ABSTRACT
Simon [1960] classified tasks as either programmable or non-programmable. Management scientists interpreted Simon's classification as structured or unstructured. Based on this classification, decision aid theorists have developed a contingency model between task structure and decision aid type where automation is suggested for structured tasks, decision support systems for semi-structured tasks, and knowledge-based expert systems for unstructured tasks. This paper argues and provides evidence from auditing that the programmable/non-programmable classification does not map onto the task structure/decision aid contingency model. Specifically, (1) tasks can be classified as programmable or non-programmable regardless of their degree of task structure; (2) only for the tasks classified as programmable, will there be a contingency between task structure and decision aid type.
One hundred and thirty four auditors from seven international accounting firms evaluated the appropriate type of decision support in the context of risk assessment where decision aids appear to have a particularly high potential for assisting auditors in professional judgments. The results provide support for the hypothesized task-structure and decision-aid-type relationship for eight of the 30 risk assessment tasks that were classified as programmable. The remaining 22 tasks were those that at least half of the subjects judged to be strictly subject to human processing regardless of their assessed level of task structure (i.e., they were judged non-programmable). The audit methodology of the accounting firms, rank of respondents, audit speciality of subjects (including electronic data processing and/or high technology audit specialities) had no effect on the overall conclusions, providing further support for the robustness of the hypothesized relationships.