EXPERT SYSTEM EXPLANATION IN RETROSPECT - A CASE-STUDY IN THE EVOLUTION OF EXPERT SYSTEM EXPLANATION

Over the last two decades, expert systems have grown from a novel research interest to a thriving industrial technology. Constant during this transition has been the emphasis on the importance of an expert system's ability to explain its solutions to the user, As expert systems have moved from research to practice, the "users" of these systems have progressed from almost exclusively the authors of the system (i.e., the. knowledge engineers), to highly trained domain experts, and are now beginning to include the novice end user. However, the methods used to produce expert system explanations have remained relatively constant. This article retrospectively examines the underlying assumptions of these methods and identifies how they fail to meet the changing needs of today's expert system end users. Three expert system projects are described to illustrate the evolution of expert system explanation methodology motivated by the failure of earlier techniques to satisfy end user explanation requirements.