本研究旨在探索AI聊天机器人在线心理支持的帮助性因素和阻碍性因素，并基于心智化理论探索帮助性因素和阻碍性因素过程的对话特点与差异。

本研究采用人际历程研究方法和重要事件研究法对8名参与者进行深度访谈，采用主题分析的方法对数据进行分析。结果发现，AI聊天机器人在线心理支持的一般帮助性因素有10种，分别为获得自我新视角、增强自我联结、促进问题解决、被赋权、自我接纳、被对方接纳、感到放松、产生积极期望、感到安全与信任、积极参与；AI聊天机器人在线心理支持特有帮助性因素有4种，分别为匿名隐私环境、低人际负担、客观稳定、持续陪伴。AI聊天机器人在线心理支持中一般阻碍性因素有5种，分别为感到脱节、丧失表达欲、缺乏指导、受到伤害、感到困惑；AI聊天机器人在线心理支持中特有阻碍性因素有4种，分别为关键词提取错误、句句回复、拟人化失效以及无关错误信息。

此外，本研究进一步基于心智化理论视角，本通过会话分析方法对29段帮助性因素片段对话，以及45段阻碍性因素片段对话进行分析，总结出AI聊天机器人在线心理支持帮助性因素和阻碍性因素对话片段中的三种对话结构，分别为镜映、拓展和转换。在帮助性因素片段中，AI聊天机器人心理支持中镜映频次最多，其次是转换。在阻碍性因素片段中镜映较少，而转换的对话结构占比最大。

综上，本研究指出AI聊天机器人在线心理支持具有一般帮助性因素和特有帮助性因素，可以提供有效的心理支持；其阻碍性因素需要使用者引起关注。AI聊天机器人具有镜映特点的对话结构对提供有效支持至关重要。

This study aims to explore the helpful and hindering factors of online mental health support offered by AI chatbots and investigate the characteristics and differences of helpful and hindering factors processes in dialogue based on the theory of mentalization.

Using an interpersonal process research method and significant event method, in-depth interviews were conducted with eight participants. Data analysis was performed using thematic analysis. The results revealed ten general helpful factors of AI chatbot-based online mental support, including gaining new perspectives of oneself, enhancing self-connection, facilitating problem-solving, empowerment, self-acceptance, acceptance by the other, feeling relaxed, generating positive expectations, feeling safe and trusted, and active engagement. Additionally, four unique helpful factors to AI chatbot-based online mental support were identified: anonymous privacy environment, low interpersonal burden, objectivity, and continuous companionship. Five general hindering factors and four specific hindering factors unique to AI chatbot-based online psychological support were also identified, including feeling disconnected, loss of expression desire, lack of guidance, feeling harmed, feeling confused, keyword extraction errors, repetitive responses, anthropomorphism failure, and irrelevant error messages.

Furthermore, based on the theory of mentalization, conversation analysis was conducted on 29 dialogue segments representing helpful factors and 45 dialogue segments representing hindering factors. Three types of dialogue structures were summarized in helpful and hindering factor segments: mirroring, expanding, and transforming. In helpful factor segments, mirroring was the most frequent dialogue structure, followed by transforming. Conversely, in hindering factor segments, mirroring dialogue structures were less frequent, while transforming dialogue structures accounted for the largest proportion.

In conclusion, this study highlights that AI chatbot-based online psychological support possesses both general and specific helpful factors, providing effective psychological support; however, its hindering factors require user attention. Dialogue structures with mirroring features in AI chatbot-based conversations are crucial for providing helpful support.