Social (Ir)Responsibility of Artificial Intelligence? Consumer Perceptions of AI-Induced Errors and Implications for Society
(A2023-114272)

AUTHORS

Alexander Mueller, University of Mannheim; Sabine Kuester, University of Mannheim; Sergej von Janda, Kaiser X Labs

ABSTRACT

Artificial intelligence (AI) still commonly errs in practice. We explore consumer responses to different error types (i.e., technical errors resulting from a technological disruption of algorithmic processes and social errors, representing social norm violations) and error severity. This distinction is crucial since our research shows different response patterns depending on error type and severity. Errors can even evoke positive reactions, as described by the pratfall effect. Four studies show that severe errors, regardless of error type, produce negative responses from consumers, while minor social errors lead to significantly fewer negative consumer responses than minor technical errors. Our results highlight the importance of AI’s adherence to social norms. We note that AI’s minor social errors could foster the stigmatization of minorities and socially disadvantaged groups, suggesting the necessity of implementing additional safeguards against social norm violations by AI.