Validation of Computational Models Based on MultipleHeterogeneous Knowledge Sources

  • Authors:
  • Daniel E. O‘Leary

  • Affiliations:
  • University of Southern California, 3660 Trousdale Parkway, Los Angeles, CA 90089-1421. E-mail: oleary@rcf.usc.edu

  • Venue:
  • Computational & Mathematical Organization Theory
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

Theories of organizations have brought together multipleheterogeneous theories in computational models. In addition, inartificial intelligence, there has been an emphasis on the generationof knowledge-based systems that include multiple heterogeneousknowledge bases. As a result, increasingly, theory and modeldevelopers have called for the need to validate these computationalmodels. Unfortunately, there has been only limited attention givento validation of multiple knowledge source programs.The primary focus of this paper is on the identification of conflict between multiple knowledge bases. The existence of conflict is particularly critical in those situations where database evaluations are “averaged”. For example, what would it mean to average the assessments of supply and demand economists, or surgeons and chemotherapists?Correlational statistics are used to identify conflict situations. In addition, a new approach, referred to as cutpoints, is developed to determine if probability distributions of multiple agents are in conflict. A case study is used to illustrate the problems of combining expertise in multiple agent systems and to demonstrate the approach.