A SOURCE OF BAYESIAN PRIORS

Establishing reasonable, prior distributions remains a significant obs tacle for the construction of probabilistic expert systems. Human asse ssment of chance is often relied upon for this purpose, but this has t he drawback of being inconsistent with axioms of probability. This art icle advances a method for extracting a coherent distribution of proba bility from human judgment. The method is based on a psychological mod el of probabilistic reasoning, followed by a correction phase using li near programming.