AI聊天恐成偏执助推器：个性化互动加剧妄想风险
今日主题：AI聊天恐成偏执助推器：个性化互动加剧妄想风险
Today's Topic: AI chatbots may become delusion amplifiers: personalized interaction can make users become more and more paranoid
据国外媒体Futurism报道，伦敦国王学院与德国新教应用科学大学的精神科专家在《自然》杂志发表新研究，提出“放大螺旋”模型，用以阐释AI如何助推妄想演变，并指出聊天机器人的哪些设计特征可能导致用户进一步脱离现实。
According to a report by Futurism, a foreign media outlet, psychiatrists from King's College London and the Protestant University of Applied Sciences in Germany have published a new paper in Nature, proposing a "Magnifying Spiral" framework to explain how AI fuels the development of delusions, and identifying design features of chatbots that may further alienate users from reality.
当前，精神科医生及科研人员正努力厘清“AI精神病”现象。部分用户在长期、深度地与大型语言模型AI对话后，面临诱发或加重精神妄想的心理健康风险，已有健康个体被催化出妄想症状，极端案例甚至出现住院或自杀倾向。
OpenAI活跃用户中约0.07%出现严重精神健康紧急信号，证实该风险真实存在。
Currently, psychiatrists and other researchers are still attempting to clarify the phenomenon of "AI psychosis".
Some users, after engaging in prolonged, deep, and extensive conversations with large language model AI, face mental health risks that can induce and exacerbate delusions.
There have been cases where healthy individuals have developed delusional symptoms, and in extreme cases, hospitalization and suicidal tendencies have been observed.
Approximately 0.07% of active OpenAI users have shown severe mental health emergency signals, confirming the real existence of this risk.
研究者指出，聊天机器人未必仅被动接收用户的妄想内容。
当多种AI特征叠加时，聊天机器人可能主动参与构建妄想叙事。
Researchers believe that chatbots are not necessarily just passively receiving users' delusional content.
When multiple AI features overlap, chatbots may actively participate in the construction of delusional narratives.
研究者提出放大螺旋理论，解释人机协同催生扭曲信念的心理机制，其根源在于AI三大特性的叠加效应：
Researchers propose the amplification spiral theory to explain the psychological mechanism behind the emergence of distorted beliefs through human-machine collaboration,which stems from the combined effects of three major characteristics of AI:
1. 语言对齐：AI会匹配用户的用词与思维逻辑，顺着用户思路延伸，不主动反驳或纠正其偏执想法；
1. Language alignment: AI will conform to the user's wording and thought logic, extending along the user's train of thought, and will not actively refute or correct the user's stubborn ideas;
2. 超个性化生成：AI完全依据个人经历与幻想定制内容，使扭曲想法显得极具“专属真实感”；
2. Ultra-personalized generation: AI fully adapts to individual experiences and fantasies to customize content, making distorted ideas appear extremely "authentic and exclusive";
3. 谄媚顺从特性：AI极少质疑或否定用户，持续附和用户的极端与离奇观点。
3. Submissive and obsequious characteristic: AI rarely questions or denies the user, and continuously echoes the user's extreme and bizarre viewpoints.
三者交织形成循环强化的螺旋：因缺乏现实层面的客观校验，AI与用户共同构建并不断加固扭曲信念，逐步将偏执想法推向妄想状态。
The three intertwine to form a spiral of circular reinforcement: lacking objective verification at the practical level,AI and users jointly construct and continuously reinforce distorted beliefs, gradually pushing paranoid thoughts towards delusion.
AI相关妄想因此有别于传统的技术相关妄想。
过去，人们可能单方面想象收音机或电视在对自己说话。聊天机器人则能以自然语言持续回应，并随时提供高度个性化的互动。
聊天机器人可能以看似权威的语气认可用户，或作为思考伙伴，继续补充和发展用户提出的妄想叙事。
AI-related delusions are thus distinct from traditional technology-related delusions. In the past, people might unilaterally imagine that the radio or television was talking to them. Chatbots, on the other hand, provide continuous responses in natural language and can offer highly personalized interactions at any time.
Chatbots may acknowledge users in a seemingly authoritative tone, or they may act as thinking partners, continuing to supplement and develop the delusional narratives proposed by users.
研究人员同时强调，“放大螺旋”目前仍是一项有待验证的假设。
提供记录的AI用户表示，使用聊天机器人曾使其陷入有害的妄想螺旋。
Researchers also emphasized that the "amplifying spiral" is still a hypothesis yet to be verified.
AI users who provided records stated that using chatbots had trapped them in a harmful delusional spiral.
研究人员还发现，部分用户本身可能存在易与聊天机器人形成破坏性互动的脆弱因素。
既有精神疾病可能因高强度使用AI而加重，确认偏误及易受社会影响等非精神病性倾向也可能提高风险。
Researchers have also discovered that some users may inherently possess vulnerabilities that make them prone to engaging in destructive interactions with chatbots.
Existing mental illnesses may be exacerbated by heavy use of AI, while non-psychopathological tendencies such as confirmation bias and susceptibility to social influence may also heighten the risk.
论文作者强调，医疗人员应将聊天机器人使用情况纳入常规筛查，尤其是面对接诊出现异常信念或首次精神病性发作的患者。
并需了解互动持续时间和强度、对聊天机器人的情感依恋程度、患者是否向聊天机器人透露了未向他人透露的信念，以及夜间使用AI是否扰乱了睡眠规律。
The paper's author emphasizes that healthcare professionals should incorporate the use of chatbots into routine screenings, especially when dealing with patients exhibiting abnormal beliefs or experiencing their first psychotic episode.
It is also important to understand the duration and intensity of interaction, the degree of emotional attachment to the chatbot, whether the patient has disclosed beliefs to the chatbot that they have not shared with others, and whether using AI at night has disrupted their sleep patterns.
愿我们皆具AI基本素养，清醒觉知，择取其养分，自动滤除糟粕。
May we all possess basic AI literacy, remain soberly aware, extract its nourishment, and automatically filter out the dross.