Can Using Artificial Intelligence in an Audit Reduce Auditor’s Liability?
Presentation
Overview
Overview
Description
Big-4 firms have begun making extensive investments in developing audit programs based on artificial intelligence (AI). However, the legal implications of using AI in the case of an audit failure are not heavily explored. In this study, we investigate whether and how using AI to partially complete an audit, compared to having human auditors complete the same work, may affect auditor negligence verdicts by jurors in case of an audit failure. We also investigate whether and how the work being outsourced or insourced interacts with the use of AI. The similarity-leniency hypothesis posits that the more a juror associates with the defendant, the more lenient the juror evaluation will be. However, this leniency depends upon the proximity of the wrongdoing to the defendant and/or the intensity of the wrongdoing, referred to as the “black-sheep effect”. Consistent with the similarity-leniency hypothesis, we find that, in an audit in which the misjudgment is attributable to human staff (rather than AI), jurors evaluate characteristics of the defendant more favorably and, subsequently, attribute the reasons for the audit failure for more circumstantial rather than personal reasons. Consequently, jurors assess more lenient verdicts. However, consistent with the black-sheep effect, the leniency is evident only when the work is outsourced (perceived to be less controllable by the audit firm) but not when it is insourced (more controllable by the audit firm).
