MODELING BEHAVIOR SEQUENCES: PRINCIPLES, PRACTICE, PROSPECTS

Date
1985-12-01
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Many knowledge-based computer programs embody finite-state models which represent procedural knowledge, and use them to predict or interpret new observations. In accordance with prevailing methods of constructing expert systems, such models are invariably created by the knowledge engineer explicitly for the program's use. However, in most cases it would be of great practical (and theorectical) value if the program could infer models automatically from observations of behavior. This paper takes a wide-ranging look at the problems involved in building casual models of discrete sequences of symbols, and surveys methods of modeling. A key question for the future is the inference of goals and plans from observations of behavior, for procedural representations are much less flexible than the goal-oriented approach which people seem to favor. This issue is discussed and some speculations are included on how goals and plans might be inferred.
Description
Keywords
Computer Science
Citation