Discussant's response to "Expert systems and AI-based decision support in auditing: Progress and perspectives";

              Discussant's Response to
"Expert Systems and AI-Based Decision
Support in Auditing: Progress and Perspectives"
Dana A. Madalon
Frederick W. Rook
Price Waterhouse
1. Introduction
A critical issue affecting progress in the development of Al-based deci-sion support systems for auditing is the relationship between research and application development. In order to present our view of the relationship be-tween these two concepts, it is useful to first discuss our perspective and back-ground in both AI technology research and expert system development.
As AI technology researchers, we have conducted research in knowledge acquisition, knowledge representation, natural language analysis and un-derstanding, planning and design, and computational theory. For example, we have examined and advanced the use of constraint satisfaction problem formulations as a method of inferencing. We recognize the extent to which the state of AI technology is driven by research in the areas of computer sci-ence, computer engineering, cognitive psychology, decision sciences, oper-ations research, human factors engineering, and mathematical logic. To ensure the most effective use of these technical developments to the applied realm, we have worked closely with a number of leading AI researchers. These include Dr. Robert Wilensky at the University of California Berkeley AI Re-search Center, Drs. Judea Pearl and Rina Dechter at the Cognitive Systems Laboratory of the University of California Los Angeles, Dr. Drew McDermott at the Yale University AI Project, Drs. B. Chandrasekaran and John Joseph-son at the Ohio State University Laboratory for Artificial Intelligence Re-search, and Dr. Andrew Sage at the George Mason University School of Information Technology and Engineering.
As expert system developers, we have designed, developed, and imple-mented over thirty prototype and operational expert systems in a variety of application areas. Our expert systems have addressed such problem types as monitoring, diagnosis, assessment, risk analysis, resource allocation, scheduling, and planning. While we have successfully fielded operational ex-pert systems, we have also met technological hurdles too great to be over-come with today's technology. The foundation of our success in building expert system applications is the ability to leverage existing AI technology, i.e., technology that in many cases has been effectively transferred from univer-sity settings.
74