Ambiguous Information and Dilation: An Experiment

With common models of updating under ambiguity, new information may increase the amount of relevant ambiguity: the set of priors may “dilate.” We test experimentally one sharp case: agents bet on a risky urn and get information that is truthful or not based on the draw from an Ellsberg urn. Under typical models, the set of priors should dilate, ambiguity averse agents should lower their value of bets, ambiguity seeking should increase it. Instead, we find that ambiguity averse agents do not change it, ambiguity seeking ones increase it substantially. We also test bets on ambiguous urns and find sizable reactions to ambiguous information.