Identifying long-term success issues of expert systems

摘要

This investigation identifies issues associated with the long-term success of expert systems and the specific nature of maintenance. Open-ended structured interviews were conducted with several expert system developers to assess factors critical to the long-term success of expert systems. The interviews provide an opportunity to examine the developer's perspective. Questions regarding general expert system issues and questions addressing maintenance-specific issues were the basis for the interviews. The interviews were analyzed by three reviewers, and the results were aggregated to create a list of relevant issues associated with long-term success of expert systems. Maintenance, verification and validation, case-based reasoning, realistic user expectations, and problem definition were the five most identified issues. Perceptions of expert systems maintenance were specifically investigated. The dynamic nature of systems was mentioned most often as a critical issue for maintaning expert systems. Results from this investigation provide an awareness of the issues associated with long-term maintenance of expert systems and its relative importance for an expert system development methodology.