The application of case-based reasoning to assist accountants in identifying top management fraud: a study of the problem domain and the methodological issues in the development, implementation and evaluation of a case-based learning and reasoning tool

Although most audit practices have developed many auditing and accounting expert systems, research and progress in this area was relatively slow until the second half of the 80s. One of the reasons for this slow progress was the difficulty of assessing the value added by these systems. Some rule-based applications in audit have been trying to tackle unstructured and dynamic problems. This has caused many design problems-such as knowledge elicitation-and maintenance. The assessment and management of risk may be a topic that exemplifies the need for a non-traditional approach to knowledge management. Research in managerial decision-making has shown that, when faced with a complex problem, an expert will often look to analogous problems for possible solutions. Top management fraud (TMF) is one of those fields for which there are neither underlying models nor formal ways of detecting them. This paper is concerned both with the methodological issues of using case based reasoning for unstructured domains-such as TMF-and with the evaluation process of such a model. It is argued that emphasis should be given to the searching-learning loop, a prerequisite for case-based learning and reasoning systems (CBLR).