Viewing Knowledge Bases as Qualitative Models

  • Authors:
  • William J. Clancey

  • Affiliations:
  • -

  • Venue:
  • IEEE Expert: Intelligent Systems and Their Applications
  • Year:
  • 1989

Quantified Score

Hi-index 0.00

Visualization

Abstract

A viewpoint on expert systems is developed that considers the qualitative nature of knowledge encoded in such programs. It is maintained that all knowledge bases contain models of systems in the world. Reasoning involves sequences of tasks (for example, monitoring and diagnosis) by which an understanding or model of specific situations is related to action plans. Programs use a simple repertoire of qualitative modeling techniques commonly called knowledge representations. The situation-specific-model concept makes concrete exactly what programs know and how to describe problem solving in terms of model-manipulation operators. The viewpoint is exemplified for the particular case of medical diagnosis. A historical perspective shows how AI's concern with adaptiveness and the rationality of the autonomous agent emphasizes the role of models as what a problem-solver knows. It is suggested that this has been to the detriment of understanding the primary characteristic of knowledge in terms of models that partition the world, viewing it selectively and making it coherent for some purpose.