Belief revision is both a philosophical and logical problem. From Popper’s logic of scientific discovery, we know that revision is ubiquitous in physics and other sciences. The AGM postulates and R-calculus are approaches from logic, where the R-calculus is a Gentzen-type concrete belief revision operator. Because deduction is undecidable in first-order logic, we apply approximate deduction to derive an R-calculus that is computational and has finite injury. We further develop approximation algorithms for SAT problems to derive a feasible R-calculus based on the relation between deduction and satisfiability. In this manner, we provide a full spectrum of belief revision: from philosophical to feasible revision.
This is a preview of subscription content, log in to check access.
This work was supported by National Basic Research Program of China (973 Program) (Grant No. 2005CB321901), Open Fund of the State Key Laboratory of Software Development Environment (Grant No. SKLSDE-2010KF-06), and Beijing University of Aeronautics and Astronautics.