What makes expert systems survive over 10 years—empirical evaluation of several engineering applications

摘要

This case study analyzes eight expert system applications that have successfully been in industrial use for a long time. We have personally been involved in the development of these applications and are thus in a good position to analyze what is important for a successful application and what kind of mistakes can be made. Since the development of the applications started in 1986–1990 and some of them are still in use we are able to observe what has happened to those applications during their lifetime. Our key observations are related to the scope of the applications, to the trade-off between usability and automation, to the role of human experts in the use and development of expert systems, on the technical solutions used, on aspects of the operation of the expert system and on the similarities between expert systems and information systems. The key findings are expressed as 20 hypotheses for successful expert systems. The support of each application to the hypotheses is discussed.