with network restructuring, in ensuring the dominance of altruistic strategies over long time-scales.Maintaining a public resource requires sustained cooperation through contributions by community members who benefit from it. Yet, a selfish individual who refuses to contribute can enjoy the benefits without paying the cost of sustaining the public good. If however, too many members of the community act selfishly, the public resource collapses to the detriment of all. The public goods game highlights such a social dilemma and provides a framework for exploring different mechanisms of strategic decision-making that allow cooperation and consequently the public good to be sustained. Among many mechanisms, the reorganization of social ties has been shown to be effective in promoting cooperation in PGG. However, the efficacy of most mechanisms in sustaining cooperation rely on individuals updating their strategy on the basis of information about the contributions of other members of the community. Often such information is either not forthcoming or cannot be effectively utilized. An alternative low-information model of behavioral updating relies on a comparison between the actual benefit received and the benefit aspired for. Individuals tend to retain their strategy if they are satisfied with the benefit received and change their strategy if they are unsatisfied. We show that such a simple reinforcement learning model along with modest restructuring of social ties over time can allow cooperation to be sustained. Our work shows that a low-information strategy-update model can be very effective in ensuring dominance of cooperators in social dilemmas.