Ben Commerford, Arthur Andersen Associate Professor recently presented his paper, "'Alexa, audit loan grades!': Does humanizing artificial intelligence enhance auditor reliance?" at Colorado State University on September 10th.
ABSTRACT: Audit firms are making substantial investments in AI with the hope that these systems, like human specialists, will provide auditors with evidence that can improve audit outcomes. However, even the most reliable AI systems will not be perfect, and auditors will inevitably observe these systems make errors. We experimentally demonstrate that auditors more heavily discount evidence from an AI system (versus a human specialist) after observing such an error. We also predict and find that humanizing an AI system mitigates the effects of this “algorithm aversion” (i.e., the tendency to discount computer-based advice more heavily than identical human advice) on auditors’ judgments. Our findings suggest that auditors are willing to rely on an AI system, so long as they do not encounter any errors by the system. Additionally, humanizing features appear to invoke human social norms (e.g., forgiveness) that facilitate auditors’ continued reliance on these systems, even after they err.