Cooperation is essential for human societies, and indirect reciprocity—where people help others to build a good reputation—is a key mechanism supporting it. But how do the social norms that govern reputation actually evolve when many norms compete? While much research has focused on the stability of norms, much less is known about their evolutionary dynamics in the early stages.

In this talk, I present a computational study addressing this question. Using large-scale simulations on a supercomputer, we systematically explore the evolution of thousands of third-order norms, each inducing its own dynamics of reputation and behavior. Our results show that sustaining cooperation is difficult in well-mixed populations but becomes feasible in group-structured populations. A particularly important social norm consistently emerges: it always assigns a good reputation to cooperation, and a bad reputation to defection—except when defection serves as justified punishment.
This research sheds light on how the interplay of social norms, reputation dynamics, and population structure shapes the evolution of cooperation.

 
Reference: Y. Murase, C. Hilbe "Computational evolution of social norms in well-mixed and group-structured populations" PNAS (2024)