Nature 456, 579 (4 December 2008) | doi:10.1038/456579a; Published online 3 December 2008
Generosity: A winner's advice
Martin A. Nowak1
- Martin A. Nowak is professor of mathematics and biology at Harvard University and director of the Program for Evolutionary Dynamics, 1 Brattle Square, Cambridge, Massachusetts 02138, USA. He is the author of Evolutionary Dynamics.
One day while I was still at Oxford, Bob May gave me some advice: "You never lose for being too generous". I was impressed because Bob is a winner. To him winning a game is everything. He has thought more deeply about winning and losing than anyone else I know. As his wife once said, "When he plays with the dog, he plays to win." At the time, Bob was not only my adviser but also one to the British government. A few years later he would become president of the Royal Society, Lord May of Oxford and the recipient of many prestigious awards.
A mathematical analysis of human behaviour suggests that Bob was right. Generosity is an essential feature of winning strategies in games that explore human interactions. These strategies underpin many of the choices people make in everyday life, and shed light on how our unusually cooperative ways have evolved.
Biologists recognize two fundamental forces of evolution: mutation and selection. I want to add a third: cooperation. Cooperation occurs when one individual pays a cost so that another receives a benefit. Here, cost and benefit are measured in terms of reproductive success. Reproduction can be genetic or cultural, the latter involving the spread of knowledge and ideas.
Only if certain mechanisms are involved can natural selection favour individuals who reduce their own fitness to increase that of a competitor. One such mechanism is direct reciprocity: my strategy depends on what you have done to me. Another is indirect reciprocity: my strategy depends on what you have done to me and on what you have done to others.
In both, mathematical analysis shows that winning strategies tend to be generous, hopeful and forgiving. Generous here means not seeking to get more than one's opponent; hopeful means cooperating in the first move or in the absence of information; and forgiving means attempting to re-establish cooperation after an accidental defection. These three traits are related. If I am generous, it is easier for me to forgive, and also to be hopeful and take the risk of cooperating with newcomers. (Italics added by Tom.)
In the Wimbledon championship, you must defeat your opponent to move to the next round. But everyday life is not like a tennis tournament. Instead, most of our interactions occur in a population of players, and pay-off accumulates over encounters with many different people. Because overall success is proportional to that pay-off sum, the other person in any one encounter is more a partner than an opponent. If I am willing to let others have a slightly bigger share of the pie, then people will want to share pies with me. Generosity bakes successful deals.
Experiments have confirmed the success of generosity. A typical set-up involves students and computer screens. The computer pairs random individuals. One person, the donor, is asked if she wishes to transfer some money to the recipient. She is informed about the recipient's decisions in previous rounds with other players. The experiment shows that people base their decision on what the recipient has done before. Generous people are more likely to receive donations.
Similar reputation-based systems operate in e-commerce. When buying a camera online, you might consider both the price and the seller's reputation. Consumers are willing to pay higher prices if the seller is thought to be reliable. Successful websites are those with good reputations.
So why aren't humans always 'generous, hopeful and forgiving'? Part of the explanation may be that cooperation is never a stable state. Mathematical studies show that it is constantly challenged by defection. In a society of defectors where no-one helps, a cluster of cooperating individuals can emerge if, by chance, a few people start playing a direct reciprocity strategy called tit-for-tat: I do whatever you have done to me. Tit-for-tat can't persist for long because its appetite for revenge is self-destructive. It is soon replaced by 'generous tit-for-tat'. Here, I cooperate whenever you have cooperated, but sometimes even when you have defected. In other words, I am forgiving. For a while, cooperation thrives. But in a generous tit-for-tat population, the emergence of unconditional cooperators eventually invites the invasion of defectors. This leads to cycles of cooperation and defection — which could account, at least in part, for the mix of cooperators and defectors that persists in human societies.
Mathematical models allow a precise investigation of fundamental aspects of human behaviour. The games described here occur in every society. Ancestral humans spent most of their time in small groups where interactions were repeated. The same is true for most dealings in modern life: repeat encounters are always possible and reputation is typically at stake. The evolution of prosocial behaviour cannot be understood outside the framework of direct or indirect reciprocity. Indeed, I believe that games of indirect reciprocity have provided the crucial selection pressures for social intelligence and language.
In such games, social intelligence is needed to monitor and interpret the interactions of others. We follow with great interest what our fellow creatures do to us and to others. When deciding how to act, we take into account — often subconsciously — the possible consequences for our own reputation. Moreover, our own observations are often not enough; we want to learn from the experiences of others. Spreading the rumours of indirect reciprocity requires language. As my colleague David Haig once remarked "for direct reciprocity you need a face, for indirect reciprocity you need a name".