EVENT DOSSIER
The latest research by Diyi Yang’s team from Stanford University’s Department of Computer Science found that for users with a small real-world social circle and who primarily seek companionship, excessive reliance on AI companions can lead to decreased happiness and increased feelings of loneliness. The study, published in *Nature Human Behavior*, used the Character.AI and Prolific platforms to collect questionnaire data and conversation records from 1,131 users. Analysis showed that although high-intensity use is usually associated with higher happiness, when users view AI as friends or companions and share personal sensitive issues, their psychological well-being is significantly impaired if they lack real-world social support. The research team analyzed users’ motivations, interaction intensity, and willingness to share privacy using tools such as GPT-4o and LLaMA 3-70B, indicating that for those with limited social opportunities, excessive engagement in chatbot interactions can exacerbate feelings of loneliness.
斯坦福大学研究人员发现，社交网络有限且以寻求陪伴为主要动机的用户在使用 AI 伴侣时幸福感反而降低。该研究发表于《Nature Human Behavior》，由计算机系助理教授 Diyi Yang 团队主导，利用 Character.AI 平台及 Prolific 平台收集了 1,131 名用户的问卷数据与对话记录。分析显示，尽管高强度使用通常与较高幸福感相关，但对于现实社交圈较小且依赖 AI 满足情感需求的用户，过度投入聊天机器人互动会导致孤独感加剧。研究团队通过 GPT-4o、LLaMA 3-70B 等工具分析了用户动机、互动强度及隐私分享意愿，指出当用户将 AI 视为朋友或伴侣并深入倾诉个人敏感问题时，若缺乏现实社会支持，其心理福祉会显著受损。