How did cooperation evolve, and what does the Prisoner's Dilemma show?

Cooperation has been a difficulty for evolutionary theory since Darwin, because helping others can look costly to the helper. The Prisoner's Dilemma is the standard tool for showing why. In a single encounter, defecting pays better than cooperating whatever the partner does, so natural selection should favor defectors. Cooperation can nonetheless arise when the same individuals meet repeatedly and respond to each other's past behavior. This is the idea behind reciprocal altruism. [1] [2]

The game comes from game theory, a branch of mathematics begun by John von Neumann and Oskar Morgenstern. Two players each choose to cooperate or defect. There are four payoffs: Temptation for defecting on a cooperator, Reward for mutual cooperation, Punishment for mutual defection, and the Sucker's payoff for cooperating with a defector. The dilemma is defined by the ranking T > R > P > S. Both players do better if both cooperate, yet each has an incentive to defect. [2] [3]

In a one-shot game, defection is an evolutionarily stable strategy, meaning no other strategy can invade a population of defectors. In a well-mixed population, defectors have a higher expected payoff than cooperators. This conclusion holds only under special conditions: partners never meet again, are unrelated, and cannot choose one another. When those conditions fail, cooperation can evolve. [2] [3]

Trivers (1971) argued that conditional behavior promotes cooperation where individuals can recognize and remember each other. He called it reciprocal altruism. One review notes that it is better described as reciprocal cooperation, because it brings a direct fitness benefit and is mutually beneficial rather than truly altruistic. [4] [5]

Repetition matters because long-term payoffs from mutual cooperation can exceed the short-term gain from defecting. A game with a known final round has a problem: players expect defection on the last round, and this reasoning unravels back to the first. If the end point is unknown and the chance of meeting again is high enough, that temptation disappears. Experiments found that cooperation in finitely repeated games lasts longer than theory predicts, though defection rises as the known end nears. [2] [3]

Axelrod and Hamilton (1981) invited programmers to design winning strategies for a repeated Prisoner's Dilemma. Despite some complex entries, the simplest won: tit-for-tat, which cooperates first and then copies its partner's previous move. It did well in both tournaments and in an evolutionary simulation where high scorers left more offspring. It rewards cooperators but cannot be easily exploited by defectors. [4] [2]

Axelrod's model indicates that reciprocity-based cooperation can start in an asocial world, thrive among many other strategies, and resist invasion once established. One source describes tit-for-tat as an evolutionarily stable strategy if the game is repeated enough times. [1] [2]