The rationale of rationality

Our research starts from the assumption that actors use a single decision theory to guide them on how to behave in all possible one-shot two-person encounters. To address which decision theories perform well, we let 17 theories compete in a large number of randomly selected symmetric 2 × 2 games. It turns out that the decision theory that optimizes its own payoff under the assumption that the other actor behaves randomly wins by a small margin. Second, we study the `evolution of rationality.' In a quasi-biological setup where more successful strategies generate more offspring, the decision theory that always plays the behavior that belongs to the risk-dominant Nash equilibrium emerges as the long-term survivor from an initially mixed pool of decision theories. We also confront the decision theories with human experimental data. The decision theory that always aims for the highest possible payoff for itself performs best against humans.