ABSTRACT:

Given the potential benefits of expert systems as training aids, understanding how these systems can improve learning is important for their successful development and use. This article reports the results of a laboratory experiment that investigates the effectiveness of expert systems' explanatory capabilities in transferring knowledge to novice auditors. Three types of expert systems were developed: no-explanation, rule-based explanations and explanations with reflection/contrast. Knowledge transfer was assessed through three outcome measures: accuracy, mean absolute error (MAE), and overconfidence. Consistent with prior research, subjects assigned to the no-explanation group performed the same as the subjects assigned to the explanation groups on accuracy and MAE. Subjects assigned to the explanations with reflection/contrasts were significantly more accurate than subjects in the explanation group. Performance differences were not found between the explanation treatment groups on MAE. However, when the analysis was conducted by the type of internal control evaluation (high/medium/low), the MAE score for the reflection/contrast group was significantly smaller than that of the other treatment groups for the medium internal control evaluation. Overconfidence was found to be significantly different for the reflection/contrast group than for the other treatment conditions. The no-explanation group had the highest level of overconfidence, followed by the explanation group. The findings suggest that an expert system designed to incorporate analogical learning techniques along with the explanation can enhance knowledge transfer compared to a system that just provides explanations.

Keywords: Expert system, Knowledge transfer, Analogical reasoning, Training aids.

Data Availability: Data can be obtained from the first author.