Logic & Philosophy of Science

Richard Jeffrey has devised a probability revision method, called reparation, that increases the probability of hypothesis H when it is discovered that H implies previously known evidence E. The speaker has generalized Jeffrey's method to the case of uncertain old evidence and probabilistic new explanation, based on the so-called "uniformity principle," which, in one of its several avatars, dictates that explanation-based revisions should preserve certain Bayes factors. A plausible alternative principle dicates that observation- and explanation-based revisions should commute. One formulation of the latter principle is equivalent to the uniformity principle, but another is not. In this talk, we will survey the mathematical properties of these competing revision principles, and raise the question of whether one has a superior claim to ground a generalization of reparation, or if each might be applicable to a separate class of uncertain old evidence problems.