Abstract Online interactions between companies and customers increasingly occur via conversational agents (CAs) (e.g., chatbots), which are commonly designed to appear humanlike. At the same time, customers are frequently guided to choose healthier or more sustainable options, which can elicit negative moral emotions (e.g., guilt). This leads to the question of how humanlike CAs influence the formation of negative emotions when users are guided. Based on the Computers-are-Social-Actors theory and Norm Activation Theory, it is theorized that humanlike CAs are viewed by users as social agents that are aware of social norms, catalyzing the formation of negative moral emotions. Two experiments suggest that a humanlike CA can induce negative moral emotions, whereas traditional point-and-click interfaces cannot. Similarly, the Macbeth effect (i.e., the act of cleansing decreases negative moral emotions) can be observed for CAs: the depiction of handwashing reduced negative moral emotions in CA users. For theory, insights are provided into the perception of humanlike CAs in the context of user guidance. For practice, the results suggest that companies should be careful when implementing CAs to guide users.
Mücksch et al. (Tue,) studied this question.