Control Issues: How Providing Input Affects Auditors’ Reliance on Artificial Intelligence

• In Event: Plenary Session

Abstract

In this study, we examine auditors’ reliance on AI-based specialist systems in the audit of complex estimates. We predict that auditors rely more readily on recommendations from human specialists, relative to AI-based specialist systems, and that giving auditors a small amount of input into specialists’ process can mitigate their aversion to AI. In an experiment with highly experienced auditors, we find evidence of aversion to AI, in line with our predictions. We further find that auditors’ locus of control, or the extent to which they feel they control outcomes in their work environment, moderates our predicted interaction. For Internals, who believe they are in control of their outcomes, providing some control does not appear to mitigate their aversion to relying on recommendations from AI-based systems. For Externals, who believe they have little control over their outcomes, providing control increases their reliance on AI-based systems, as expected. Audit firms continue to develop advanced AI that can improve the audit of complex estimates, and this work sheds additional light on auditors’ aversion to use of AI systems.

Authors

• Benjamin Commerford, University of Kentucky

• Aasmund Eilifsen, Norwegian School of Economics

• Richard Hatfield, University of Alabama-Tuscaloosa

• Kathryn Holmstrom, Iowa State University

• Finn Kinserdal, NHH Norwegian School of Economics