Super Rationality

Douglas Hofstadter sketched out a theory of super rationality in a couple of his Metamagical Themas columns in Scientific American in the mid-80s (he later collected his columns into a very interesting book). It was an attempt to solve the prisoner’s dilemma. The basic idea was that you should act as though everybody else is going to think the problem through the same way you do, and will therefore come to the same conclusion. Thus, in the prisoner’s dilemma, you can conclude that the other player will make the same decision as you no matter how much you think about it, so it follows that you should cooperate.

This is sort of like an amusing logic puzzle. Everybody on an island is a perfect logician. Some people’s eyes are blue. There are no mirrors on the island, but everybody can see everybody else and tell what color their eyes are. Somebody decrees that all blue eyed people must leave the island at midnight. Four days later, some people leave. How many people left?

It is also, of course, sort of like Kant’s categorical imperative, in that you should act as though everybody will act the same way you do.

The problem with super rationality, and, for that matter, with the categorial imperative in actual practice, is that people simply aren’t that rational. It’s hard enough for people to know what they themselves really want to do. It’s downright impossible to figure out what everybody should do.

In practice we completely fail to be super rational, and instead we solve the prisoner’s dilemma through simple guidelines: “nobody likes a tattle-tale.” These guidlines have generally developed to maintain cooperation in the face of temptation. I think our present society has a tendency to weaken those guidelines, stressing self-interest above all else, and thus weakening cooperation. Whether this is good or bad in the long run remains to be seen.