Book: The Evolution of Cooperation by Robert Axelrod
Five stars out of five.
Suppose two generals each have two choices: attack or defend. The decisions are simultaneous and private. Military strategy favors the offensive, so both really want the other guy to defend while he attacks and really does not want to defend while the other guy attacks. On the other hand, war is extremely bloody. Both generals agree that mutual defense is better than mutual aggression. What should we expect the generals to do?
Intuitively, you might think that mutual defense is a reasonable outcome since peace is an agreeable outcome. However, this fails to appreciate individual incentives. If one general knows the other will play defensively, he should take advantage of his rival’s cooperation and attack. As a result, mutual aggression is the only sustainable outcome. But war is worse for both parties. This is the tragedy known as the prisoner’s dilemma: both parties end up in a mutually despised outcome but cannot commit to the better result due to their selfish individual incentives.
The prisoner’s dilemma has been around since the 1950s. For the next three decades or so, game theorists speculated that repeated interaction could solve the cooperation problem. Perhaps war favors the aggressor, but only a slight degree. If so, the generals could agree to maintain the peace as long as the other guy did. But the moment one slips up, the generals will fight all-out war. The threat of a painful breakdown in peace might incentivize the generals to never start conflict, even if a surprise attack might yield short-term benefits.
However, the cooperative solution remained elusive…until Robert Axelrod’s The Evolution of Cooperation. For a three sentence summary, Axelrod shows that these generals can adopt a “grim trigger” strategy and credibly promise infinite punishment in the future to enforce cooperation in the present. Thus, even bitter rivals can maintain friendly relations over the long term. In essence, we can rationally expect cooperative relationships in even the worst of environments.
Despite how I glossed over all of the intricacies of the repeated prisoner’s dilemma, The Evolution of Cooperation is a must-read for that result alone. But the book is so much more. I first picked it up during my junior year of college. I hadn’t taken a math class in five years, and the grade in that class was a C. Yet, despite the sophistication of the argument, I understood exactly what was going on. Axelrod’s exposition of formal theory in this book is quite simply the best you will ever see.
The fourth chapter is nothing short of awesome. Axelrod takes cooperation to the limit in his study of the “live and let live” trench warfare system during World War I. For a significant chunk of the war, troops spent most of their time deliberately shooting to miss their enemies in the opposing trench. While shooting and killing an enemy soldier provided a marginal gain should a battle take place, said act of shooting risked sparking a larger battle which would cause great causalities on both sides. Thus, for the sake of self-preservation, armies avoided fighting. This culminated in the famous Christmas Truce, in which the troops actually got out of their trenches and began fraternizing with the so-called enemies. (In that vain, you should watch Joyeux Noel if you have not already.)
If there is one issue with the book, it is the emphasis on tit-for-tat. Tit-for-tat is a less aggressive way of responding to your opponent’s aggression than grim trigger; rather than punishing forever, you merely punish at the next available opportunity. Axelrod correctly identifies a bunch of nice properties of tit-for-tat, especially how well it plays nice with others. However, as every modern game theorist knows, tit-for-tat is not subgame perfect and thus is extremely questionable on theoretical grounds. Of course, we would not have found out about that had this book not existed, so this just further solidifies how important The Evolution of Cooperation is.
In sum, go out and buy it. The book has applications to game theory, economics, political science, sociology, evolutionary biology, and psychology. If you are reading this blog, you likely have an interest in one or more of those fields, so you should pick it up.