Auditor experience in preliminary control risk assessments : neural network models of auditor's knowledge structures

Auditing pronouncements typically cite auditor judgment as an integral part of the audit process. Research in expertise suggests that increased experience in a specific task increases accuracy and efficiency of a decision maker's judgment. Contrary to these findings, the results from most internal control judgment research have not found that experience influences auditor judgment. However, none of these studies included all three integral elements of the decision process-selection of cues from the set of potential cues, cue processing and decision response—into one experiment or model. This study examines how experience influences auditors' abilities to select and use relevant cues to make efficient, appropriate preliminary control risk assessments (CRAs). Using the preliminary CRAs of two auditor groups with differing levels of experience, this study developed two neural network models of auditors' CRA decision processes. The models provided a means of capturing and measuring auditors' selective attention, a cognitive concept describing a decision process that selects and uses cues based on cue dimensions, the different qualities or aspects of a cue perceived to provide a contrast in level of significance for making categorical decisions. The results showed that the typical experienced senior auditor exhibited a higher level of selective attention to cue dimensions. The experienced group also used fewer cues and exhibited less decision time than did the less experienced group. The results are consistent with the conclusions in the expertise literature (e.g. Anderson, 1983; Johnson, et al. 1984; Dreyfus & Dreyfus, 1986) that experience does lead to a more top-down approach to cue selection and a more highly developed knowledge structure that provides more information content for relating the selected cues to the resulting decision response. However, decision error in relation to the firm's CRA solution was nearly the same for both groups and suggested that both groups could improve their CRA process. In order for the inexperienced seniors to adjust their control risk assessment to match the firm's solution, they need to use the information provided by their selected cue set better. Although the experienced seniors' CRAs were more consistent with their cue sets than the inexperienced seniors' CRAs were with their chosen cue sets, in terms of increasing consensus to the firm's prescribed solution, the experienced group may need to consider planning to rely more on internal controls and less on tests of account balances for this case that exhibited a fairly strong internal control structure.