Abstract

This interdisciplinary dissertation studies uncertain inference pursuant to the purposes of artificial intelligence, while following the tradition of philosophy of science. Its major achievement is the extension and integration of work in epistemology and knowledge representation. This results in both a better system for evidential reasoning and a better system for qualitative non-monotonic reasoning. ;By chapter, the contributions are: a comparison of non-monotonic and inductive logic; the effective implementation of Kyburg's indeterminate probability system; an extension of that system; a proposal for decision-making with indeterminate probabilities; a system of non-monotonic reasoning motivated by the study of probabilistic reasoning; some consequences of this system; a conventionalistic foundation for decision theory and non-monotonic reasoning.