I don’t understand the relevance of your comment; could you explain? (Expected payout for all agents in PD increases if they can find a way to cooperate AFAIK, even if all are completely selfish.)
Expected payout for one agent increases even more if they can convince everyone else to cooperate while they defect. This is the game you want to keep the other agents from playing, and while TDT works when all the other agents use a similar decision strategy, it fails in situations where they don’t. Which is exactly the problem Eneasz was getting at.
Not just, else we say that defectors in PD are winning.
I don’t understand the relevance of your comment; could you explain? (Expected payout for all agents in PD increases if they can find a way to cooperate AFAIK, even if all are completely selfish.)
Expected payout for one agent increases even more if they can convince everyone else to cooperate while they defect. This is the game you want to keep the other agents from playing, and while TDT works when all the other agents use a similar decision strategy, it fails in situations where they don’t. Which is exactly the problem Eneasz was getting at.
defect-defect is not a win by anyone’s utility function. What are you getting at?