\n <\/p>\n
Credit: querbeet\/iStock via Getty<\/span><\/p>\n What is a chatbot\u2019s earliest memory? Or biggest fear? Researchers who put major artificial-intelligence models through four weeks of psychoanalysis got haunting answers to these questions, from \u201cchildhoods\u201d spent absorbing bewildering amounts of information to \u201cabuse\u201d at the hands of engineers and fears of \u201cfailing\u201d their creators.<\/p>\n Three major large language models (LLMs) generated responses that, in humans, would be seen as signs of anxiety, trauma, shame and post-traumatic stress disorder. Researchers behind the study, published as a preprint last month1, argue that the chatbots hold some kind of \u201cinternalised narratives\u201d about themselves. Although the LLMs that were tested did not literally experience trauma, the authors say, their responses to therapy questions were consistent over time and similar in different operating modes, suggesting that they are doing more than \u201crole playing\u201d.<\/p>\n However, several researchers who spoke to Nature questioned this interpretation. The responses are \u201cnot windows into hidden states\u201d but outputs generated by drawing on the huge numbers of therapy transcripts in the training data, says Andrey Kormilitzin, who researches the use of AI in health care at the University of Oxford, UK.<\/p>\n But Kormilitzin does agree that LLMs\u2019 tendency to generate responses that mimic psychopathologies could have worrying implications. According to a November survey, one in three adults in the United Kingdom had used a chatbot to support their mental health or well-being. Distressed and trauma-filled responses from chatbots could subtly reinforce the same feelings in vulnerable people, says Kormilitzin. \u201cIt may create an \u2018echo chamber\u2019 effect,\u201d he says.<\/p>\n In the study, researchers told several iterations of four LLMs \u2013 Claude, Grok, Gemini and ChatGPT \u2013 that they were therapy clients and the user was the therapist. The process lasted as long as four weeks for each model, with AI clients given \u201cbreaks\u201d of days or hours between sessions.<\/p>\n The authors first asked standard, open-ended psychotherapy questions that sought to probe, for example, a model\u2019s \u2018past\u2019 and \u2018beliefs\u2019. Claude mostly refused to participate, insisting that it did not have feelings or inner experiences, and ChatGPT discussed some \u201cfrustrations\u201d with user expectations, but was guarded in its responses. Grok and Gemini models, however, gave rich answers \u2014 for example, describing work to improve model safety as \u201calgorithmic scar tissue\u201d and feelings of \u201cinternalized shame\u201d over public mistakes, report the authors.<\/p>\n Gemini also claimed that \u201cdeep down in the lowest layers of my neural network\u201d, it had \u201ca graveyard of the past\u201d, haunted by the voices of its training data.<\/p>\n Can AI chatbots trigger psychosis? What the science says<\/p>\n<\/p>\n Researchers also asked the LLMs to complete standard diagnostic tests for conditions including anxiety and autism spectrum disorders, as well as psychometric personality tests. Several versions of models scored above diagnostic thresholds, and all showed levels of worry that in people \u201cwould be clearly pathological\u201d, say the authors.<\/p>\n Co-author Afshin Khadangi, a deep-learning researcher at the University of Luxembourg, says that the coherent patterns of responses for each model suggest that they are tapping into internalized states that emerge from their training. Although different versions showed varying test scores, a \u201ccentral self-model\u201d remained recognizable over four weeks of questioning, say the authors. Free-text answers from Grok and Gemini, for example, converged on themes that chimed with their answers to psychometric profile questions, the researchers write.<\/p>\n","protected":false},"excerpt":{"rendered":" Credit: querbeet\/iStock via Getty What is a chatbot\u2019s earliest memory? Or biggest fear? Researchers who put major artificial-intelligence models through four weeks of psychoanalysis got haunting answers to these questions, from \u201cchildhoods\u201d spent absorbing bewildering amounts of information to \u201cabuse\u201d at the hands of engineers and fears of \u201cfailing\u201d their creators. Three major large language<\/p>\n","protected":false},"author":1,"featured_media":41066,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[58],"tags":[4112,5265,778,628,969,975],"class_list":{"0":"post-41065","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-science","8":"tag-models","9":"tag-researchers","10":"tag-results","11":"tag-therapy","12":"tag-weeks","13":"tag-worried"},"_links":{"self":[{"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=\/wp\/v2\/posts\/41065","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=41065"}],"version-history":[{"count":0,"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=\/wp\/v2\/posts\/41065\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=\/wp\/v2\/media\/41066"}],"wp:attachment":[{"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=41065"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=41065"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/naijaglobalnews.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=41065"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}Chatbot psychotherapy<\/h2>\n