skip to main content

Title: Heuristic reasoning about uncertainty: an artificial intelligence approach

A model of reasoning about uncertainty, called the model of endorsement, is presented. Part of the model is implemented in an artificial intelligence (AI) program called SOLOMON, which is also discussed. Uncertainty has a passive role in current AI programs. It is typically represented as a number that is incremented or decremented as evidence becomes available. The number has little if any influence on the control of reasoning, and the relationship between the number associated with a proposition and the evidence for the proposition is unclear. In contrast, the model of endorsement represents uncertainty as the body of reasons for believing and disbelieving a proposition. Numerical degrees of belief, which might be derived from these reasons, have not proved necessary. The reasons are called endorsements, and guide SOLOMON's reasoning. The most important characteristic of the model of endorsement is that it provides the information needed for uncertainty to influence the way a program solves problems. The dissertation focusses on this idea, its implementation, and its ramifications.
Authors:
Publication Date:
OSTI Identifier:
5900878
Resource Type:
Thesis/Dissertation
Resource Relation:
Other Information: Thesis (Ph. D.)
Publisher:
Stanford Univ.,Stanford, CA
Country of Publication:
United States
Language:
English
Subject:
99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE; ARTIFICIAL INTELLIGENCE; COMPUTER CODES; DATA COVARIANCES 990200* -- Mathematics & Computers