Active expectation management in chatbot conversations

Abstract

Chatbots have gained strong popularity in customer service, although users regularly experience unsatisfactory interactions and service failures, often due to highly exaggerated performance expectations. As a viable option, firms might therefore consider using ‘active expectations management’ by describing chatbot limitations. However, the question remains whether this strategy has a positive or negative impact on customer satisfaction and reuse intentions. Drawing on expectancy violation theory and the computers are social actors (CASA) paradigm, we empirically examine the effects of different expectation management strategies on user satisfaction and reuse intention. The results of a between-subjects experiment (n = 346) demonstrate that expectation management is an effective strategy to at least partly recover the failure. Furthermore, we show that different message types (i.e., ‘adapt’ vs. ‘understand’) have differential effects, whereas message positioning was found to be irrelevant. Our results enrich the service and chatbot literature and give managerial guidance for successful chatbot design.