[vacuity]

The optimal decision in the Prisoner's Dilemma is to defect, but in the iterated version, where multiple Dilemmas occur and people remember previous results, Tit-For-Tat is optimal. The real world is even less reminiscent of the Dilemma, so it's not at all clear that the Dilemma's conclusion applies.

(Tit-For-Tat: Prefer cooperating, but if the other person defected on the previous turn, defect on the current turn.)

[dragonwriter]

> The optimal decision in the Prisoner's Dilemma is to defect, but in the iterated version, where multiple Dilemmas occur and people remember previous results, Tit-For-Tat is optimal.

That’s not true. There is no optimal strategy in iterated Prisoner's Dilemma in the sense that defection is optimal in the single-round version; Tit-for-Tat performs well in certain conditions in iterated Prisoner's Dilemma, and less well in others (dependent particularly on the strategies played on the other side); in single-round, defection always produces a better outcome than defection independently of the choice made against it.

[vacuity]

Perhaps I am wrong about Tit-for-Tat. It's been a while since I checked my source. In any case, my point (not to say that you deny it) is not to take any result in an idealized game too literally, and that consistent defection is bad.