A Bayesian view of assessing uncertainty and comparing expert opinion

A Bayesian approach to the problem of comparing experts or expert systems is presented. The question of who is an expert is considered and comparisons among well-calibrated experts are studied. The concept of refinement, in various equivalent forms, is used in this study. An informative example of the combination of the opinions of well-calibrated experts is described. Total orderings of the class of well-calibrated experts are derived from strictly proper scoring rules.