Using an Interactive Artificial Intelligence System to Augment Auditor Judgment in a Complex Task

Abstract

Emerging research highlights the benefits and challenges of advanced technologies in auditing. However, research examining how these technologies impact auditor judgment is limited, particularly when performing complex tasks. Thus, this study experimentally examines whether using an artificial intelligence system with interactive and structured information processing features augments auditor judgment when performing a fraud risk assessment. Prior research reveals fraud risk assessments are challenging for auditors, causing susceptibility to judgment biases. Auditors’ exposure to increased and complex information due to intensified digitization by firms likely results in greater susceptibility to judgment bias. While prior research report mixed results regarding the use of decision aids to mitigate judgment biases, we posit that interactive and structured information processing features will reduce cognitive load resulting in enhanced auditor judgment. As such, auditors will make more appropriate fraud risk assessments. Experimental results are consistent with our expectations. Specifically, after using the artificial intelligence system with interactive and structured information processing features, auditors disregarded irrelevant information and made fraud risk assessments consistent with an expert panel of experienced auditors. We further investigate experimental results with a verbal protocol analysis using a subset of the participants and find auditors interacting with the AI system demonstrate information processing strategies consistent with a reduction in cognitive load and effort.

Keywords: auditor judgment bias, artificial intelligence, augmented, verbal protocol analysis, fraud risk, cognitive load theory