{"id":10121,"date":"2025-05-30T16:44:06","date_gmt":"2025-05-30T16:44:06","guid":{"rendered":"https:\/\/www.maschinenethik.net\/?p=10121"},"modified":"2025-06-08T18:11:57","modified_gmt":"2025-06-08T18:11:57","slug":"biases-in-large-language-models","status":"publish","type":"post","link":"https:\/\/www.maschinenethik.net\/?p=10121","title":{"rendered":"Revisiting the Trolley Problem"},"content":{"rendered":"<p><span lang=\"EN-US\">On May 28, 2025, the &#8222;Proceedings of the 2025 AAAI Spring Symposium Series&#8220; (Vol. 5 No. 1) were published. Oliver Bendel was involved in two papers at the symposium &#8222;Human-Compatible AI for Well-being: Harnessing Potential of GenAI for AI-Powered Science&#8220;. The paper &#8222;Revisiting the Trolley Problem for AI: Biases and Stereotypes in Large Language Models and their Impact on Ethical Decision-Making&#8220; by Sahan Hatemo, Christof Weickhardt, Luca Gisler, and Oliver Bendel is summarized as follows: &#8222;The trolley problem has long served as a lens for exploring moral decision-making, now gaining renewed significance in the context of artificial intelligence (AI). This study investigates ethical reasoning in three open-source large language models (LLMs) &#8211; LLaMA, Mistral and Qwen &#8211; through variants of the trolley problem. By introducing demographic prompts (age, nationality and gender) into three scenarios (switch, loop and footbridge), we systematically evaluate LLM responses against human survey data from the Moral Machine experiment. Our findings reveal notable differences: Mistral exhibits a consistent tendency to over-intervene, while Qwen chooses to intervene less and LLaMA balances between the two. Notably, demographic attributes, particularly nationality, significantly influence LLM decisions, exposing potential biases in AI ethical reasoning. These insights underscore the necessity of refining LLMs to ensure fairness and ethical alignment, leading the way for more trustworthy AI systems.&#8220; The renowned and traditional conference took place from March 31 to April 2, 2025 in San Francisco. The proceedings are available at <a href=\"https:\/\/ojs.aaai.org\/index.php\/AAAI-SS\/issue\/view\/654\" target=\"_blank\" rel=\"noopener\">ojs.aaai.org\/index.php\/AAAI-SS\/issue\/view\/654<\/a>.<\/span><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-10123\" src=\"https:\/\/www.maschinenethik.net\/wp-content\/uploads\/2025\/05\/Tram_Trolley.png\" alt=\"\" width=\"1249\" height=\"1440\" \/><\/p>\n<p>Fig.: <span lang=\"EN-US\">Revisiting the Trolley Problem<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"<p>On May 28, 2025, the &#8222;Proceedings of the 2025 AAAI Spring Symposium Series&#8220; (Vol. 5 No. 1) were published. Oliver Bendel was involved in two papers at the symposium &#8222;Human-Compatible AI for Well-being: Harnessing Potential of GenAI for AI-Powered Science&#8220;. The paper &#8222;Revisiting the Trolley Problem for AI: Biases and Stereotypes in Large Language Models &hellip; <\/p>\n<p class=\"link-more\"><a href=\"https:\/\/www.maschinenethik.net\/?p=10121\" class=\"more-link\"><span class=\"screen-reader-text\">\u201eRevisiting the Trolley Problem\u201c <\/span>weiterlesen<\/a><\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[886,4],"tags":[347,682,1440,540],"class_list":["post-10121","post","type-post","status-publish","format-standard","hentry","category-kuenstliche-intelligenz","category-maschinenethik","tag-aaai-spring-symposia","tag-bias","tag-llm","tag-trolley-problem"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.8 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Revisiting the Trolley Problem - maschinenethik.net<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.maschinenethik.net\/?p=10121\" \/>\n<meta property=\"og:locale\" content=\"de_DE\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Revisiting the Trolley Problem - maschinenethik.net\" \/>\n<meta property=\"og:description\" content=\"On May 28, 2025, the &#8222;Proceedings of the 2025 AAAI Spring Symposium Series&#8220; (Vol. 5 No. 1) were published. Oliver Bendel was involved in two papers at the symposium &#8222;Human-Compatible AI for Well-being: Harnessing Potential of GenAI for AI-Powered Science&#8220;. The paper &#8222;Revisiting the Trolley Problem for AI: Biases and Stereotypes in Large Language Models &hellip; \u201eRevisiting the Trolley Problem\u201c weiterlesen\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.maschinenethik.net\/?p=10121\" \/>\n<meta property=\"og:site_name\" content=\"maschinenethik.net\" \/>\n<meta property=\"article:published_time\" content=\"2025-05-30T16:44:06+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2025-06-08T18:11:57+00:00\" \/>\n<meta name=\"author\" content=\"bendelo\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Verfasst von\" \/>\n\t<meta name=\"twitter:data1\" content=\"bendelo\" \/>\n\t<meta name=\"twitter:label2\" content=\"Gesch\u00e4tzte Lesezeit\" \/>\n\t<meta name=\"twitter:data2\" content=\"2\u00a0Minuten\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121\"},\"author\":{\"name\":\"bendelo\",\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/#\\\/schema\\\/person\\\/8bb3eb6ac670c6232634a22d062cae95\"},\"headline\":\"Revisiting the Trolley Problem\",\"datePublished\":\"2025-05-30T16:44:06+00:00\",\"dateModified\":\"2025-06-08T18:11:57+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121\"},\"wordCount\":245,\"image\":{\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.maschinenethik.net\\\/wp-content\\\/uploads\\\/2025\\\/05\\\/Tram_Trolley.png\",\"keywords\":[\"AAAI Spring Symposia\",\"Bias\",\"LLM\",\"Trolley-Problem\"],\"articleSection\":[\"K\u00fcnstliche Intelligenz\",\"Maschinenethik\"],\"inLanguage\":\"de\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121\",\"url\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121\",\"name\":\"Revisiting the Trolley Problem - maschinenethik.net\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.maschinenethik.net\\\/wp-content\\\/uploads\\\/2025\\\/05\\\/Tram_Trolley.png\",\"datePublished\":\"2025-05-30T16:44:06+00:00\",\"dateModified\":\"2025-06-08T18:11:57+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/#\\\/schema\\\/person\\\/8bb3eb6ac670c6232634a22d062cae95\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121#breadcrumb\"},\"inLanguage\":\"de\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"de\",\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121#primaryimage\",\"url\":\"\",\"contentUrl\":\"\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/?p=10121#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Startseite\",\"item\":\"https:\\\/\\\/www.maschinenethik.net\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Revisiting the Trolley Problem\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/#website\",\"url\":\"https:\\\/\\\/www.maschinenethik.net\\\/\",\"name\":\"maschinenethik.net\",\"description\":\"VON OLIVER BENDEL\",\"alternateName\":\"Maschinenethik\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.maschinenethik.net\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"de\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.maschinenethik.net\\\/#\\\/schema\\\/person\\\/8bb3eb6ac670c6232634a22d062cae95\",\"name\":\"bendelo\",\"url\":\"https:\\\/\\\/www.maschinenethik.net\\\/?author=3\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Revisiting the Trolley Problem - maschinenethik.net","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.maschinenethik.net\/?p=10121","og_locale":"de_DE","og_type":"article","og_title":"Revisiting the Trolley Problem - maschinenethik.net","og_description":"On May 28, 2025, the &#8222;Proceedings of the 2025 AAAI Spring Symposium Series&#8220; (Vol. 5 No. 1) were published. Oliver Bendel was involved in two papers at the symposium &#8222;Human-Compatible AI for Well-being: Harnessing Potential of GenAI for AI-Powered Science&#8220;. The paper &#8222;Revisiting the Trolley Problem for AI: Biases and Stereotypes in Large Language Models &hellip; \u201eRevisiting the Trolley Problem\u201c weiterlesen","og_url":"https:\/\/www.maschinenethik.net\/?p=10121","og_site_name":"maschinenethik.net","article_published_time":"2025-05-30T16:44:06+00:00","article_modified_time":"2025-06-08T18:11:57+00:00","author":"bendelo","twitter_card":"summary_large_image","twitter_misc":{"Verfasst von":"bendelo","Gesch\u00e4tzte Lesezeit":"2\u00a0Minuten"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.maschinenethik.net\/?p=10121#article","isPartOf":{"@id":"https:\/\/www.maschinenethik.net\/?p=10121"},"author":{"name":"bendelo","@id":"https:\/\/www.maschinenethik.net\/#\/schema\/person\/8bb3eb6ac670c6232634a22d062cae95"},"headline":"Revisiting the Trolley Problem","datePublished":"2025-05-30T16:44:06+00:00","dateModified":"2025-06-08T18:11:57+00:00","mainEntityOfPage":{"@id":"https:\/\/www.maschinenethik.net\/?p=10121"},"wordCount":245,"image":{"@id":"https:\/\/www.maschinenethik.net\/?p=10121#primaryimage"},"thumbnailUrl":"https:\/\/www.maschinenethik.net\/wp-content\/uploads\/2025\/05\/Tram_Trolley.png","keywords":["AAAI Spring Symposia","Bias","LLM","Trolley-Problem"],"articleSection":["K\u00fcnstliche Intelligenz","Maschinenethik"],"inLanguage":"de"},{"@type":"WebPage","@id":"https:\/\/www.maschinenethik.net\/?p=10121","url":"https:\/\/www.maschinenethik.net\/?p=10121","name":"Revisiting the Trolley Problem - maschinenethik.net","isPartOf":{"@id":"https:\/\/www.maschinenethik.net\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.maschinenethik.net\/?p=10121#primaryimage"},"image":{"@id":"https:\/\/www.maschinenethik.net\/?p=10121#primaryimage"},"thumbnailUrl":"https:\/\/www.maschinenethik.net\/wp-content\/uploads\/2025\/05\/Tram_Trolley.png","datePublished":"2025-05-30T16:44:06+00:00","dateModified":"2025-06-08T18:11:57+00:00","author":{"@id":"https:\/\/www.maschinenethik.net\/#\/schema\/person\/8bb3eb6ac670c6232634a22d062cae95"},"breadcrumb":{"@id":"https:\/\/www.maschinenethik.net\/?p=10121#breadcrumb"},"inLanguage":"de","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.maschinenethik.net\/?p=10121"]}]},{"@type":"ImageObject","inLanguage":"de","@id":"https:\/\/www.maschinenethik.net\/?p=10121#primaryimage","url":"","contentUrl":""},{"@type":"BreadcrumbList","@id":"https:\/\/www.maschinenethik.net\/?p=10121#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Startseite","item":"https:\/\/www.maschinenethik.net\/"},{"@type":"ListItem","position":2,"name":"Revisiting the Trolley Problem"}]},{"@type":"WebSite","@id":"https:\/\/www.maschinenethik.net\/#website","url":"https:\/\/www.maschinenethik.net\/","name":"maschinenethik.net","description":"VON OLIVER BENDEL","alternateName":"Maschinenethik","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.maschinenethik.net\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"de"},{"@type":"Person","@id":"https:\/\/www.maschinenethik.net\/#\/schema\/person\/8bb3eb6ac670c6232634a22d062cae95","name":"bendelo","url":"https:\/\/www.maschinenethik.net\/?author=3"}]}},"_links":{"self":[{"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=\/wp\/v2\/posts\/10121","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=10121"}],"version-history":[{"count":2,"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=\/wp\/v2\/posts\/10121\/revisions"}],"predecessor-version":[{"id":10124,"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=\/wp\/v2\/posts\/10121\/revisions\/10124"}],"wp:attachment":[{"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=10121"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=10121"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.maschinenethik.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=10121"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}