Interim report on the development of an expert system for the auditor's loan loss evaluation;

              Interim Report on the Development of an Expert System for the Auditor's Loan Loss Evaluation
Kirk P. Kelly Gary S. Ribar John J. Willingham
Peat, Marwick, Mitchell & Co.
Introduction
The Audit Research Group at Peat, Marwick, Mitchell & Co. has been interested in Artificial Intelligence (AI) and Expert Systems for a number of years. Under the auspices of the Research Opportunities in Auditing program, we have funded a number of academic research projects on the application of AI to the audit task. With the growing interest in the field and the advances in technology, it was decided to undertake a project oriented toward the development of an application model. The initial thrust was to build a prototype model for test and evaluation with the implicit intent that the model would eventually be developed into a useful audit tool for field work. This paper reports on that project in terms of the rationale for the project, the current status of the project, and the future directions for this project.
Rationale for Expert Systems
The rapid advances in computer technology and ensuing applications require that those engaged in the accounting and auditing profession be involved in exploring new application opportunities. Artificial intelligence and expert systems are clearly in the forefront of these technologies; however the conventional wisdom of expert system developers suggests that considered applications ought to be limited to environments that exhibit certain character-istics. For example, it is suggested that there should be clearly definable experts in the problem task, that there should be appropriate measures of correct vs incorrect judgments, and problems should be small yet have a high payoff.
The auditing environment has some unique characteristics that tend to make it a less likely candidate for successful deployment of expert systems. For example, many areas of auditing do not have a feedback mechanism that allows for determination of correct vs incorrect decisions. Auditing is more process oriented than results oriented, wherein the quality of work is judged not by results, but by traces of process to be found in the work papers. Moreover, auditors learn acceptance of processes that may diverge signifi-
167