AI 챗봇 사용 목적에 따른 복구 전략의 신뢰도 차이‑ 도구적 목적과 정서적 목적을 중심으로 = Differences in Trust in Repair Strategies According to the Purpose of AI Chatbot Use - Focusing on Utilitarian and Emotional Purposes

다국어 초록 (Multilingual Abstract)

Background: Large Language Model(LLM)-based AI chatbots are used in a wide range of situations. However, inaccurate or contextually inappropriate esponses can undermine user trust. In particular, AI chatbots are used not only for utilitarian purposes, such as information seeking and task completion, but also for emotional purposes, such as discussing personal concerns and seeking emotional support. As a result, the type of response users expect in error situations may differ depending on the purpose of use. Accordingly, this study aimed to classify chatbot use into utilitarian and emotional purposes and to explore repair strategies that enhance user trust for each purpose.
Methods: This study first reviewed the literature on the usage purposes of AI chatbots and their repair strategies. Next, scenario-based online surveys were developed for utilitarian and emotional purposes, and items for measuring trust were collected from prior studies. Finally, an online survey was conducted to measure users' trust in repair strategies across each purpose, and the results were compared and analyzed.
Results: The findings showed that, for utilitarian purposes, solve and information strategies that provide users with concrete guidance on what to do next received higher trust. For emotional purposes, social and disclosure strategies that respond in a friendly and honest manner received higher trust. In addition, repair strategies that received high trust for one purpose showed relatively lower trust for the other, confirming that effective repair strategies may differ depending on the purpose of use.
Conclusion: The findings suggest that chatbot repair strategies do not function in the same way across all situations and that the most effective strategy varies depending on the purpose of use. For utilitarian purposes, strategies that support problem solving and guide users’ next actions generated higher trust, whereas for emotional purposes, strategies that express friendliness and honesty generated higher trust. This implies that chatbot error responses should not be designed in a one-size-fits-all manner, but rather should be tailored to the purpose of use and the role users expect the chatbot to play.