Informational or emotional? Exploring the relative effects of chatbots’ self-recovery strategies on consumer satisfaction

The rapid development of artificial intelligence technology has enabled chatbots to perform human-like service work. Chatbots commonly fail in service due to their limitations. However, previous research has simply focused on the difference between self-recovery and human recovery, and the details of self-recovery have largely been overlooked. This study attempts to classify chatbots' self-recovery strategies and elucidate their relative impact on consumer satisfaction. Drawing on social support theory, this study proposes two types of self-recovery strategies, informational self-recovery and emotional self-recovery. Meanwhile, this study further details consumer satisfaction from quality and attitude. The results of three scenario-based experiments reveal that informational self-recovery has a more substantial effect on consumer quality satisfaction than emotional self-recovery, while emotional self-recovery has a greater effect on consumer attitude satisfaction than informational self-recovery. Further, the anthropomorphic level of chatbots' self-recovery also positively moderates the relative relationship. The underlying mechanisms suggest that perceived competence partially mediates the relationship between informational self-recovery and consumer quality satisfaction and that perceived warmth partially mediates the relationship between emotional self-recovery and consumer attitude satisfaction. In addition, this study implies that an informational self-recovery strategy is well-suited to the context of service process failure, whereas an emotional self-recovery strategy is better adopted in the service outcome failure context. This study provides new insights into chatbots’ self-recovery strategies in response to service failure and offers implications to service providers of chatbots to leverage different self-recovery strategies to improve consumer satisfaction.