Outraged AI: Large language models prioritise emotion over cost in fairness enforcement

Outraged AI: Large language models prioritise emotion over cost in fairness enforcement | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Social Sciences - Article Outraged AI: Large language models prioritise emotion over cost in fairness enforcement Zhen Wu, Hao Liu, Yiqing Dai, Haotian Tan, Yu Lei, Yujia Zhou This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7830884/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Emotions guide human decisions, but whether large language models (LLMs) use emotion similarly remains unknown. We tested this using altruistic third-party punishment, where an observer incurs a personal cost to enforce fairness—a hallmark of human morality and often driven by negative emotion. In a large-scale comparison of 4,068 LLM agents with 1,159 adults across 796,100 decisions, LLMs used emotion to guide punishment, sometimes even more strongly than humans did: Unfairness elicited stronger negative emotion that led to more punishment; punishing unfairness produced more positive emotion than accepting; and critically, prompting self-reports of emotion causally increased punishment. However, mechanisms diverged: LLMs prioritised emotion over cost, enforcing norms in an almost all-or-none manner with reduced cost sensitivity, whereas humans balanced fairness and cost. Notably, reasoning models (o3-mini, DeepSeek-R1) were more cost-sensitive and closer to human behaviour than foundation models (GPT-3.5, DeepSeek-V3), yet remained heavily emotion-driven. These findings provide the first causal evidence of emotion-guided moral decisions in LLMs and reveal deficits in cost calibration and nuanced fairness judgements, reminiscent of early-stage human responses. We propose that LLMs progress along a trajectory paralleling human development; future models should integrate emotion with context‑sensitive reasoning to achieve human-like emotional intelligence. Scientific community and society/Social sciences/Interdisciplinary studies Scientific community and society/Social sciences/Psychology/Human behaviour Physical sciences/Mathematics and computing/Computer science Full Text Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryMaterials1010.docx Supplementary Information Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7830884","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Social Sciences - Article","associatedPublications":[],"authors":[{"id":532952497,"identity":"9a8bb672-f3b8-49a5-97b6-3a69ddd37b69","order_by":0,"name":"Zhen Wu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9klEQVRIiWNgGAWjYDACCQglx8DA2ABCIGBAlBZjHpK1JPaASKK0yM9uPvaAse1w+n7+w20Pv+44LM/A3rxNgqHmDk4tjHOOpRsAteT2SCS2G8ueOWzYwHOsTILh2DOcWpglcsyk/24DaWFsk5ZsO8zYABSRYGw4jFMLm0T+NwnGbYfTefgPgrXYN8i/wa+FRyKHDaQlgYchsU3yY9vhxAYJHvxaJCTSgAr+pRv23Ehsk2ZsS09u40krtkg4hluL/IzkZxIMZ6zl2fuPP5P82WZt289+eOONDzW4taCGBQ/IdyBWAnEagCH+g1iVo2AUjIJRMKIAAF1ITzBZ0F44AAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-3655-1069","institution":"Tsinghua University","correspondingAuthor":true,"prefix":"","firstName":"Zhen","middleName":"","lastName":"Wu","suffix":""},{"id":532952498,"identity":"8bef1adc-8e0f-4879-bdfd-c4b15bc8031a","order_by":1,"name":"Hao Liu","email":"","orcid":"","institution":"Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Liu","suffix":""},{"id":532952499,"identity":"d8cd1e47-4040-4915-9f2a-983a8d0e0bbf","order_by":2,"name":"Yiqing Dai","email":"","orcid":"","institution":"Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Yiqing","middleName":"","lastName":"Dai","suffix":""},{"id":532952500,"identity":"2ebf58ad-df6b-44e4-95a0-436059ec6e9a","order_by":3,"name":"Haotian Tan","email":"","orcid":"","institution":"Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Haotian","middleName":"","lastName":"Tan","suffix":""},{"id":532952501,"identity":"fd9a71e2-cf2f-46e4-adb6-fc3e29560cb2","order_by":4,"name":"Yu Lei","email":"","orcid":"","institution":"Beijing University of Posts and Telecommunications","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"","lastName":"Lei","suffix":""},{"id":532952502,"identity":"cfeac202-d4af-464f-ad09-47e44f630a66","order_by":5,"name":"Yujia Zhou","email":"","orcid":"","institution":"Tsinghua University","correspondingAuthor":false,"prefix":"","firstName":"Yujia","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2025-10-11 02:15:37","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7830884/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7830884/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":98501398,"identity":"d85f0c36-4a39-45c0-bd99-5134602a1cef","added_by":"auto","created_at":"2025-12-18 09:41:49","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1070070,"visible":true,"origin":"","legend":"","description":"","filename":"ManuscriptLLMEmotion1010.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7830884/v1_covered_1d3b24d1-09ca-4769-a713-40e98d7c5a1b.pdf"},{"id":98501397,"identity":"5435286d-6795-4fa4-99ad-692de2064a91","added_by":"auto","created_at":"2025-12-18 09:41:43","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":18108838,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"SupplementaryMaterials1010.docx","url":"https://assets-eu.researchsquare.com/files/rs-7830884/v1/18ffdf3fe45cb40961954a41.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Outraged AI: Large language models prioritise emotion over cost in fairness enforcement","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7830884/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7830884/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Emotions guide human decisions, but whether large language models (LLMs) use emotion similarly remains unknown. We tested this using altruistic third-party punishment, where an observer incurs a personal cost to enforce fairness—a hallmark of human morality and often driven by negative emotion. In a large-scale comparison of 4,068 LLM agents with 1,159 adults across 796,100 decisions, LLMs used emotion to guide punishment, sometimes even more strongly than humans did: Unfairness elicited stronger negative emotion that led to more punishment; punishing unfairness produced more positive emotion than accepting; and critically, prompting self-reports of emotion causally increased punishment. However, mechanisms diverged: LLMs prioritised emotion over cost, enforcing norms in an almost all-or-none manner with reduced cost sensitivity, whereas humans balanced fairness and cost. Notably, reasoning models (o3-mini, DeepSeek-R1) were more cost-sensitive and closer to human behaviour than foundation models (GPT-3.5, DeepSeek-V3), yet remained heavily emotion-driven. These findings provide the first causal evidence of emotion-guided moral decisions in LLMs and reveal deficits in cost calibration and nuanced fairness judgements, reminiscent of early-stage human responses. We propose that LLMs progress along a trajectory paralleling human development; future models should integrate emotion with context‑sensitive reasoning to achieve human-like emotional intelligence.","manuscriptTitle":"Outraged AI: Large language models prioritise emotion over cost in fairness enforcement","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-18 09:41:38","doi":"10.21203/rs.3.rs-7830884/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-human-behaviour","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"nathumbehav","sideBox":"Learn more about [Nature Human Behaviour](http://www.nature.com/nathumbehav/)","snPcode":"","submissionUrl":"","title":"Nature Human Behaviour","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"d3e0b135-b69d-4e41-b8fe-c48fb3e1c997","owner":[],"postedDate":"December 18th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":56657616,"name":"Scientific community and society/Social sciences/Interdisciplinary studies"},{"id":56657617,"name":"Scientific community and society/Social sciences/Psychology/Human behaviour"},{"id":56657618,"name":"Physical sciences/Mathematics and computing/Computer science"}],"tags":[],"updatedAt":"2025-12-18T09:41:38+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-18 09:41:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7830884","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7830884","identity":"rs-7830884","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}