The Evolution of Cooperation

Robert Axelrod

6 ideas

1. Tit-for-tat wins the computer tournaments

   Axelrod invited game theorists to submit strategies for an iterated Prisoner's Dilemma round-robin tournament, and Anatol Rapoport's four-line tit-for-tat, which cooperates first and then copies the opponent's last move, won. In a second tournament entrants knew the result and tried to beat it, and tit-for-tat won again. It never scored higher than any single opponent in a head-to-head match. It won on total points because it drew out mutual cooperation from many different partners.

2. Four properties of successful cooperative strategies

   Strategies that did well shared four traits. They were nice, meaning never the first to defect. They were retaliatory, punishing defection at once so exploitation did not pay, and forgiving, returning to cooperation after the other side did. They were also clear, simple enough that opponents could recognize the pattern and adapt to it.

3. The shadow of the future

   Cooperation between self-interested parties is stable only when future interactions weigh enough against present gains. When the chance of meeting again is high and future payoffs are not heavily discounted, the long-run cost of losing cooperation outweighs the one-time gain from defecting. Anything that makes interactions more durable or frequent, or makes the future count for more, therefore promotes cooperation.

4. Live-and-let-live in WWI trenches

   On stretches of the Western Front, opposing units facing each other for long periods developed unspoken truces. They fired at predictable times and places, or deliberately missed, and answered violations with measured retaliation. It arose because the same units stayed in repeated contact, and it broke down when high command imposed raids that disrupted that stable reciprocity.

5. Cooperation can invade via small clusters

   A population of unconditional defectors cannot be invaded by lone reciprocators. A small cluster of reciprocators who interact mostly with each other can earn enough from mutual cooperation to outscore the defectors and spread. Once reciprocity is established, it resists re-invasion by defectors. Cooperation's evolution is therefore a ratchet that needs no central authority, friendship, or altruism.

6. Don't be envious in non-zero-sum games

   People often judge their performance by comparing it with their counterpart's score, which pushes them to defect to stay ahead. In non-zero-sum relationships the right benchmark is how well you could do with another strategy given the other player's behavior, not whether you beat them. Seen this way, letting a partner do as well as or better than you is compatible with maximizing your own outcome.

Save and mark ideas in the app