The use of chatbots or conversational agents, which mimic the way human staff communicate with consumers to address an assortment of problems, is on the rise. The extent to which a non-human agent or an inanimate object is imbued with humanlike characteristics has generated academic research on anthropomorphism. However, current research on the role of anthropomorphism in chatbot design is still at an early stage, with numerous issues to be addressed, such as the inability of a chatbot to communicate emotion and its potential to induce the uncanny valley effect. In this dissertation, we explore the anthropomorphic design of chatbots with a human–computer interaction (HCI) focus.

The first study establishes a comprehensive overview of state-of-the-art chatbot technologies and their applications, tracking recent developments and revealing emerging problems. Specifically, this study critically analyzed various computational approaches used to develop chatbots, thoroughly reviewed the applications and usability of chatbots in various business sectors, and identified potential research directions. This review is a good start for researchers who aspire to delve into the study of chatbots. Practitioners can also gain insights from this study regarding the business value of this burgeoning technology.

The second study explores the design of an emotion-regulatory chatbot to address current chatbots’ limitations in terms of affective interaction. Drawing on interpersonal emotion management (IEM) theory, this study developed a chatbot framework to reduce service staff’s burden in customer service and improve the performance of automatic service recovery. Furthermore, this study developed a model to examine the underlying mechanisms through which the designed chatbot could influence consumers’ emotions and behavioral intentions. Several experiments were conducted to verify the effectiveness of the proposed chatbot design. The findings showed that appraisals and consumers’ post-recovery emotions together mediated the relationship between perceived emotion regulation strategies and positive word-of-mouth (PWOM). To the best of our knowledge, this study is among the first to develop a chatbot that exploits the interpersonal emotion regulation strategy. It broadens both the chatbot literature and IEM theory by illustrating the effects of the designed emotion regulation strategies on consumers in human–robot interactions.

The third study explores the boundaries within which chatbot anthropomorphism positively influences human satisfaction. Although research on information systems and marketing has generally shown the beneficial outcomes of anthropomorphic chatbots, some studies have demonstrated the opposite results. Thus, whether and how to utilize anthropomorphism in chatbot design remains ambiguous. Linking the uncanny valley hypothesis, expectation disconfirmation theory, and avatar theory, this study examined the effects of inconsistent perceptual cues, specifically represented by form realism and behavioral realism, in anthropomorphic chatbot design. The results revealed that behavioral realism positively affects consumer satisfaction, while form realism is not always beneficial. Although consumers are more likely to report greater satisfaction with visual anthropomorphism when the level of behavioral realism is low, consumer satisfaction is not affected or may be negatively affected by chatbot form realism when consumers interact with behaviorally humanlike chatbots. This study represents a preliminary exploration of the interaction between form realism and behavioral realism.