Robert Axelrod’s Computer Tournament and Tit for Tat
Around 1980, the political scientist Robert Axelrod invites experts to pit computer programs against each other. The simplest one wins: it starts friendly and afterward always does what the other did last.
Two dots take turns sending arcs to each other; golden arcs make a line grow between them, a teal arc makes it buckle. After a clear teal response and an immediate golden arc, the line straightens and branches out in gold, while a second pair below, after the first break, only exchanges pale arcs and lets nothing grow between them.
Principle #62
Tit for Tat: Start friendly, respond clearly to betrayal and forgive quickly – then cooperation can grow and last even between rivals.
4 min read5 sectionsSituationLink to the principle
Contents
How does cooperation arise between people who think only of themselves and answer to no higher referee? Robert Axelrod asked himself this question at the University of Michigan. Instead of merely theorizing about it, he organized a competition. The result, a victory for Tit for Tat, surprised the experts and became the basis of his 1984 book “The Evolution of Cooperation.”
A grid of all the encounters, next to it score bars, and at the very top, in gold, the shortest program – like the winner of Axelrod’s computer tournament.
01
The game: the prisoner’s dilemma
Situation
In the prisoner’s dilemma, two players choose at the same time whether or not to cooperate. If both cooperate, both receive a good reward. If both refuse, both get little. If one refuses while the other cooperates, the one who refuses gets the most and the trusting one gets nothing at all. In a single round, it pays for each player not to cooperate. But Axelrod had the game played for many rounds in a row.
Link to the principle
Repetition changes everything. If you know you will meet the other person again, you have to take their response into account. Axelrod calls this the shadow of the future. It is what makes cooperation between egoists possible in the first place.
02
The first round
Situation
Axelrod invited game theorists from economics, psychology, sociology, political science and mathematics to submit programs. Fourteen programs and one random strategy competed against each other in a round robin. The winner was the simplest of all the programs submitted, “Tit for Tat,” sent in by the mathematician and psychologist Anatol Rapoport of the University of Toronto. It cooperates on the first move and afterward always repeats the opponent’s last move.
Link to the principle
Tit for Tat is the principle in its purest form. It is never the first to be unfriendly. It does not put up with betrayal. And it holds no grudge: if the other side cooperates again, it cooperates again too.
03
The second round
Situation
Axelrod published the results in 1980 in the Journal of Conflict Resolution and invited entries for a second round. This time there were 62 entries from six countries, from hobby programmers as well as professors of evolutionary biology, physics and computer science. All of them knew the result of the first round and tried to beat Tit for Tat. Rapoport submitted it again, and it won again.
Link to the principle
A second victory among experts says more than the first. Sophisticated programs that tried to exploit their opponents failed because of their own cleverness. It was precisely its simplicity that helped Tit for Tat: other programs could quickly tell where they stood.
04
What the losers did wrong
Situation
A program called JOSS basically played like Tit for Tat but defected unexpectedly in about 10% of cases, even though the other side had cooperated. It did considerably worse because its slips set off an echo of mutual retaliation. The program FRIEDMAN never forgave: once betrayed, it defected forever. That did not pay off either. Even when Axelrod repeated the first round with an error rate of 1%, that is, with misunderstandings, Tit for Tat won.
Link to the principle
Small acts of exploitation don’t pay off if the other side strikes back. Endless retaliation doesn’t pay off either, because it blocks any return to cooperation. The principle needs both: response and forgiveness.
05
The four rules
Situation
Axelrod sums up the lesson in a set of recommendations. Don’t be envious of the other’s success. Don’t be the first to defect. Reciprocate both cooperation and defection. Don’t be too clever. Together with the biologist William D. Hamilton, he showed in 1981 in the journal Science that the same logic can also explain cooperation in nature.
Link to the principle
“Tit for tat” sounds like revenge. In Axelrod’s findings, it is above all an attitude of clarity: you always know where you stand with me. That creates trust, even without friendship.
Key takeaways
Key lessons
01
Start friendly, even with someone you don’t know.
02
Don’t put up with being exploited; respond quickly and clearly.
03
Forgive as soon as the other side cooperates again.
04
Be predictable, so that others know where they stand with you.
05
Think of the next encounter: it is what makes cooperation worthwhile.
Summary
Around 1980, Robert Axelrod pitted computer programs against each other in the repeated prisoner’s dilemma. Twice the winner was Anatol Rapoport’s Tit for Tat, the simplest program. It starts friendly, answers betrayal and forgives immediately when the other side cooperates again. This made Tit for Tat the best-known strategy for lasting cooperation.