Making Assumptions for Efficiency Reasons

Abstract

Problem-solving methods are present in most of the current knowledge engineering frameworks.¹⁾ They are used to describe the reasoning steps and types of knowledge which are needed to perform a task by a knowledge-based system. However, a question that has not been answered clearly is the connection between problem-solving methods and the efficiency of the problem-solving process. Most descriptions of problem-solving method frameworks do point to problem-solving methods as being somehow related to efficiency, however no framework makes this relation explicit, i.e. explains how problem-solving methods achieve their efficiency. Other approaches interpret the knowledge level paradigm of [Newell, 1982] such that the description of problem-solving methods should exclude all efficiency concerns, which are regarded as being dealt with at the symbol level. Finally, some approaches claim not to be concerned with efficiency since their problem-solving methods are only used to capture the expert’s problem-solving behavior. But we must be aware that experts also have to solve the task given their real-life limitations. In fact, a large share of expert-knowledge is concerned exactly with efficient reasoning given these limitations [VanLehn, 1989]. The conceptualization of a domain by an expert differs from the conceptualization of a novice because the former reflects the learning process that led to efficiency in problem solving.

[Chandrasekaran, 1986], [Marcus, 1988], [Steels, 1990], [Chandrasekaran et al., 1992], [Puppe, 1993], [Schreiber et al., 1993], [Terpstra et al., 1993], [Breuker & Van de Velde, 1994], [Schreiber et al., 1994], [Benjamins, 1995], [Eriksson et al., 1995], [Motta & Zdrahal, 1996], [Angele et al., 1998].