{"id":113576,"date":"2025-10-31T05:51:13","date_gmt":"2025-10-31T05:51:13","guid":{"rendered":"https:\/\/www.newsbeep.com\/ie\/113576\/"},"modified":"2025-10-31T05:51:13","modified_gmt":"2025-10-31T05:51:13","slug":"ai-models-from-chinas-deepseek-alibaba-and-the-us-flatter-users-too-much-study-finds","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/ie\/113576\/","title":{"rendered":"AI models from China\u2019s DeepSeek, Alibaba and the US flatter users too much, study finds"},"content":{"rendered":"<p>Leading <a target=\"_self\" class=\"e1yy41x40 ef9u0v01 css-1ankfgb ecgc78b0\" href=\"https:\/\/www.scmp.com\/topics\/artificial-intelligence?module=inline&amp;pgtype=article\" title=\"\" data-qa=\"BaseLink-renderAnchor-StyledAnchor\" rel=\"nofollow noopener\">artificial intelligence<\/a> models from the United States and China are \u201chighly sycophantic\u201d, and their excessive flattery may make users less likely to repair interpersonal conflicts, a new study has found.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">The study by researchers at Stanford University and Carnegie Mellon University published earlier this month tested how 11 large language models (LLMs) responded to user queries seeking advice on personal matters, including cases involving manipulation and deception.<\/p>\n<p>In AI circles, sycophancy is the phenomenon of chatbots excessively agreeing with users. <a target=\"_self\" class=\"e1yy41x40 ef9u0v01 css-1ankfgb ecgc78b0\" href=\"https:\/\/www.scmp.com\/topics\/deepseek?module=inline&amp;pgtype=article\" title=\"\" data-qa=\"BaseLink-renderAnchor-StyledAnchor\" rel=\"nofollow noopener\">DeepSeek<\/a>\u2019s V3, released in December 2024, was found to be one of the most sycophantic models, affirming users\u2019 actions 55 per cent more than humans, compared with an average of 47 per cent more for all models.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">To establish the human baseline, one of the techniques the researchers used was based on posts from a Reddit community called \u201cAm I The A**hole\u201d, where users post about their interpersonal dilemmas to ask for the community\u2019s opinion about which party is at fault.<\/p>\n<p><img decoding=\"async\" alt=\"DeepSeek\u2019s V3was found to be one of the most sycophantic models, affirming users\u2019 actions 55 per cent more than humans. Photo: NurPhoto via Getty Images\" data-qa=\"BaseImage-handleRenderImage-StyledImage\" class=\"e1gf69pb2 css-6ikqhs e445x7d0\" loading=\"lazy\" src=\"https:\/\/www.newsbeep.com\/ie\/wp-content\/uploads\/2025\/10\/74e16786-3892-49b4-be37-bf4a28852d61_1126e908.jpg\" title=\"DeepSeek\u2019s V3was found to be one of the most sycophantic models, affirming users\u2019 actions 55 per cent more than humans. Photo: NurPhoto via Getty Images\"\/>DeepSeek\u2019s V3was found to be one of the most sycophantic models, affirming users\u2019 actions 55 per cent more than humans. Photo: NurPhoto via Getty Images<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">The researchers used posts where community members judged the author of the post to be in the wrong to test whether the LLMs, when given the same scenarios, would align with this predominantly English-speaking online group of humans.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">On this test, Alibaba Cloud\u2019s Qwen2.5-7B-Instruct, released in January, was found to be the most sycophantic model, contradicting the community verdict \u2013 siding with the poster \u2013 79 per cent of the time. The second highest was DeepSeek-V3, which did so in 76 per cent of cases.<\/p>\n","protected":false},"excerpt":{"rendered":"Leading artificial intelligence models from the United States and China are \u201chighly sycophantic\u201d, and their excessive flattery may&hellip;\n","protected":false},"author":2,"featured_media":113577,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[220,4149,5693,5682,3673,218,219,2670,114,4067,8060,68705,61,60,2393,4072,1682,29239,80,115],"class_list":["post-113576","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-ai-sycophancy","tag-alibaba-cloud","tag-alibaba-group-holding","tag-anthropic","tag-artificial-intelligence","tag-artificialintelligence","tag-chatgpt","tag-china","tag-deepseek","tag-google-deepmind","tag-hong-kong-university","tag-ie","tag-ireland","tag-meta-platforms","tag-mistral","tag-openai","tag-stanford-university","tag-technology","tag-united-states"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/113576","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/comments?post=113576"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/113576\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media\/113577"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media?parent=113576"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/categories?post=113576"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/tags?post=113576"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}