Population fluctuation promotes cooperation in networks

TL;DR

This paper demonstrates that population fluctuations in dynamic networks significantly enhance the robustness and emergence of cooperation, even from non-cooperative initial states, without requiring complex agent cognition.

Contribution

It introduces a fluctuating population model that promotes cooperation more effectively than previous models, broadening understanding of cooperation evolution in dynamic networks.

Findings

Fluctuating populations support cooperation across various initial conditions.

The model enables cooperation to emerge from non-cooperative networks.

Cooperation is more robust to different strategies and temptations.

Abstract

We consider the problem of explaining the emergence and evolution of cooperation in dynamic network-structured populations. Building on seminal work by Poncela et al, which shows how cooperation (in one-shot prisoner's dilemma) is supported in growing populations by an evolutionary preferential attachment (EPA) model, we investigate the effect of fluctuations in the population size. We find that the fluctuating model is more robust than Poncela et al's in that cooperation flourishes for a wide variety of initial conditions. In terms of both the temptation to defect, and the types of strategies present in the founder network, the fluctuating population is found to lead more securely to cooperation. Further, we find that this model will also support the emergence of cooperation from pre-existing non-cooperative random networks. This model, like Poncela et al's, does not require agents to have memory, recognition of other agents, or other cognitive abilities, and so may suggest a more general explanation of the emergence of cooperation in early evolutionary transitions, than mechanisms such as kin selection, direct and indirect reciprocity.