Why CDT agents always defect in the Twin Prisoner's Dilemma
Trotztd · x · 2026-10-18
A fragment of a decision-theory exchange between @jessicata and @Trotztd. jessicata imagines Alice and Bob as CDT+game-theoretic agents who know who they are and each care only about their own payoff; in a Twin Prisoner's Dilemma both defect, and this doesn't obviously violate ex ante optimality. Trotztd counters: whether I'm at position A or B, defecting maximizes value regardless of the other's choice — but if the two players are copies of one brain emulation, "I might be the other copy" becomes relevant and the analysis shifts.
Related event: Decision Theory Debate: Copying an AI Constitutes Value Drift(4 posts)→