City Research Online

Abstract

This study presents the results of an investigation into the methods of auditing Expert Systems. Such systems have already proved to be, and are increasingly becoming, a very powerful tool in many areas such as medicine, geology, finance and banking. They embody unique risks which are not treated by conventional audit methods of operating or developing software. The lack of awareness and information about Expert Systems in general and their auditability in particular are somewhat surprising.

The author, in tackling this new area, has developed and proposed two models of auditing Expert Systems: a) the Audit of an Operating Expert System (AOES), b) the Audit of an Expert System under development (AESD). The first model incorporates the" control band" which aims at eliminating the exceptional risks and to allow the internal auditor to treat it as conventional software. The second proposed model is based on "NESDEM'; a normative evaluation model for Expert Systems.

The test of the proposed AOES model was conducted in two different organisations: ARJO-WIGGINS APPLETON which developed and still uses an Expert System for its paper mill and the CITY UNIVERSITY DEPARTMENT OF OPTOMETRY AND VISUAL SCIENCE which developed an Expert System for eye tests.

Unfortunately the author was unable to test his proposed AESD model under a "live" development process due to lack of cooperation from organisations which the author contacted. Consequently he tested this model by mailing questionnaires to internal/external auditors within the U.K.

Given the research performed in this study and subject to the limitations detailed, the proposed models appear reliable, flexible, practical and suitable to the internal auditor in assessing the effectiveness of the internal controls within Expert Systems.

Publication Type:	Thesis (Doctoral)
Subjects:	H Social Sciences > HF Commerce H Social Sciences > HG Finance
Departments:	Bayes Business School Doctoral Theses Bayes Business School > Bayes Business School Doctoral Theses

Export

Downloads