“Our AI might have failed, but did better than humans”: Using the better-than-humans response after discriminating AI actions
(A2022-107212)

AUTHORS

Andrea Weihrauch, University of Amsterdam; Chunya Xie, Renmin University of China

ABSTRACT

The use of artificial intelligence (AI) in companies’ activities is becoming more and more common, and while AI technology progresses, current media coverage provides a good overview for incidents where AI fails. One of the most reputationally hurtful AI failures is related to AI being biased or discriminating (i.e., Google’s AI showing prestigious jobs to men, but not to women). Often, when such discriminating AI failure occurs, companies opt to acknowledge the failure, but at the same time clarify that the AI still performed better than the system of human decision making previously in place. In one pilot study, and five experimental studies, we examine the efficacy of such a response strategy against other common response strategies (i.e., a simple apology), and provide evidence that the “better-than-response” is less successful in restoring brand reputation, especially since the majority of consumers are fearful of AI being overly powerful/autonomous.