Modeling of individual beliefs is essential to the computer understanding of natural languages. Phenomena at all levels -- syntactic, semantic, and pragmatic -- cannot be fully analyzed in the absence of models of a hearer and of the hearer's model of other believers. The heart of this text is the presentation of an artificial intelligence (AI) program intended to simulate certain aspects of a human believer. This book provides a prolog program, Viewgen, that maintains belief structures about the world and other believers, and is able to ascribe beliefs to others without direct evidence by using a form of default reasoning. The authors contend that a plausible model such as this can -- in the best cognitive science tradition -- shed light on the long-standing philosophical problem of what belief is.
The issues presented here will be of considerable interest to an informed general reader as well as those with a background in any of the disciplines that make up what is now called cognitive science: philosophy, linguistics, psychology, neuropsychology, and also AI itself.
"What, therefore, is especially important about this book for anthropologists is the rigorously intensionalist(p.6) view of cognition, meaning, and hence "belief", a view necessary for any plausible comparative theory of cultural systems, and the demonstration that only such a view can be coherent, adequate, and finitely specifiable…. a highly motivated and rigorous computational model…"
"In about the fortieth year of its existence, artificial intelligence may just be beginning to show signs of maturation….undoubtedly stimulating contents."
"…this book is interesting, applies a broad range of new belief attribution heuristics in very creative ways…"
—Minds and Machines
Contents: Introduction. Preliminaries on the Nature of Belief. Belief Ascription. Experiments in Belief. Global Issues: Reasoning with Viewpoints. Further Extensions and Speculations.