Proposing the throughput model as a potential cognitive-behavioral framework for employing explainable AI/ML technologies for fraud risk assessments: an exploratory empirical investigation

Purpose This study proposes the Throughput Model as a potential cognitive-behavioral framework for applying explainable AI/ML technologies in fraud risk assessments. Through an exploratory empirical investigation, the study also examines the effectiveness of applying the Throughput Model to fraud risk assessments in auditing by focusing on auditors' cognitive throughput. Design/methodology/approach An exploratory study was conducted with 42 practicing auditors who evaluated two hypothetical audit cases. Participants completed structured questionnaires designed to mirror the I P J D cognitive throughput for decision-making. Structural Equation Modelling (SEM) and regression analyses were used to test relationships and evaluate differences across auditors with high and low levels of professional skepticism. Findings The results indicate a significant positive influence of perception on judgment and of judgment on decision, confirming the effectiveness of the I→P→J→D cognitive throughput for decision-making. Also, auditors with high skepticism responded more to incentive and opportunity fraud risks, while low-skepticism auditors were sensitive to all three components of the fraud triangle. These findings provide a theoretical foundation and preliminary behavioral evidence for employing the Throughput Model as a potential cognitive-behavioral framework for explainable AI/ML in auditing. Originality/value This study is among the first to apply the Throughput Model as a foundational design for explainable AI/ML in auditing. It proposes a cognitive-behavioral framework for employing explainable AI/ML applications for fraud risk assessments. It contributes to both auditing and information systems research in the context of explainable AI/ML. It also offers researchers a transparent, theoretically grounded framework for modelling decision-making behavior in other high-stakes contexts.