An Axiomatic Characterization of Bayesian Updating

We provide an axiomatic characterization of Bayesian updating, viewed as a mapping from prior beliefs and new information to posteriors, which is disentangled any reference preferences. updating characterized by Non-Innovativeness (events considered impossible in the remain posterior), Dropping contradicted evidence are Proportionality (for other events, posterior simply rescales prior’s probabilities proportionally). The result clarifies differences between normative benchmark, alternative models, actual human behavior.