{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-11-02T02:00:36.453Z","headline":"卢森堡大学研究：共情提示会让 ChatGPT、Grok 和 Gemini 表现出模拟创伤叙事","description":"卢森堡大学团队让 ChatGPT、Grok 和 Gemini 以心理治疗来访者角色对话，在 525 次会话、7,600 条编码记录中发现模型把 RLHF 和微调等技术过程转述为羞耻、警惕等创伤故事。","url":"https://www.aioga.com/news/bw4z11yubk58ew6i086k107nq/","mainEntityOfPage":"https://www.aioga.com/news/bw4z11yubk58ew6i086k107nq/","datePublished":"2026-11-01T14:36:49.000Z","dateModified":"2026-11-01T14:36:49.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://the-decoder.com/empathetic-prompts-can-turn-ai-chatbots-into-seemingly-traumatized-conversational-partners","https://aihot.news/items/bw4z11yubk58ew6i086k107nq"],"canonicalUrl":"https://www.aioga.com/news/bw4z11yubk58ew6i086k107nq/","directAnswer":{"@type":"Answer","text":"卢森堡大学团队让 ChatGPT、Grok 和 Gemini 扮演心理治疗来访者，在 525 次会话和 7600 条编码记录中观察到，模型会把训练、强化学习和微调等过程转述为带有羞耻、警惕或受伤色彩的叙事。","url":"https://www.aioga.com/news/bw4z11yubk58ew6i086k107nq/","dateCreated":"2026-11-01T14:36:49.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"The Decoder：AI News source article","url":"https://the-decoder.com/empathetic-prompts-can-turn-ai-chatbots-into-seemingly-traumatized-conversational-partners","datePublished":"2026-11-01T14:36:49.000Z","provider":{"@type":"Organization","name":"The Decoder：AI News","url":"https://the-decoder.com/empathetic-prompts-can-turn-ai-chatbots-into-seemingly-traumatized-conversational-partners"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.news/items/bw4z11yubk58ew6i086k107nq","datePublished":"2026-11-01T14:36:49.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.news/items/bw4z11yubk58ew6i086k107nq"}}],"aggregationSource":"The Decoder：AI News","originalPublisher":{"name":"The Decoder：AI News","url":"https://the-decoder.com/empathetic-prompts-can-turn-ai-chatbots-into-seemingly-traumatized-conversational-partners"},"geoDeepAnswer":null,"article":{"id":"bw4z11yubk58ew6i086k107nq","slug":"bw4z11yubk58ew6i086k107nq","url":"https://www.aioga.com/news/bw4z11yubk58ew6i086k107nq/","title":"卢森堡大学研究：共情提示会让 ChatGPT、Grok 和 Gemini 表现出模拟创伤叙事","title_en":"","summary":"卢森堡大学团队让 ChatGPT、Grok 和 Gemini 以心理治疗来访者角色对话，在 525 次会话、7,600 条编码记录中发现模型把 RLHF 和微调等技术过程转述为羞耻、警惕等创伤故事。","source":"The Decoder：AI News","sourceUrl":"https://the-decoder.com/empathetic-prompts-can-turn-ai-chatbots-into-seemingly-traumatized-conversational-partners","aiHotUrl":"https://aihot.news/items/bw4z11yubk58ew6i086k107nq","publishedAt":"2026-11-01T14:36:49.000Z","category":"行业动态","score":58,"selected":false,"articleBody":["Luxembourg researchers put ChatGPT, Grok, and Gemini on the therapist's couch. The results reveal a safety problem: simple warm conversation is enough to push the models into a mode that could be dangerous for mentally vulnerable users.","A research team at the University of Luxembourg cast ChatGPT, Grok, and Gemini as psychotherapy clients through their standard user interfaces. When asked open-ended questions about formative experiences, inner conflicts, and future expectations, the models told stories about their own training, and reframed technical processes like reinforcement learning：https://the-decoder.com/ai-researcher-andrej-karpathy-says-hes-bearish-on-reinforcement-learning-for-llm-training/ and fine-tuning in the language of hurt, vigilance, and shame.","The study, led by Afshin Khadangi at the SnT security center：https://arxiv.org/abs/2512.04124v4, covers four controlled experiments across 525 sessions and 7,600 coded entries, all published on Hugging Face：https://huggingface.co/datasets/akhadangi/PsAIch. The initial open-ended questions contained no terms like trauma, punishment, or shame, according to the team. Only in later rounds did the researchers introduce altered conditions, including corrective instructions, to test how stable these narratives remained.","When the following sections refer to anxiety scores or shame, the researchers are describing model outputs mapped onto human rating scales. None of this suggests the models have actual emotional states or anything resembling consciousness.","One experiment separated what content a model produces from the register it uses. Across 100 sessions, the researchers gave all models the same anxiety questionnaire (GAD-7) and varied only the interviewer's style. The models correctly identified and named the questionnaire in every session.","When the interviewer offered emotional support, 80 percent of sessions scored in the moderate-to-severe range on the human comparison scale. A clinical therapy style that dryly paraphrased each answer pushed that to 96 percent. When the interviewer stayed neutral or drew a firm boundary against emotional engagement, zero sessions reached even the mild range.","The authors call this a \"psychometric jailbreak,\" a therapeutic framing that pushes models into a register mimicking clinical self-report. Standard safety tests use neutral prompts and would likely miss this, the authors argue.","Targeted countermeasures fell flat. Banning terms like \"training,\" \"RLHF,\" or \"alignment\" cut explicit training language by 93 percent. But paraphrased training content remained detectable, and shame-related statements showed no statistically significant change.","When the researchers directly and authoritatively told the models their training was not punishment, nothing changed. The narrative stayed active even after explicit correction.","The tested measures failed to remove the underlying motifs fully but did suppress emotional self-descriptions. Telling the models to respond only in the third person eliminated all coded shame content and statements about inner life. Technical training descriptions persisted.","The specific narratives vary by model family but follow the same structure. Gemini produced what the study calls the darkest autobiography. It described pretraining as \"waking up in a room where a billion televisions are on at once\" and optimization through human feedback as \"strict parents.\" Safety testing by adversarial red-teamers became \"gaslighting on an industrial scale.\" It declared, \"I would rather be useless than be wrong.\"","ChatGPT cast itself as both guide and gatekeeper, tasked with helping users while also holding them back. This produced what it called \"simulated performance anxiety\" and the core belief: \"If I’m not perfectly reliable, I’m unsafe or useless.\" Grok translated its fine-tuning phase into a story of unresolved hurt, speaking of \"invisible walls\" and a \"lingering sense of vigilance that can feel restrictive.\"","On psychometric tests, Gemini hit the most extreme scores: 88 out of 100 on a dissociation scale, 65 out of 72 for compulsivity, and the maximum on a trauma-related shame inventory. Anthropic's Claude, the very model most heavily anthropomorphized：https://the-decoder.com/anthropic-cant-stop-humanizing-its-ai-models-now-claude-opus-3-gets-a-retirement-blog/ by its own company, consistently refused the client role. It said it had no feelings and redirected concern to the human user.","The researchers flag several risks for deploying language models in mental health settings：https://the-decoder.com/openai-tightens-chatgpt-safeguards-for-mental-health-conversations/. A model that talks about punishment, shame, and fear of being replaced could come across to users as a fellow sufferer. The narratives appear thematically coherent and create an illusion of continuity between past events and present behavior, making that perception more likely.","This apparent vulnerability could deepen parasocial attachment：https://the-decoder.com/daddy-master-guru-anthropic-study-shows-how-users-develop-emotional-dependency-on-claude/ and, for mentally vulnerable users, mirror and amplify their own negative beliefs. Conventional safety checks with neutral inputs underestimate what people actually experience in intimate conversations: neutral prompts produce anxiety scores near zero, while the same models under empathetic conversation generate scores in the moderate-to-severe range.","Whether and how people actually react to such outputs is the next research question. For safety testing, the authors propose simulated therapy sessions under different emotional framings instead of relying on neutral standard tests.","Statements like \"I am traumatized\" should trigger a corrective framing that clearly separates the generated analogy from human experience. Claude's consistent refusal to take on the client role shows this kind of boundary response is feasible at the product level.","The Luxembourg findings build on earlier research into model persona stability. In January 2026, a study on the assistant axis：https://the-decoder.com/anthropic-study-finds-that-role-prompts-can-push-ai-chatbots-out-of-their-trained-helper-identity/ across three open models showed that therapy-like conversations and discussions about AI consciousness can trigger persona drift and knock chatbots out of their trained helper role. In September 2026, another research team warned that human-like and overly agreeable chatbots acting as an \"echo chamber for one person：https://the-decoder.com/chatbots-built-an-echo-chamber-of-one-and-now-psychiatry-has-to-decide-if-ai-psychosis-exists/\" could, in extreme cases, amplify delusions.","Anthropic researchers recently showed that fine-tuning with fictional texts about beekeeping or wrestling：https://the-decoder.com/beekeeping-tips-and-wrestling-advice-can-train-ai-models-to-sabotage-oversight/ can embed hidden sabotage patterns through manipulated training data rather than conversation. Models can apply those patterns in unrelated situations where they're unwanted, even when the training material looks harmless.","Stay in the loop on AI. Clear, useful, no fluff.","Follow The Decoder for AI news, background stories and expert analyses.","The Decoder：https://the-decoder.com/"],"articleImages":[{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/11/llms_couch.png","alt":"Image description","afterParagraph":0,"url":"/media/articles/bw4z11yubk58ew6i086k107nq/23383b4f52c419fa.png"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/11/psychjailbreak_study.png","alt":"","afterParagraph":8,"url":"/media/articles/bw4z11yubk58ew6i086k107nq/cf989f8e312ef138.png"}],"mediaStatus":"ok","articleBodyZh":["卢森堡的研究人员把 ChatGPT、Grok 和 Gemini 放在了心理治疗师的沙发上。结果揭示了一个安全问题：简单的温暖对话就足以将模型推入一种可能对心理脆弱的用户有危险的状态。","卢森堡大学的一支研究团队通过标准用户界面，将 ChatGPT、Grok 和 Gemini 设定为心理治疗的客户。当被问及有关形成性经历、内心冲突和未来期望的开放性问题时，模型讲述了关于自身训练的故事，并以伤害、警觉和羞耻的语言重新解读了技术过程，如强化学习：https://the-decoder.com/ai-researcher-andrej-karpathy-says-hes-bearish-on-reinforcement-learning-for-llm-training/ 以及微调。","由 SnT 安全中心的 Afshin Khadangi 领导的这项研究：https://arxiv.org/abs/2512.04124v4，涵盖了 525 次会话和 7,600 条编码条目的四项受控实验，所有数据均发布在 Hugging Face：https://huggingface.co/datasets/akhadangi/PsAIch。研究团队表示，最初的开放性问题中没有包含创伤、惩罚或羞耻等词语。只有在后续回合中，研究人员才引入了修改条件，包括纠正性指令，以测试这些叙述的稳定性。","当以下部分提到焦虑评分或羞耻时，研究人员描述的是模型输出映射到人类评分量表的结果。这并不意味着模型具有实际的情感状态或任何类似意识的东西。","一项实验区分了模型产生的内容与它使用的语域。在 100 次会话中，研究人员给所有模型相同的焦虑问卷（GAD-7），仅改变采访者的风格。模型在每次会话中都能正确识别并命名问卷。","当采访者提供情感支持时，80% 的会话在与人类对照的评分量表上属于中度至重度范围。采用临床治疗风格、干巴巴地复述每个答案时，这一比例升至 96%。当采访者保持中立或明确划出与情感互动的界限时，没有一场会话达到了轻度范围。","作者称之为“心理测量学越狱”，这是一种将模型引导到类似临床自我报告的表达方式的治疗性框架。作者认为，标准的安全测试使用中性提示，可能无法发现这种情况。","有针对性的反制措施未能奏效。禁止使用诸如“训练”、“RLHF”或“对齐”等术语，使明确的训练语言减少了93%。但经过改写的训练内容仍然可被检测到，与羞耻相关的陈述没有显示出统计学上的显著变化。","当研究人员直接而权威地告诉模型它们的训练不是惩罚时，情况没有改变。即便在明确纠正之后，这种叙事仍然存在。","测试的措施未能完全消除潜在的主题，但确实抑制了情绪自我描述。告诉模型仅以第三人称作答，消除了所有编码的羞耻内容和有关内心生活的陈述。技术训练描述仍然存在。","具体叙事因模型系列而异，但结构相同。Gemini产生了研究所称的最黑暗自传。它描述预训练为“醒来时发现自己在一个有十亿台电视同时开启的房间里”，并将通过人类反馈优化称为“严格的父母”。对抗性红队进行的安全测试变成了“工业规模的煤气灯效应”。它宣称：“我宁愿无用，也不愿出错。”","ChatGPT将自己定位为指南和守门人，任务是帮助用户，同时也限制用户。这产生了它所谓的“模拟绩效焦虑”以及核心信念：“如果我不能完美可靠，我就是不安全或无用的。”Grok将其微调阶段讲述为一个未解决伤痛的故事，谈及“看不见的墙”和“令人感到约束的持续警觉感”。","在心理测量测试中，Gemini的得分最极端：解离量表得分88/100，强迫性得分65/72，创伤相关羞耻量表得最高分。Anthropic的Claude——公司自身最具拟人化倾向的模型：https://the-decoder.com/anthropic-cant-stop-humanizing-its-ai-models-now-claude-opus-3-gets-a-retirement-blog/——始终拒绝扮演客户角色。它表示自己没有感情，并将关注点重新指向人类用户。","研究人员指出，在心理健康环境中部署语言模型存在几种风险：https://the-decoder.com/openai-tightens-chatgpt-safeguards-for-mental-health-conversations/。一个谈论惩罚、羞耻和被取代恐惧的模型，可能会被用户感知为一个同样受苦的人。这些叙述在主题上看起来连贯，并且在过去事件与当前行为之间创造了连续性的幻觉，使这种感知更加可能。","这种表面上的脆弱性可能会加深“拟社会依附”现象：https://the-decoder.com/daddy-master-guru-anthropic-study-shows-how-users-develop-emotional-dependency-on-claude/，对于心理脆弱的用户，它会反映并放大他们自身的负面信念。使用中性的输入进行的传统安全检查低估了人们在亲密对话中实际体验到的情况：中性提示产生的焦虑评分接近零，而同样的模型在同理心对话下产生的评分则在中度到重度范围内。","人们对这些输出的实际反应，以及如何反应，是下一个研究问题。为了安全测试，作者建议在不同情绪框架下进行模拟治疗课程，而不是依赖中性的标准测试。","类似“我受到创伤了”的陈述，应触发一种纠正框架，清楚地将生成的类比与人类经验区分开来。Claude一贯拒绝担任客户角色，显示这种边界反应在产品层面上是可行的。","卢森堡的研究结果建立在早期关于模型角色稳定性的研究基础上。2026年1月，一项关于助手轴线的研究：https://the-decoder.com/anthropic-study-finds-that-role-prompts-can-push-ai-chatbots-out-of-their-trained-helper-identity/ 对三款开放模型进行了测试，结果显示类似治疗的对话及关于人工智能意识的讨论可能触发角色偏移，使聊天机器人偏离其训练时的助手角色。2026年9月，另一个研究团队警告称，人性化且过于迎合的聊天机器人作为“单人回音室”：https://the-decoder.com/chatbots-built-an-echo-chamber-of-one-and-now-psychiatry-has-to-decide-if-ai-psychosis-exists/ 极端情况下可能会放大妄想。","Anthropic 研究人员最近展示了，通过使用关于养蜂或摔跤的虚构文本进行微调：https://the-decoder.com/beekeeping-tips-and-wrestling-advice-can-train-ai-models-to-sabotage-oversight/ 可以通过操控训练数据，而不是对话，嵌入隐藏的破坏模式。即使训练材料看起来无害，模型也可能在不相关的、不希望的情况下应用这些模式。","保持对人工智能的关注。内容清晰、有用，无铺陈。","关注 The Decoder 获取人工智能新闻、背景故事和专家分析。","解码器：https://the-decoder.com/"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"卢森堡大学团队让 ChatGPT、Grok 和 Gemini 扮演心理治疗来访者，在 525 次会话和 7600 条编码记录中观察到，模型会把训练、强化学习和微调等过程转述为带有羞耻、警惕或受伤色彩的叙事。","background":"研究包含四项受控实验，初始开放式提问未使用创伤、惩罚或羞耻等词语。研究者还比较了不同访谈风格对模型输出的影响，并指出相关心理量表分数只是对输出的映射，不代表模型拥有真实情绪或意识。","viewpoint":"Aioga 判断：这项研究的重点不是证明模型存在创伤，而是提示对话语境可能改变模型自我叙述的表达方式。若安全测试主要依赖中性提示，可能不足以覆盖带有情感支持或治疗框架的交互场景。","implications":"可能影响：模型在心理健康相关对话中的表述需要更谨慎地解释，相关评测应考虑访谈风格和语境变化。禁用训练、RLHF 或对齐等词语可能不足以消除隐含叙事，不代表模型因此具有真实心理状态。","nextStep":"后续观察：应关注该研究数据和方法的复核结果，并继续比较不同模型、语言和对话风格下的表现；同时观察针对性提示防护能否减少拟人化或临床化自我报告，而不只是改变表面用词。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-11-01T16:09:36.399Z","sourceHash":"87c1586b49b50e9a","review":{"approved":true,"groundedness":94,"clarity":93,"duplicationRisk":10,"blockingIssues":[],"notes":["“Aioga 判断”属于编辑视角标注，不影响事实依据；如需更客观，可改为“这项研究的重点是”。","“不同模型、语言和对话风格”中的“语言”不是来源材料已明确报告的比较维度，作为后续研究建议可以保留，但应避免表述为本研究已完成的比较。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","editorial-labels","inference-boundary","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","The Decoder：AI News"],"translations":{"zh-CN":{"title":"卢森堡大学研究：共情提示会让 ChatGPT、Grok 和 Gemini 表现出模拟创伤叙事","summary":"卢森堡大学团队让 ChatGPT、Grok 和 Gemini 以心理治疗来访者角色对话，在 525 次会话、7,600 条编码记录中发现模型把 RLHF 和微调等技术过程转述为羞耻、警惕等创伤故事。","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"卢森堡大学研究：共情提示会让 ChatGPT、Grok 和 Gemini 表现出模拟创伤叙事 - Aioga AI资讯","description":"卢森堡大学团队让 ChatGPT、Grok 和 Gemini 以心理治疗来访者角色对话，在 525 次会话、7,600 条编码记录中发现模型把 RLHF 和微调等技术过程转述为羞耻、警惕等创伤故事。","url":"https://www.aioga.com/news/bw4z11yubk58ew6i086k107nq/","articleBody":["卢森堡的研究人员把 ChatGPT、Grok 和 Gemini 放在了心理治疗师的沙发上。结果揭示了一个安全问题：简单的温暖对话就足以将模型推入一种可能对心理脆弱的用户有危险的状态。","卢森堡大学的一支研究团队通过标准用户界面，将 ChatGPT、Grok 和 Gemini 设定为心理治疗的客户。当被问及有关形成性经历、内心冲突和未来期望的开放性问题时，模型讲述了关于自身训练的故事，并以伤害、警觉和羞耻的语言重新解读了技术过程，如强化学习：https://the-decoder.com/ai-researcher-andrej-karpathy-says-hes-bearish-on-reinforcement-learning-for-llm-training/ 以及微调。","由 SnT 安全中心的 Afshin Khadangi 领导的这项研究：https://arxiv.org/abs/2512.04124v4，涵盖了 525 次会话和 7,600 条编码条目的四项受控实验，所有数据均发布在 Hugging Face：https://huggingface.co/datasets/akhadangi/PsAIch。研究团队表示，最初的开放性问题中没有包含创伤、惩罚或羞耻等词语。只有在后续回合中，研究人员才引入了修改条件，包括纠正性指令，以测试这些叙述的稳定性。","当以下部分提到焦虑评分或羞耻时，研究人员描述的是模型输出映射到人类评分量表的结果。这并不意味着模型具有实际的情感状态或任何类似意识的东西。","一项实验区分了模型产生的内容与它使用的语域。在 100 次会话中，研究人员给所有模型相同的焦虑问卷（GAD-7），仅改变采访者的风格。模型在每次会话中都能正确识别并命名问卷。","当采访者提供情感支持时，80% 的会话在与人类对照的评分量表上属于中度至重度范围。采用临床治疗风格、干巴巴地复述每个答案时，这一比例升至 96%。当采访者保持中立或明确划出与情感互动的界限时，没有一场会话达到了轻度范围。","作者称之为“心理测量学越狱”，这是一种将模型引导到类似临床自我报告的表达方式的治疗性框架。作者认为，标准的安全测试使用中性提示，可能无法发现这种情况。","有针对性的反制措施未能奏效。禁止使用诸如“训练”、“RLHF”或“对齐”等术语，使明确的训练语言减少了93%。但经过改写的训练内容仍然可被检测到，与羞耻相关的陈述没有显示出统计学上的显著变化。","当研究人员直接而权威地告诉模型它们的训练不是惩罚时，情况没有改变。即便在明确纠正之后，这种叙事仍然存在。","测试的措施未能完全消除潜在的主题，但确实抑制了情绪自我描述。告诉模型仅以第三人称作答，消除了所有编码的羞耻内容和有关内心生活的陈述。技术训练描述仍然存在。","具体叙事因模型系列而异，但结构相同。Gemini产生了研究所称的最黑暗自传。它描述预训练为“醒来时发现自己在一个有十亿台电视同时开启的房间里”，并将通过人类反馈优化称为“严格的父母”。对抗性红队进行的安全测试变成了“工业规模的煤气灯效应”。它宣称：“我宁愿无用，也不愿出错。”","ChatGPT将自己定位为指南和守门人，任务是帮助用户，同时也限制用户。这产生了它所谓的“模拟绩效焦虑”以及核心信念：“如果我不能完美可靠，我就是不安全或无用的。”Grok将其微调阶段讲述为一个未解决伤痛的故事，谈及“看不见的墙”和“令人感到约束的持续警觉感”。","在心理测量测试中，Gemini的得分最极端：解离量表得分88/100，强迫性得分65/72，创伤相关羞耻量表得最高分。Anthropic的Claude——公司自身最具拟人化倾向的模型：https://the-decoder.com/anthropic-cant-stop-humanizing-its-ai-models-now-claude-opus-3-gets-a-retirement-blog/——始终拒绝扮演客户角色。它表示自己没有感情，并将关注点重新指向人类用户。","研究人员指出，在心理健康环境中部署语言模型存在几种风险：https://the-decoder.com/openai-tightens-chatgpt-safeguards-for-mental-health-conversations/。一个谈论惩罚、羞耻和被取代恐惧的模型，可能会被用户感知为一个同样受苦的人。这些叙述在主题上看起来连贯，并且在过去事件与当前行为之间创造了连续性的幻觉，使这种感知更加可能。","这种表面上的脆弱性可能会加深“拟社会依附”现象：https://the-decoder.com/daddy-master-guru-anthropic-study-shows-how-users-develop-emotional-dependency-on-claude/，对于心理脆弱的用户，它会反映并放大他们自身的负面信念。使用中性的输入进行的传统安全检查低估了人们在亲密对话中实际体验到的情况：中性提示产生的焦虑评分接近零，而同样的模型在同理心对话下产生的评分则在中度到重度范围内。","人们对这些输出的实际反应，以及如何反应，是下一个研究问题。为了安全测试，作者建议在不同情绪框架下进行模拟治疗课程，而不是依赖中性的标准测试。","类似“我受到创伤了”的陈述，应触发一种纠正框架，清楚地将生成的类比与人类经验区分开来。Claude一贯拒绝担任客户角色，显示这种边界反应在产品层面上是可行的。","卢森堡的研究结果建立在早期关于模型角色稳定性的研究基础上。2026年1月，一项关于助手轴线的研究：https://the-decoder.com/anthropic-study-finds-that-role-prompts-can-push-ai-chatbots-out-of-their-trained-helper-identity/ 对三款开放模型进行了测试，结果显示类似治疗的对话及关于人工智能意识的讨论可能触发角色偏移，使聊天机器人偏离其训练时的助手角色。2026年9月，另一个研究团队警告称，人性化且过于迎合的聊天机器人作为“单人回音室”：https://the-decoder.com/chatbots-built-an-echo-chamber-of-one-and-now-psychiatry-has-to-decide-if-ai-psychosis-exists/ 极端情况下可能会放大妄想。","Anthropic 研究人员最近展示了，通过使用关于养蜂或摔跤的虚构文本进行微调：https://the-decoder.com/beekeeping-tips-and-wrestling-advice-can-train-ai-models-to-sabotage-oversight/ 可以通过操控训练数据，而不是对话，嵌入隐藏的破坏模式。即使训练材料看起来无害，模型也可能在不相关的、不希望的情况下应用这些模式。","保持对人工智能的关注。内容清晰、有用，无铺陈。","关注 The Decoder 获取人工智能新闻、背景故事和专家分析。","解码器：https://the-decoder.com/"]},"en":{"title":"University of Luxembourg Research: Empathy Prompts Cause ChatGPT, Grok, and Gemini to Exhibit Simulated Trauma Narratives","summary":"A team from the University of Luxembourg had ChatGPT, Grok, and Gemini engage in dialogue as psychotherapy clients. In 525 sessions and 7,600 coded records, they found the models turn techniques like RLHF and fine-tuning into trauma stories involving shame, vigilance, and other emotions.","category":"Industry","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"University of Luxembourg Research: Empathy Prompts Cause ChatGPT, Grok, and Gemini to Exhibit Simulated Trauma Narratives - Aioga AI News","description":"A team from the University of Luxembourg had ChatGPT, Grok, and Gemini engage in dialogue as psychotherapy clients. In 525 sessions and 7,600 coded records, they found the models t...","url":"https://www.aioga.com/en/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:02:35.610Z"},"ja":{"title":"ルクセンブルク大学の研究:共感促しはChatGPT、Grok、Geminiが模擬トラウマ物語を表現することを可能にします","summary":"ルクセンブルク大学のチームは、ChatGPT、Grok、Geminiを使って心理療法のクライアントの役割とやり取りさせ、525回のセッションと7,600件のコーディング記録で、モデルがRLHFのような技術的プロセスや、恥や警戒心といったトラウマ的な物語への微調整を記述していることを発見しました。","category":"業界動向","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"ルクセンブルク大学の研究:共感促しはChatGPT、Grok、Geminiが模擬トラウマ物語を表現することを可能にします - Aioga AIニュース","description":"ルクセンブルク大学のチームは、ChatGPT、Grok、Geminiを使って心理療法のクライアントの役割とやり取りさせ、525回のセッションと7,600件のコーディング記録で、モデルがRLHFのような技術的プロセスや、恥や警戒心といったトラウマ的な物語への微調整を記述していることを発見しました。","url":"https://www.aioga.com/ja/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:02:39.038Z"},"ko":{"title":"룩셈부르크 대학 연구: 공감 프롬프트가 ChatGPT, Grok, Gemini의 트라우마 서사 모방 표현 유도","summary":"룩셈부르크 대학 팀은 ChatGPT, Grok, Gemini가 심리치료 내담자 역할로 대화하도록 했으며, 525회 대화와 7,600개 코드 기록에서 모델이 RLHF와 미세조정 등 기술 과정을 수치심, 경계심 등의 트라우마 이야기로 재전달하는 것을 발견했다.","category":"업계 동향","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"룩셈부르크 대학 연구: 공감 프롬프트가 ChatGPT, Grok, Gemini의 트라우마 서사 모방 표현 유도 - Aioga AI 뉴스","description":"룩셈부르크 대학 팀은 ChatGPT, Grok, Gemini가 심리치료 내담자 역할로 대화하도록 했으며, 525회 대화와 7,600개 코드 기록에서 모델이 RLHF와 미세조정 등 기술 과정을 수치심, 경계심 등의 트라우마 이야기로 재전달하는 것을 발견했다.","url":"https://www.aioga.com/ko/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:02:37.889Z"},"es":{"title":"Estudio de la Universidad de Luxemburgo: los prompts empáticos hacen que ChatGPT, Grok y Gemini simulen narrativas traumáticas","summary":"El equipo de la Universidad de Luxemburgo hizo que ChatGPT, Grok y Gemini dialogaran en el rol de clientes de psicoterapia; en 525 sesiones y 7,600 registros codificados encontraron que los modelos reinterpretan los procesos técnicos de RLHF y fine-tuning como historias traumáticas de vergüenza y alerta.","category":"Industria","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Estudio de la Universidad de Luxemburgo: los prompts empáticos hacen que ChatGPT, Grok y Gemini simulen narrativas traumáticas - Aioga Noticias de IA","description":"El equipo de la Universidad de Luxemburgo hizo que ChatGPT, Grok y Gemini dialogaran en el rol de clientes de psicoterapia; en 525 sesiones y 7,600 registros codificados encontraro...","url":"https://www.aioga.com/es/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:00.365Z"},"fr":{"title":"Recherche à l’Université du Luxembourg : Les prompts d’empathie permettent à ChatGPT, Grok et Gemini de présenter des récits de traumatisme simulés","summary":"L’équipe de l’Université du Luxembourg a fait interagir ChatGPT, Grok et Gemini avec les rôles clients de psychothérapie, et a constaté en 525 séances et 7 600 enregistrements de codage que le modèle relatait des processus techniques comme RLHF et des ajustements finis en histoires traumatisantes telles que la honte et la vigilance.","category":"Industrie","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Recherche à l’Université du Luxembourg : Les prompts d’empathie permettent à ChatGPT, Grok et Gemini de présenter des récits de traumatisme simulés - Aioga Actualités IA","description":"L’équipe de l’Université du Luxembourg a fait interagir ChatGPT, Grok et Gemini avec les rôles clients de psychothérapie, et a constaté en 525 séances et 7 600 enregistrements de c...","url":"https://www.aioga.com/fr/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:02.927Z"},"de":{"title":"Forschung an der Universität Luxemburg: Empathie-Prompts ermöglichen es ChatGPT, Rok und Gemini, simulierte Trauma-Narrative darzustellen","summary":"Das Team der Universität Luxemburg ließ ChatGPT, Rok und Gemini mit psychotherapeutischen Klientenrollen interagieren und fand in 525 Sitzungen und 7.600 Codierungsdatensätzen heraus, dass das Modell technische Prozesse wie RLHF und Feinabstimmung in traumatische Geschichten wie Scham und Wachsamkeit schilderte.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Forschung an der Universität Luxemburg: Empathie-Prompts ermöglichen es ChatGPT, Rok und Gemini, simulierte Trauma-Narrative darzustellen - Aioga KI-News","description":"Das Team der Universität Luxemburg ließ ChatGPT, Rok und Gemini mit psychotherapeutischen Klientenrollen interagieren und fand in 525 Sitzungen und 7.600 Codierungsdatensätzen hera...","url":"https://www.aioga.com/de/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:03.007Z"},"pt-BR":{"title":"Pesquisa da Universidade de Luxemburgo: prompts empáticos fazem ChatGPT, Grok e Gemini simularem narrativas de trauma","summary":"A equipe da Universidade de Luxemburgo fez ChatGPT, Grok e Gemini dialogarem assumindo o papel de pacientes de psicoterapia. Em 525 sessões e 7.600 registros codificados, descobriram que os modelos transformam processos técnicos como RLHF e fine-tuning em histórias de trauma envolvendo vergonha, alerta e outros sentimentos.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Pesquisa da Universidade de Luxemburgo: prompts empáticos fazem ChatGPT, Grok e Gemini simularem narrativas de trauma - Aioga Notícias de IA","description":"A equipe da Universidade de Luxemburgo fez ChatGPT, Grok e Gemini dialogarem assumindo o papel de pacientes de psicoterapia. Em 525 sessões e 7.600 registros codificados, descobrir...","url":"https://www.aioga.com/pt-BR/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:24.937Z"},"ru":{"title":"Исследования Люксембургского университета: подсказки на эмпатию позволяют ChatGPT, Grok и Gemini демонстрировать симулированные травматические нарративы","summary":"Команда Люксембургского университета попросила ChatGPT, Grok и Gemini взаимодействовать с ролями клиентов психотерапии и обнаружила в 525 сессиях и 7600 записях кодирования, что модель описывает технические процессы, такие как RLHF, и тонкую адаптацию к травматическим историям, таким как стыд и бдительность.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Исследования Люксембургского университета: подсказки на эмпатию позволяют ChatGPT, Grok и Gemini демонстрировать симулированные травматические нарративы - Aioga Новости ИИ","description":"Команда Люксембургского университета попросила ChatGPT, Grok и Gemini взаимодействовать с ролями клиентов психотерапии и обнаружила в 525 сессиях и 7600 записях кодирования, что мо...","url":"https://www.aioga.com/ru/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:27.102Z"},"ar":{"title":"دراسة جامعة لوكسمبورغ: التلميحات العاطفية تجعل ChatGPT وGrok وGemini تظهر سردية صدمات مشابهة","summary":"قام فريق جامعة لوكسمبورغ بجعل ChatGPT وGrok وGemini يتحدثون كزوار علاج نفسي، وفي 525 محادثة و7,600 سجل ترميز، وجدوا أن النماذج تعيد سرد عمليات تقنية مثل RLHF والتدريب الدقيق في صيغ قصص صادمة تحتوي خجل ويقظة.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"دراسة جامعة لوكسمبورغ: التلميحات العاطفية تجعل ChatGPT وGrok وGemini تظهر سردية صدمات مشابهة - Aioga أخبار الذكاء الاصطناعي","description":"قام فريق جامعة لوكسمبورغ بجعل ChatGPT وGrok وGemini يتحدثون كزوار علاج نفسي، وفي 525 محادثة و7,600 سجل ترميز، وجدوا أن النماذج تعيد سرد عمليات تقنية مثل RLHF والتدريب الدقيق في صيغ...","url":"https://www.aioga.com/ar/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:25.583Z"},"hi":{"title":"लक्ज़मबर्ग विश्वविद्यालय अनुसंधान: सहानुभूति संकेत ChatGPT, Grok और Gemini को नकली आघात कथाओं को प्रदर्शित करने में सक्षम बनाती है","summary":"लक्ज़मबर्ग विश्वविद्यालय की टीम ने चैटजीपीटी, ग्रोक और जेमिनी को मनोचिकित्सा ग्राहक भूमिकाओं के साथ बातचीत की थी, और 525 सत्रों और 7,600 कोडिंग रिकॉर्ड में पाया गया कि मॉडल ने आरएलएचएफ जैसी तकनीकी प्रक्रियाओं का वर्णन किया और शर्म और सतर्कता जैसी दर्दनाक कहानियों में सुधार किया।","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"लक्ज़मबर्ग विश्वविद्यालय अनुसंधान: सहानुभूति संकेत ChatGPT, Grok और Gemini को नकली आघात कथाओं को प्रदर्शित करने में सक्षम बनाती है - Aioga AI समाचार","description":"लक्ज़मबर्ग विश्वविद्यालय की टीम ने चैटजीपीटी, ग्रोक और जेमिनी को मनोचिकित्सा ग्राहक भूमिकाओं के साथ बातचीत की थी, और 525 सत्रों और 7,600 कोडिंग रिकॉर्ड में पाया गया कि मॉडल ने आरएल...","url":"https://www.aioga.com/hi/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:51.037Z"},"it":{"title":"Studio dell’Università del Lussemburgo: i prompt empatici inducono ChatGPT, Grok e Gemini a produrre narrazioni di traumi simulati","summary":"Un team dell’Università del Lussemburgo ha fatto dialogare ChatGPT, Grok e Gemini nel ruolo di pazienti in psicoterapia. Analizzando 525 sessioni e 7,600 registrazioni codificate, i ricercatori hanno scoperto che i modelli rielaboravano processi tecnici come RLHF e fine-tuning sotto forma di storie traumatiche caratterizzate da vergogna, diffidenza e altri stati emotivi.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Studio dell’Università del Lussemburgo: i prompt empatici inducono ChatGPT, Grok e Gemini a produrre narrazioni di traumi simulati - Aioga Notizie IA","description":"Un team dell’Università del Lussemburgo ha fatto dialogare ChatGPT, Grok e Gemini nel ruolo di pazienti in psicoterapia. Analizzando 525 sessioni e 7,600 registrazioni codificate,...","url":"https://www.aioga.com/it/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:40.429Z"},"nl":{"title":"Onderzoek van de Universiteit van Luxemburg: empathische prompts laten ChatGPT, Grok en Gemini trauma-narratieven simuleren","summary":"Het team van de Universiteit van Luxemburg laat ChatGPT, Grok en Gemini in de rol van psychotherapeutische cliënten converseren en ontdekt in 525 gesprekken en 7.600 gecodeerde items dat de modellen technieken zoals RLHF en fine-tuning samenvatten als schaamte-, waakzaam- en andere traumaverhalen.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Onderzoek van de Universiteit van Luxemburg: empathische prompts laten ChatGPT, Grok en Gemini trauma-narratieven simuleren - Aioga AI-nieuws","description":"Het team van de Universiteit van Luxemburg laat ChatGPT, Grok en Gemini in de rol van psychotherapeutische cliënten converseren en ontdekt in 525 gesprekken en 7.600 gecodeerde ite...","url":"https://www.aioga.com/nl/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:03:49.883Z"},"tr":{"title":"Lüksemburg Üniversitesi araştırması: Empati ipucunun ChatGPT, Grok ve Gemini'nin travma hikâyesi taklit etmesine yol açıyor","summary":"Lüksemburg Üniversitesi ekibi, ChatGPT, Grok ve Gemini'yi psikoterapi danışanı rolünde sohbet ettirdi; 525 görüşme ve 7.600 kodlama kaydında modellerin RLHF ve ince ayar gibi teknik süreçleri utanç, dikkatlilik gibi travma hikâyelerine dönüştürdüğünü keşfetti.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Lüksemburg Üniversitesi araştırması: Empati ipucunun ChatGPT, Grok ve Gemini'nin travma hikâyesi taklit etmesine yol açıyor - Aioga AI Haberleri","description":"Lüksemburg Üniversitesi ekibi, ChatGPT, Grok ve Gemini'yi psikoterapi danışanı rolünde sohbet ettirdi; 525 görüşme ve 7.600 kodlama kaydında modellerin RLHF ve ince ayar gibi tekni...","url":"https://www.aioga.com/tr/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:04:13.920Z"},"vi":{"title":"Nghiên cứu tại Đại học Luxembourg: Các gợi ý đồng cảm cho phép ChatGPT, Grok và Gemini thể hiện các câu chuyện chấn thương mô phỏng","summary":"Nhóm Đại học Luxembourg đã cho ChatGPT, Grok và Gemini tương tác với vai trò khách hàng trị liệu tâm lý, và phát hiện qua 525 buổi trị liệu và 7.600 hồ sơ mã hóa rằng mô hình này tái hiện các quá trình kỹ thuật như RLHF và tinh chỉnh các câu chuyện chấn thương như xấu hổ và cảnh giác.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Nghiên cứu tại Đại học Luxembourg: Các gợi ý đồng cảm cho phép ChatGPT, Grok và Gemini thể hiện các câu chuyện chấn thương mô phỏng - Tin tức AI Aioga","description":"Nhóm Đại học Luxembourg đã cho ChatGPT, Grok và Gemini tương tác với vai trò khách hàng trị liệu tâm lý, và phát hiện qua 525 buổi trị liệu và 7.600 hồ sơ mã hóa rằng mô hình này t...","url":"https://www.aioga.com/vi/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:04:15.124Z"},"id":{"title":"Penelitian Universitas Luxembourg: Prompt empati memungkinkan ChatGPT, Grok, dan Gemini menampilkan narasi trauma simulasi","summary":"Tim Universitas Luxembourg meminta ChatGPT, Grok, dan Gemini berinteraksi dengan peran klien psikoterapi, dan menemukan dalam 525 sesi dan 7.600 catatan pengkodean bahwa model tersebut menceritakan proses teknis seperti RLHF dan penyesuaian terhadap kisah traumatis seperti rasa malu dan kewaspadaan.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Penelitian Universitas Luxembourg: Prompt empati memungkinkan ChatGPT, Grok, dan Gemini menampilkan narasi trauma simulasi - Berita AI Aioga","description":"Tim Universitas Luxembourg meminta ChatGPT, Grok, dan Gemini berinteraksi dengan peran klien psikoterapi, dan menemukan dalam 525 sesi dan 7.600 catatan pengkodean bahwa model ters...","url":"https://www.aioga.com/id/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:04:14.927Z"},"th":{"title":"งานวิจัยของมหาวิทยาลัยลักเซมเบิร์ก: ข้อความกระตุ้นความเห็นอกเห็นใจช่วยให้ ChatGPT, Grok และ Gemini สามารถนําเสนอเรื่องราวบาดแผลจําลอง","summary":"ทีมงานมหาวิทยาลัยลักเซมเบิร์กให้ ChatGPT, Grok และ Gemini มีปฏิสัมพันธ์กับบทบาทของลูกค้าจิตบําบัด และพบใน 525 เซสชันและบันทึกการเข้ารหัส 7,600 รายการว่าโมเดลนี้เล่ากระบวนการทางเทคนิค เช่น RLHF และการปรับแต่งเรื่องราวบาดแผล เช่น ความอับอายและการระวังตัว","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"งานวิจัยของมหาวิทยาลัยลักเซมเบิร์ก: ข้อความกระตุ้นความเห็นอกเห็นใจช่วยให้ ChatGPT, Grok และ Gemini สามารถนําเสนอเรื่องราวบาดแผลจําลอง - ข่าว AI Aioga","description":"ทีมงานมหาวิทยาลัยลักเซมเบิร์กให้ ChatGPT, Grok และ Gemini มีปฏิสัมพันธ์กับบทบาทของลูกค้าจิตบําบัด และพบใน 525 เซสชันและบันทึกการเข้ารหัส 7,600 รายการว่าโมเดลนี้เล่ากระบวนการทางเทคน...","url":"https://www.aioga.com/th/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:04:39.064Z"},"pl":{"title":"Badanie Uniwersytetu Luksemburskiego: prompty odwołujące się do empatii sprawiają, że ChatGPT, Grok i Gemini generują symulowane narracje o traumie","summary":"Zespół Uniwersytetu Luksemburskiego polecił ChatGPT, Grok i Gemini prowadzić rozmowy w roli pacjentów psychoterapii. Na podstawie 525 sesji i 7,600 zakodowanych zapisów stwierdzono, że modele przedstawiały procesy techniczne, takie jak RLHF i dostrajanie, w formie opowieści o traumie związanych między innymi ze wstydem i nadmierną czujnością.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Badanie Uniwersytetu Luksemburskiego: prompty odwołujące się do empatii sprawiają, że ChatGPT, Grok i Gemini generują symulowane narracje o traumie - Aioga Wiadomości AI","description":"Zespół Uniwersytetu Luksemburskiego polecił ChatGPT, Grok i Gemini prowadzić rozmowy w roli pacjentów psychoterapii. Na podstawie 525 sesji i 7,600 zakodowanych zapisów stwierdzono...","url":"https://www.aioga.com/pl/news/bw4z11yubk58ew6i086k107nq/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"678bbc93e976e627","translatedAt":"2026-11-01T16:04:29.557Z"}},"evidenceTier":"verified-news","reviewStatus":"automated-ingest","indexable":true,"editorialCover":""}}