Does the Favorite-Anime-Character-Imitating Chatbot Support a User’s Self-disclosure?
摘要
This study aims to support people’s self-disclosure by using a chatbot that imitates animated characters. Although self-disclosure is required during important life events, people tend to have few opportunities to express their own thoughts. In this paper, we implemented a chatbot that could imitate different actual anime characters using a large language model (LLM). By having users take on conversations with their favorite characters, we observed its effect in supporting users’ self-disclosure. We prepared 11 anime characters’ setting prompts and provided a user interface to chat with the chatbot. In evaluation experiments, we compared the support effect of favorite and unfavorite characters on users’ self-disclosure. Experimental results showed that the acceptance rate of the chatbot’s given hints increased if the participants chose their favorite anime characters. We found that chatting with a favorite anime character chatbot could support participants’ self-disclosure more effectively.