A constructive synthesis approach to a knowledge-based internal control evaluation system design

摘要

In knowledge engineering interviews with experienced auditors, it was observed that the auditor's review processes in evaluating the client's internal control system can be characterized by the following: using selective attention, forward chaining with a hypothesis-test as a back-up, and multiple knowledge sources. Based on these observations and cognitive science theories, a computational model, called the red-flagging model, was developed and implemented in a microcomputer environment.In the red-flagging model, a problem solver pays attention to relevant information and users a blackboard architecture for allowing multiple knowledge sources to communicate with each other. The computerized internal control evaluation review system, ICE1, built on the red-flagging model, processes more than 400 checklist question items and incrementally constructs review conclusions from relevant information. In a performance test against 29 real audit cases (correspond to 232 cases in previous research), the review system showed more than 80% compatibility with experienced auditor's conclusions.