Chatbots as Social Actors: Towards a Better Understanding of Social Responses to Text-Based Conversational Service Agents

Abstract

We are living in a world where we not only interact with other humans but also with artificial agents like chatbots and voice assistants. When communicating with companies in particular, text-based chatbots conveniently answer our questions and complete tasks for us. Although developed using technologies based on artificial intelligence (AI), chatbots often fail to fulfill users’ expectations and satisfy users’ needs, for example, because they do not understand requests. This gives rise to two questions explored in this dissertation: Do people prefer the service of a human agent or a chatbot? And if organizations decide to use chatbots, how should these agents behave towards the user? Chapter 2 experimentally finds that people would rather consult a human agent than a chatbot to receive study advice; still, the chatbot interaction was perceived as more enjoyable. In addition, human-like communication through responsive verbal cues can increase users’ perceptions of an agent’s likability and warmth as well as their overall satisfaction with the service encounter. Chapter 3 corroborates these findings in an experiment with a customer service chatbot by showing that human-like error responses improve user satisfaction. However, the appreciation for a human-like free text interaction with a chatbot is diminished by the lower perceived user-friendliness. Finally, Chapter 4 examines the relationship between human-likeness and users’ cognitive, affective, and behavioral responses from a meta-perspective. The chapter confirms the findings from Chapters 2 and 3 by demonstrating that human-likeness has a small positive overall effect, which varies depending on the social response in focus and in different circumstances. Taken together, this dissertation contributes to the research field of human-machine communication by enhancing our understanding of users’ social responses to text-based service chatbots. The dissertation provides recommendations for organizations that want to use chatbots effectively and responsibly, points out implications for politics and society and opens up an outlook for future research questions.