Cooperation, Punishment, and the Evolution of Human Institutions

Explaining the scale, diversity, and historical dynamics of human cooperation is increasingly bringing together diverse empirical and theoretical approaches. For decades, this challenge has energized evolutionary and economic researchers to ask: Under what conditions will decision-makers sacrifice their own narrow self-interest to help others? Although classic evolutionary models based on relatedness and reciprocity have explained substantial swaths of the cooperation observed in many species, including our own, theoretical work in the 1980s demonstrated that the puzzle of cooperation in large groups, or in situations without much repeated interaction, remained unsolved and would likely require alterative theoretical formulations.

Such cooperative dilemmas, or "public goods" problems, involve situations in which individuals incur a cost to create a benefit for the group. In our society, think of recycling, buying a hybrid car, valor in combat, voting, and donating blood. The dilemma arises from free-riders who enjoy the group benefits created by the contributions of others without paying the costs. Even if nearly everyone is initially cooperative and contributes, free-riders can profit and proliferate, leading to the eventual collapse of cooperation. So, understanding how public goods problems can be solved has provoked great interest, both because human societies have somehow managed to solve many such problems to varying degrees, and because some of the world's most pressing issues, such as global climate change, are essentially public goods dilemmas. On page 108 of this issue, Gürerk et al. take an important step in understanding how self-sustaining cooperative institutions may have emerged over the course of human history.

Recent models have demonstrated how evolutionary processes (genetic or cultural) can maintain cooperation in large groups or without repeated interaction. Costly signaling models have shown how cooperation by "high-quality individuals" (those who are potentially desirable as allies or mates) can be sustained if such individuals can accurately signal their quality by making substantial cooperative contributions to public goods. For example, great hunters might supply all the meat for a public feast, or millionaires might donate a recreational center to their community. Similarly, reputation-based models have shown how cooperation can be sustained if individuals' reputations for not contributing to public goods reduce their payoffs (or fitness) by altering how others treat them in certain dyadic social interactions. Finally, models that allow individuals to both contribute to the public good and to sanction noncontributors have revealed stable cooperative solutions, especially when the strategies for cooperation and punishment are influenced by social learning. Thus, a number of possible stable solutions to the puzzle of cooperation in large groups, or cooperation without repeated interaction, have now emerged.
It turns out, however, that finding a stable solution is only the first step in confronting the dilemma of cooperation. Each of the above approaches can actually stabilize any behavior or practice, independent of whether it delivers any benefit to anyone. This includes behaviors that reduce the payoff or fitness of the group. For example, instead of public goods contributions, costly signaling could maintain behaviors involving dangerous physical feats (like scaling icy mountain peaks), aggressive displays (like beating up your neighbor), or extravagantly wasteful feasts. Similarly, the same reputational and sanctioning mechanisms that can stabilize cooperation can also sustain maladaptive practices such as consuming the brains of dead relatives, flattening the foreheads of infants, or binding the feet of young girls. Thus, there are actually a multitude of stable equilibria, only some of which are cooperative. What determines which equilibria emerge and/or spread?
Three broad theoretical approaches confront the problem of equilibrium selection. The first, and perhaps the most intuitive, is that rational, forward-looking individuals recognize the long-term payoffs available at stable cooperative equilibria, assume others are similarly sensible, and choose the cooperative state. The second approach is based on the stochasticity inherent in any interaction. Different stable equilibria are more or less susceptible to this stochasticity, meaning that in the long-run, some equilibria will be substantially more common than others. The third mechanism, cultural group selection, gives priority to the competition among social groups who have arrived at different culturally evolved equilibria. This intergroup competition favors the spread of individuals and practices from groups stabilized at more cooperative equilibria. In humans, competition between groups can take the form of warfare, demographic production (some social groups reproduce faster than others), or more subtle forms in which individuals learn decisions and strategies by preferentially observing more successful individuals, many of whom are more successful because they live in groups at stable cooperative equilibria. This can lead to a flow of decisions, strategies, and even preferences from more cooperative groups to less cooperative ones, or to a migration of individuals among groups that favors the spread of the more cooperative equilibria.
Gürerk et al. address the issue of equilibrium selection with an elegant addition to the existing experimental work on public goods. In their experiment, individuals (the "players") choose between two different "institutions." In one institution, players can contribute money to a group project. The sum of all contributions to the project is augmented by a fixed percentage and then is divided equally among all players, regardless of their contributions. Previous experiments established that when this interaction is repeated, mean contributions to the public good drop to near zero (a noncooperative equilibrium). The other "sanctioning" institution is very similar, except that after players have contributed, they can pay to punish (reduce the payoff of) other players. When this interaction is played repeatedly a substantial fraction of players punish low contributors, causing mean contributions to rise and stabilize near full cooperation (a cooperative equilibrium). Both institutions were run concurrently for 30 interactions and players could, initially and after each subsequent interaction (after seeing others' payoffs), choose their institution for the next interaction.
The principal findings of Gürerk et al. can be summarized simply. Initially, most players picked the institution without sanctioning possibilities. But, as usual, free-riders in the nonsanctioning institution started driving mean contributions downward, so cooperators, who hate being exploited by free-riders, started reducing their contributions. Meanwhile, in the sanctioning institution, punishers started driving contributions up by inflicting costs on noncontributors, despite the personal cost of punishing. After a few interactions, players from the nonsanctioning institution--presumably seeing the higher payoffs of those choosing the sanctioning institution--increasingly switched institutions. Notably, despite the incoming flow of migrants from the nonsanctioning institution, the mean contributions in the sanctioning institution consistently increased or held stable near full cooperation. In fact, most incoming migrants, consistent with local norms in their new setting, increased their contributions during their first interaction in the sanctioning institution, and a majority administered some punishment.
What does this tell us about equilibrium selection? First, the players' degree of rationality did not permit them to foresee the final outcome and select the higher payoff institution on the first interaction. Second, despite the stochasticity of human decisions, neither institution drifted to another equilibrium. What did happen is that once players from the lower payoff institution observed the higher payoffs of the other institution, they wanted to adopt either the practices of the higher payoff institution, or the decisions and strategies of those other players. Consistent with ethnographic and historical case studies, the present work provides an important experimental demonstration of cultural group selection in action, as the two alternative equilibria compete for shares of the total population.
The course charted by Gürerk et al. should spur more empirical work on how processes of equilibrium selection influence the evolution of institutional forms. Many questions remain to be tackled: for example, what happens if switching institutions is costly, or if information about the payoffs in the other institution is poor? Or, what happens if individuals cannot migrate between institutions, but instead can vote on adopting alternative institutional modifications? Such work can both help us understand how humans became such a cooperative species, and teach us how to build durable cooperative institutions that solve public goods problems and are readily spread.