Book review: An Introduction to Default Logic Philippe Besnard (Springer Verlag, New York, 1989)

One of Hubert Dreyfus's more biting criticisms of Artificial Intelligence has been that its success stories have been largely confined to toy environments. Some researchers are unperturbed by this criticism, apparently believing that an accretion of little success stories will eventually reward us with a cumulative global success story. Others have turned to renewed efforts to understand what Herbert Simon calls "weak methods" of inference, that is methods of inference which do not presuppose specific knowledge of a domain. I too believe that knowledge-poor methods of inference offer greater long-term promise for the development of AI. Philippe Besnard's book is a codification of efforts to formalize one species (genus?) of knowledge-poor reasoning, what has been variously called default, defeasible, and nonmonotonic reasoning.