Anytime deduction for probabilistic logic (Q1337678)

The authors present a set of inference rules that in any reasoning on probabilistic logic (in the sense of Nilsson) based on conditional probabilities will yield the tightest interval of confidence given the information processed up to that point, under the proposed rules. In contrast to Nilsson's approach, all arguments are based directly on these rules, no systems of linear equations are solved. The rules are based on the familiar ways of assigning intervals in probabilistic logic; however, no consistency proof is offered.