Inequity aversion and the evolution of cooperation

Evolution of cooperation is a widely studied problem in biology, social science, economics, and artificial intelligence. Most of the existing approaches that explain cooperation rely on some notion of direct or indirect reciprocity. These reciprocity based models assume agents recognize their partner and know their previous interactions, which requires advanced cognitive abilities. In this paper we are interested in developing a model that produces cooperation without requiring any explicit memory of previous game plays. Our model is based on the notion of, a concept introduced within behavioral economics, whereby individuals care about payoff equality in outcomes. Here we explore the effect of using income inequality to guide partner selection and interaction. We study our model by considering both the well-mixed and the spatially structured population and present the conditions under which cooperation becomes dominant. Our results support the hypothesis that inequity aversion promotes cooperative relationship among nonkin.

💡 Research Summary

The paper tackles the classic problem of the evolution of cooperation without invoking the usual assumptions of direct or indirect reciprocity, which require agents to recognize partners and remember past interactions—a cognitive load that many biological and artificial systems cannot realistically support. Instead, the authors build a model grounded in the behavioral‑economics concept of inequity aversion: individuals experience disutility when their payoff differs from that of others, and the strength of this aversion is captured by a continuous sensitivity parameter λ.

In the model, a fixed population of N agents each adopts a pure strategy—either always cooperate (C) or always defect (D). Agents also possess a λ value that determines how strongly they avoid partners whose accumulated payoff r differs from their own. When an agent i seeks a partner, it samples a subset N_s of the population (the “search space”) and computes the acceptance probability φ(i,j)=exp(‑λ_i|r_i‑r_j|) for each candidate j. Interaction occurs only if both agents accept, yielding a joint probability P(i,j)=exp