I have long been interested in the question of how cooperation can emerge in a world of egoists without central authority. Over a period of five years, culminating in my book on the Evolution of Cooperation (1984), I published a series of studies that explored the emergence and maintenance of cooperation in the context of the two-person iterated Prisoner's Dilemma (PD). I was well aware that two-person interactions can tell only part of the story of cooperation in societies, so I thought about various ways of building and sustaining cooperation when one person's actions can affect many others. Unfortunately, the most straightforward way to extend the two-person PD game would not sustain cooperation unless something else was added to solve the collective action problem by preventing free riders.

In looking for mechanisms that actually work in societies, I was impressed with the power of social norms to maintain specific types of cooperation. For example, the norm against cheating can be sustained by individuals who punish cheaters whenever they are detected. The problem is that punishment is typically costly to the punisher even if beneficial to the community. In the paradigm of egoists without central authority, no one would want to pay the private costs of punishing a violator, and soon the norm itself would collapse. My article proposed that one way to build and sustain cooperation in such a setting would be for individuals not only to punish the norm violator but also to take vengeance on anyone who did not punish the norm violator. I explored the dynamics of this situation with an agent-based model in which relatively effective strategies would be used more than less effective strategies in the future.

The principal result of this evolutionary approach was that if the level of vengeance was initially high enough, selfish acts (such as cheating) would be suppressed. To be successful you had to punish defectors lest you be punished for not doing so. The simulations of the agent-based model demonstrated how and under what conditions cooperation would emerge in a setting where large numbers of people are affected by each selfish act.

The article also informally discussed social norms in relation to other mechanisms for sustaining cooperation in groups, such as law, power differentials, internalization, social proof, and voluntary membership. For most readers, this informal discussion of how various mechanisms complement each other was probably far more useful than the formal model. For example, the article discussed how a law is likely to be effective when supported by an existing social norm, whereas the existing norm, in turn, is reinforced by the law.

This article proved far more compelling than I expected it to be. For example, it was mentioned (along with my two-person work) in the citations for two awards I won shortly thereafter, a Five-Year McArthur Prize Fellowship and the National Academy of Sciences Prize for Behavioral Research Relevant to the Prevention of Nuclear War.

It is hard to evaluate the influence of this work. The formal modeling of social norms did pick up markedly after this article was published in 1986, but several earlier works also promoted this trend. What is clear is that my terminology of “metanorms” did not catch on, and the concept had to be “reinvented” under other labels such as one of the (many) meanings of “indirect reciprocity.” The influence of the model is perhaps more clear in the growing literature on agent-based models and evolutionary approaches to game theory. I attribute much of whatever influence it did have to my decision to treat norms in behavioral terms, rather than rely on the standard definition of social norms in terms of beliefs and values.

Another aspect of the article's influence was its relevance to a wide range of problems in political science and beyond. Within two years of publication, it was cited not only in political science journals, but also in journals of economics, sociology, law, philosophy, anthropology, and animal behavior.

The article has apparently stood the test of time. Unlike the typical scholarly article, which receives fewer and fewer citations after the first four or five years, the annual citation rate for this article shows no signs of declining after 20 years. The citations in the second decade after publication actually exceeded the citations in the first decade by 30%.

I attribute much of the attention this article has received to its role as an extension of my earlier work on pair-wise cooperation. More generally, both projects benefited from the widespread desire to provide a “hardheaded” rationale for cooperation.