http://plato.stanford.edu/entries/prisoner-dilemma/

The Stanford Encyclopedia of Philosophy has an article on the Prisoner’s Dilemma (cited above). I chose to write on this article because it is 1) interesting, 2) comprehensive and 3) at the right level for Info204 students. While some of the material goes beyond the scope of this class, there are a few sections of the article that are especially relevant to course material (here and here, for example). Another reason why this article was so appealing was its view of PD in the context of Evolutionary gaming, one of the newer topics in the class.

This article explores what happens when PD is played over and over again, an occurrence commonly referred to as IPD (iterative prisoner’s dilemma). As mentioned briefly in class, this version of the game is different because each player gets a chance to “punish” or “reward” their fellow player by either cooperating or defecting in future rounds. To further explore this idea, political scientist Robert Axelrod invited game theorists to submit strategies for an IPD tournament. The highest scoring strategy went by the name of tit-for-tat. The strategy can be effectively described in two steps:

1) Cooperate on the first round
2) Every round thereafter, emulate your opponent’s  last played move

Axelrod suggested four reasons that the strategy did so well (outlined in the article).  Also, tit-for-tat appears to be the best scoring strategy for an evolutionary verson of PD (EPD). (In this game, the better a strategy does, the more “offspring” that strategy will have. We can consider offspring to be more people adopting the strategy). The authors claim that TFT is evolutionarily stable, but by the definition presented in class that isn’t technically the case. For instance, in a large population of TFTers, any cooperators will not die out (as the populace will just cooperate all the time). But in that case, the invaders are acting exactly like TFTers. They just would act differently if a defecter came along (which in the hypothetical constraint posed by our professor, will not happen).