Reporter：Wang Weiguang
Abstract：The advancement of conversational AI techniques is driving the rise of chatbot applications in various fields. Pioneering business studies have examined the effects of functional chatbots for specific business tasks, such as customer service. This study extends the business literature by focusing on a new type of chatbot, the social chatbot. We develop a GPT-based social chatbot for fitness-related topics and implement it on Twitter. Specifically, this field experiment identifies a strong negative impact of the chatbot’s identity on its ability to engage social media users. We further explore the mechanisms of a reduction in perceived socialization value caused by chatbot identity disclosure by diving into a commonly studied factor, gender matching. The chatbot presents as female, whose perceived socialization value is higher for male social media users. Correspondingly, male users are likely to experience a stronger reduction in perceived socialization value if the account is disclosed as a chatbot. Moreover, we investigate the role of gender cues in the chatbot’s textual replies. While the chatbot is constructed as a female, male cues could potentially present gender inconsistency, which we find to be further detrimental to user engagement after chatbot identity is disclosed.
