To Err is Bot, Not Human: Asymmetric Reactions to Chatbot Service Failures

The reaction of individuals to AI failures is a complex and controversial topic that remains widely debated. Existing research on this issue is fragmented and inconclusive, with multiple perspectives and conflicting findings. To further explore this issue, we conducted a study examining how customers respond to service failures committed by chatbots and human employees. We used an online vignette-based method to assess participants’ emotional, cognitive, and behavioral intentions following such failures. A total of 307 people participated in this study and the results indicated that participants were more tolerant of chatbot failures. They exhibited fewer negative emotions and were more likely to forgive erring chatbots. These results were moderated by service failure type. This study provides a more comprehensive understanding of how people respond to AI errors in service context, offering a more nuanced perspective than previously available in the literature.