On the relevance of irrelevant information: evidence from a Monty Hall experiment

The Monty Hall (MH) Problem refers to a game where decision-makers systematically fail to be Bayesian when provided with a noisy signal of the true state of the world. Presenting the problem in a more intuitive framework has been shown to increase Bayesian updating by 40 percentage points. The authors extend this intuitive framework, offering participants even more “obviously dominant” strategies and, after choosing such a strategy, ask them to make a second decision following the revelation of irrelevant information. The main finding is that in spite of choosing ex ante dominant strategies, some agents nonetheless react to irrelevant information – information that should not affect the ranking of actions after Bayesian updating – by choosing strategies that (i) were ex ante dominated, and (ii) are not consistent with Bayesian updating. The authors find that exposure to their extension does not lead to more Bayes-consistent behavior in the original MH problem.