Skip to content
mentua.org

"The Evolution of Cooperation" by Robert Axelrod: Summary and Review

Line drawing of two rows of circles joined by crossing arrows, each circle answering the one before it in the other row, the first circle filled and the last two rounds only dottedLine drawing of two rows of circles joined by crossing arrows, each circle answering the one before it in the other row, the first circle filled and the last two rounds only dotted

10 min read≈ 2,150 words5 questions · up to 5 points

What this is about

Around 1980 a political scientist asked experts to send in computer programs for a game of trust and betrayal, and the shortest and plainest program beat all the clever ones. You learn the main ideas of the book that grew from those tournaments, from the shadow of the future to the truces in the trenches of World War I. You also learn which of its claims later research corrected and how to read it in a few evenings.

The book in three sentences· skip ahead

Axelrod asks how selfish players can come to cooperate when no authority forces them to. In two computer tournaments of the repeated prisoner's dilemma, the simple strategy Tit for Tat won: cooperate first, then copy the other player's last move. The book shows that such reciprocity can start in small groups, defend itself once it is common and grow even between enemies, as long as the players are likely to meet again.

The Evolution of Cooperation

Author
Robert Axelrod
Published
1984
Publisher
Basic Books, revised edition, 2025 printing
Length
272 pages
Audiobook
6h, read by Roger Wayne
Best for
Readers who want to understand from the source why reciprocity works between selfish players, and who enjoy a clear argument built on a simple game.

Around 1980 a political scientist at the University of Michigan found that the simplest computer program in a contest of trust and betrayal had beaten all the clever ones. Robert Axelrod had asked game theorists to send in strategies for a game played over and over, and the winner did nothing more than cooperate first and then copy its partner’s last move. The Evolution of Cooperation (1984) explains why.

The book appeared at the height of the Cold War, when many people believed that two rivals without a common ruler could only arm against each other, and it argued that selfish players can learn to cooperate on their own. This review explains the book’s main steps, compares it with two neighbors and says what later research corrected.

Cooperation among selfish players is a repeated prisoner’s dilemma

Axelrod opens with one question: under what conditions will cooperation emerge among egoists when there is no central authority? By egoists he means players who care only about their own gain, and by central authority a government, a boss or anyone else who can force them. His first example is two countries with trade barriers. Both would gain from free trade, but whatever the other country does, each is better off keeping its own barriers, so both keep them and both lose. He also describes the habit of trading favors in the United States Senate, where senators help each other because they know they will need each other again.

To study what all these cases share, Axelrod uses the prisoner’s dilemma, a game in which each of two players can cooperate or defect. In his version, both get 3 points if both cooperate and 1 point each if both defect. If one defects while the other cooperates, the defector gets 5 points and the cooperator nothing. Whatever the other player does, defecting pays more in a single round, which is why the game is explained in What Is the Prisoner’s Dilemma? as a trap. Axelrod’s move is to ask what happens when the same two players meet again and again, because then a defection today can be answered tomorrow.

The simplest program, “Tit for Tat”, won both computer tournaments

Chapter 2 tells the story that made the book famous. Axelrod invited experts to submit programs for a round-robin tournament in which every program played every other one for 200 moves. Fourteen entries and one program that chose at random took part. Some entries were complex; one built a statistical model of its opponent after every move. The winner was the simplest program, Tit for Tat, sent in by the game theorist and peace researcher Anatol Rapoport: cooperate on the first move, then do whatever the other player did on the move before.

Axelrod published the results and held a second round. This time he received 62 entries from six countries, from computer hobbyists as well as professors of biology, physics and computer science, and nobody knew in advance how many moves a game would last. Everyone had seen the first results, and Rapoport sent in Tit for Tat again. It won again.

Axelrod then analyzed why. The successful strategies were nice, meaning they were never the first to defect. They were provocable, hitting back at once after an unprovoked defection. They were forgiving, returning to cooperation as soon as the other did, and they were clear, so that the other player could see the pattern and adapt to it. He also ran an “ecological” version, in which each program came back in the next round in proportion to its earlier success, and over many rounds Tit for Tat displaced all the other programs.

Cooperation needs a long “shadow of the future”

The tournaments show what works, and chapter 3 and the work with the biologist W. D. Hamilton show when it works. If two players know which round is the last, each defects in that round, then in the round before, and cooperation comes apart all the way back to the first move. Cooperation therefore needs games of unknown length and what Axelrod calls a large shadow of the future: a high chance that the same players meet again, and a next meeting that matters to them.

Axelrod turns this into a precise result. A strategy is collectively stable when a population using it cannot be invaded by a newcomer playing anything else. With the book’s points, Tit for Tat is stable only if the chance of meeting again is at least two in three. Below that, a newcomer who defects on every second move does better against it, and below one in two, so does a player who always defects.

Always defecting, by contrast, is stable at every level of that chance, because a lone cooperator among defectors only loses. In Axelrod’s adapted essay, the soldiers of World War I and the superpowers of the Cold War both illustrate the same point: what keeps cooperation going is the expectation of a long relationship.

Reciprocity can start in small clusters and then works like a ratchet

If everyone defects and always-defect is stable, how could cooperation ever begin? Axelrod’s answer is clustering. A single reciprocator among defectors is exploited, but a small group of reciprocators who have even a modest share of their meetings with each other can score more than the defectors around them. The theory also allows a second start: among relatives, helping a partner also helps the genes they share, as explained in Kin Selection and Hamilton’s Rule, Explained.

Once a nice strategy like Tit for Tat is common and stable, the process cannot easily run backward. A cluster of defectors who arrive together still meets mostly reciprocators, and since nice players already get the best score with each other, the newcomers cannot do better. Axelrod and Hamilton put it in one image in their Science paper of 1981:

The gear wheels of social evolution have a ratchet.

Robert Axelrod and William D. Hamilton, “The Evolution of Cooperation”, Science (1981)

A ratchet is a mechanism that turns one way only. Cooperation can grow from a few clusters, and once it is established, it holds.

Enemies in the trenches cooperated without friendship

Chapter 4 is the book’s most surprising case. On quiet parts of the Western Front in World War I, British, French and German soldiers often avoided shooting to kill, as long as the other side held back too. Axelrod draws on the sociologist Tony Ashworth’s 1980 study of soldiers’ diaries and letters, and he calls the result the live-and-let-live system. It worked because trench warfare kept the same small units facing each other for months, so every shot could be answered. Soldiers also fired in routine, predictable ways to show that they could hit back, a ritual that told the other side two things at once: that they were holding back, and that they were ready.

Axelrod also finds something the simple model leaves out. Over time, the soldiers came to value the restraint for its own sake and to see breaking it as wrong. He sums this up in one sentence: “not only did preferences affect behavior and outcomes, but behavior and outcomes also affected preferences.” The system ended where the high command ordered raids into enemy trenches. A raid could be checked from headquarters, so it could not be faked, and it had to be answered, which destroyed the conditions that the tacit restraint needed.

Cooperation can evolve without foresight, even in bacteria

Chapter 5, written with Hamilton, applies the same logic to living things that cannot think ahead at all. A strategy does not need a brain; it only needs to be inherited and to respond to what others do. Bacteria qualify, and so do plants and fish.

Fig trees, for example, often drop a young fig when the wasp inside lays eggs in almost all its flowers instead of pollinating them, and the wasp’s young die with it. Cleaner fish, which pick parasites off larger fish, live in fixed home ranges on reefs, where the same partners meet again, and they seem to be unknown in the open sea. The authors also offer speculative ideas about illness: a microbe living inside a host might turn harmful when the host is sick or old, because the shadow of the future has become short. They mark these ideas openly as speculation.

Four rules for players and five ways to promote cooperation

The last part of the book turns the results into advice. Chapter 6 gives four rules: do not be envious, do not be the first to defect, reciprocate both cooperation and defection, and do not be too clever. The first rule is the least obvious. Most people judge themselves by whether they did better than the other side, but in a game that is not zero-sum, envy leads you to defect and to lose points you could have shared. Axelrod admits in his adapted essay that the results changed his own view: he had believed one should be slow to anger, and the tournaments showed that a quick, measured response sends a clearer signal.

Chapter 7 is addressed to reformers, meaning people who design the setting rather than play the game. Axelrod’s first way is to enlarge the shadow of the future by making meetings more frequent or more durable. The others are to change the payoffs, teach people to care about each other, teach reciprocity, and help players recognize each other and remember how each one behaved. Chapter 8 adds how labels, reputations, regulation and territory shape who meets whom, and chapter 9 draws the results together. These later chapters are shorter and more suggestive than the tournament chapters, but they are where the book becomes useful for daily life.

How it differs from other books on cooperation

Two neighbors show what is special about the book. Richard Dawkins added a chapter called “Nice guys finish first” to the 1989 edition of The Selfish Gene, and it retells Axelrod’s tournaments for readers interested in evolution. As the review “The Selfish Gene” by Richard Dawkins: Summary and Review explains, Dawkins uses the result as one piece of a gene-centered view of life. Axelrod gives the full argument, the proofs and the human cases, and Dawkins later wrote the foreword to its revised edition.

Elinor Ostrom’s Governing the Commons (1990) starts where Axelrod stops. His players are two at a time, cannot talk and cannot change the rules, while Ostrom’s farmers and fishers are many, meet, write rules and punish cheats. “Governing the Commons” by Elinor Ostrom: Summary and Review shows what that adds. Axelrod’s distinctive method is the tournament itself: instead of proving which strategy is best in theory, he let many experts’ ideas compete and then explained the result.

Where it falls short

The book has three main weaknesses. The first concerns its central result. In 1987 Robert Boyd and Jeffrey Lorberbaum showed in Nature that no fixed strategy, Tit for Tat included, is evolutionarily stable in the repeated prisoner’s dilemma, because other strategies can always spread into it under some conditions. Axelrod’s strongest statements about Tit for Tat need that qualification.

The second is errors. Axelrod tested moderate mistakes and found Tit for Tat fairly robust, but later studies made the problem clearer: when moves are sometimes misread, two Tit for Tat players can fall into a long chain of revenge. In 1993 Martin Nowak and Karl Sigmund showed that a strategy called win-stay, lose-shift does better under such conditions, and more generous versions of Tit for Tat also do better.

The third is the trench case. The political scientist Joanne Gowa in 1986 and the statistician Andrew Gelman in 2008 argued that the soldiers were not really in a prisoner’s dilemma. Gelman’s reason is that holding fire protected each soldier from return fire, so restraint needed no special explanation. The history remains fascinating, but it fits the model less neatly than the book suggests. Some examples, especially those about arms control between the United States and the Soviet Union, also show the book’s age.

Which edition to read

Read the revised edition, which adds Dawkins’s foreword; the 2025 Basic Books printing is the easiest to find. Print or ebook is best if you want to look at the tournament tables and the appendices with proofs, which you can skip without losing the argument. The unabridged audiobook, read by Roger Wayne, runs about six hours and works well for the trench and biology chapters, but the payoff numbers and stability conditions are easier to follow on the page.

A good plan is chapters 1 to 4 first, then 6 and 7, and Axelrod’s free eight-page essay as a summary to keep. If the topic holds your interest, Nowak and Sigmund’s short paper of 1993 shows what changes when players make mistakes.

Quiz

5 questions · up to 5 points

Open a question and pick one answer. Only your first answer counts: +1 if it is right, −1 if it is wrong.

  1. Why did Tit for Tat get the highest total score in Axelrod's tournaments?
    Show answer

    It earned the reward for mutual cooperation with every strategy that was willing to cooperate, and it never let itself be exploited for long. Tit for Tat never scores more than the strategy it plays against. It wins on its total, because it gets the cooperative payoff with many partners and answers every defection at once.

  2. Two companies supply the same customer and could fight a price war. According to Axelrod, what makes stable cooperation between them more likely?
    Show answer

    A high chance that they will keep dealing with each other for a long time, so that next year matters to them as much as this year. This is the 'shadow of the future'. When the next encounter is likely and important enough, defecting today costs more than it gains.

  3. In a world where almost everyone always defects, how can cooperation based on reciprocity get started, according to the book?
    Show answer

    A small cluster of reciprocating players who meet each other often enough can do better than the defectors around them. One cooperator alone is exploited. But a cluster gives each member a real chance of meeting a partner who returns the cooperation, and that is enough to gain a foothold.

  4. What does Axelrod say ended the 'live and let live' system in many quiet sectors of the Western Front?
    Show answer

    Raids ordered by the high command, which could be checked from headquarters and made each side's restraint impossible to keep. Axelrod argues that the raids destroyed the conditions that the tacit restraint needed, because they could not be faked and had to be answered.

  5. How does this review suggest you read The Evolution of Cooperation?
    Show answer

    Read the tournament chapters, the trench chapter and the advice chapters closely, skim the mathematical appendices, and then look at the later research on errors. The argument is carried by the tournaments, the trench case and the advice. The proofs are optional, and the later work on mistakes corrects the strongest claims for Tit for Tat.

The takeaway

The Evolution of Cooperation argues that selfish players without a ruler can learn to cooperate when they are likely to meet again, which Axelrod calls a long shadow of the future. Tit for Tat won his two tournaments because it was nice, hit back at once, forgave quickly and was easy to read. The book adds evidence from the trenches of World War I and from biology, and practical advice for players and reformers. Later work showed that Tit for Tat suffers from errors, so read the book as a clear model, not a final answer.

Key quotes

“Under what conditions will cooperation emerge in a world of egoists without central authority?”
Robert Axelrod, The Evolution of Cooperation, essay adapted from the book, in Breakthrough: Emerging New Thinking (1988)
“The foundation of cooperation is not really trust, but the durability of the relationship.”
Robert Axelrod, The Evolution of Cooperation, essay adapted from the book, in Breakthrough: Emerging New Thinking (1988)

Should you read it?

Read it if

  • You know the prisoner's dilemma and want to see why it changes completely when the same players meet again.
  • You deal with people you will meet many times, such as colleagues, suppliers, neighbors or rival firms, and want rules for dealing with them.
  • You want to understand a classic that shaped economics, political science and evolutionary biology.

Skip it if

  • You want the current state of research; the book is from 1984, and later work changed the picture, especially on errors.
  • Your problem has many players who can talk and make rules, such as a shared resource; Ostrom's Governing the Commons fits better.
  • You only need the main result; the review of the prisoner's dilemma and Axelrod's short essay give it in a few pages.

As an Amazon Associate, mentua.org earns from qualifying purchases. It costs you nothing and never shapes the verdict.

Ideas from this book

References

  1. The Evolution of Cooperation. Robert Axelrod and William D. Hamilton, Science 211 (4489), 1390–1396, 1981

    The paper that became chapter 5 of the book, with the tournaments, the stability proof and the biological examples.

  2. The Evolution of Cooperation. Robert Axelrod, Essay adapted from the book, in Breakthrough: Emerging New Thinking, 1988

    Axelrod's own eight-page summary of the book: the tournaments, the trenches, the shadow of the future and his advice.

  3. A strategy of win-stay, lose-shift that outperforms tit-for-tat in the Prisoner's Dilemma game. Martin Nowak and Karl Sigmund, Nature 364, 56–58, 1993

    The best-known correction: when players sometimes make mistakes, a different simple strategy does better than Tit for Tat.

  4. Methodology as ideology: some comments on Robert Axelrod's The Evolution of Cooperation. Andrew Gelman, QA Rivista dell'Associazione Rossi-Doria, 2008

    A critique arguing that the soldiers in the trenches were not really in a prisoner's dilemma.

the-evolution-of-cooperationrobert-axelrodbook-reviewsummarycooperationreciprocitytit-for-tatprisoners-dilemmaiterated-prisoners-dilemmagame-theoryshadow-of-the-futurecomputer-tournamentanatol-rapoportwilliam-d-hamiltonlive-and-let-livetrench-warfareevolutionary-game-theorycollective-action