People often believe what they want to believe rather than what the evidence implies. Here Melnikoff and Strohminger find that this seemingly irrational tendency may emerge from fully rational Bayesian calculations.