Research on Expert Problem Solving in Physics

Gordon S. Novak Jr.
and Agustin A. Araya
Department of Computer Sciences
University of Texas, Austin, TX 78712

Copyright © 1980 by AAAI.

This article appears in Proc. First Annual National Conf. on Artificial Intelligence (AAAI-80), Aug. 1980, pp. 178-180.

This research was supported by NSF award No. SED-7912803 in the Joint National Science Foundation - National Institute of Education Program of Research on Cognitive Processes and the Structure of Knowledge in Science and Mathematics.


Physics problems cannot in general be solved by methods of deductive search in which the laws of physics are stated as axioms. In solving a real physics problem, it is necessary to treat the problem as a "nearly decomposable system" and to design a method of analysis which accounts for the salient factors in the problem while ignoring insignificant factors. The analysis method which is chosen will depend not only on the objects in the problem and their interactions, but also on the context, the accuracy needed, the factors which are known, the factors which are desired, and the magnitudes of certain quantities. Expert problem solvers are able to recognize many frequently occurring problem types and use analysis methods which solve such problems efficiently. Methods by which a program might learn such expertise through practice are discussed.

Gordon S. Novak Jr.