Emergent Moral Representations in Large Language Models Aligns with Human Conceptual, Neural, and Behavioral Moral Structure

Emergent Moral Representations in Large Language Models Aligns with Human Conceptual, Neural, and Behavioral Moral Structure | 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 Research Article Emergent Moral Representations in Large Language Models Aligns with Human Conceptual, Neural, and Behavioral Moral Structure Behnam Karami, Fatemeh Zandi, Javad Hatami This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8270539/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Large language models (LLMs) increasingly operate in ethically sensitive settings, yet it remains unclear whether they internally encode structured representations of morality. Here we examine the activation space representations of several mid-sized LLMs to test whether statistical learning over text gives rise to moral distinctions that parallel human conceptual, behavioral, and neural organization. Using multivariate decoding, representational similarity analysis, partial least squares correlation, and behavioral prediction, we show that moral foundations are linearly decodable from hidden activations, with peak discriminability in mid layers. Representational similarity analysis uncovered a hierarchical moral geometry consistent with Moral Foundations Theory, with the posterior cingulate cortex (PCC) showing robust multivariate decoding and a representational structure that aligned most strongly with mid-layer LLM activations. These same mid-layers also predicted human wrongness judgments, indicating a shared computational substrate for moral evaluation. Partial least squares correlation further revealed orthogonal activation dimensions corresponding to individual foundations and their higher-order abstractions, yielding interpretable axes along which moral meaning is encoded. These results reveal a striking convergence across conceptual, behavioral, neural, and model representations, positioning LLMs as emerging neurocognitive models of moral reasoning and offering a window into the internal mechanisms that shape their behavior in sensitive domains. Such alignment may enable more explainable and transparent AI systems and support future efforts to ground LLMs in human values. Computational Neuroscience NeuroAI NeuroConnectionism Moral Foundations Theory NeuroCogntive Model Large Language Models Full Text Additional Declarations The authors declare no competing interests. Supplementary Files NMISupplementary.docx NMISupplementaryExtendedFigures.docx Cite Share Download PDF Status: Posted 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-8270539","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":554697573,"identity":"c2bef1c6-304f-415a-b58b-4ebbc82bbe15","order_by":0,"name":"Behnam Karami","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIiWNgGAWjYBACNgYeBgbGBoYEIJvxAZDg4SNFC7MBSAsbYXsQWtgkIIYQAHzsZw9/YNxhk2fOfjqt8muOnQwbA/PDRzfwOYwnL02C8UxasWVP7rbbstuSgQ5jMzbOwadFgseMgbHtcOKGA0AtktuYgVp42KQJaDH+wNj2P3HD+bfbiiW31ROlxUCCse1A4oYbudsYP247TIQWkF8SzyQDtbzdLM247TgPGzMBv8i3A0Ps4w47oMNyN378ua3anp+9+eFjfFrAIAFKM/OASULKkQHjD1JUj4JRMApGwYgBAJOQRcdjWFrBAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0001-8868-3384","institution":"University of Tehran","correspondingAuthor":true,"prefix":"","firstName":"Behnam","middleName":"","lastName":"Karami","suffix":""},{"id":554697805,"identity":"45aab530-3041-4f0a-9406-51624d8b8367","order_by":1,"name":"Fatemeh Zandi","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Fatemeh","middleName":"","lastName":"Zandi","suffix":""},{"id":554697806,"identity":"b55413d9-5a5c-4537-821e-5e71c784c275","order_by":2,"name":"Javad Hatami","email":"","orcid":"","institution":"University of Tehran","correspondingAuthor":false,"prefix":"","firstName":"Javad","middleName":"","lastName":"Hatami","suffix":""}],"badges":[],"createdAt":"2025-12-03 13:00:46","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8270539/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8270539/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97424080,"identity":"f169cb00-969b-4cc2-8446-9404488a865a","added_by":"auto","created_at":"2025-12-04 08:54:01","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2152734,"visible":true,"origin":"","legend":"","description":"","filename":"NMIEmergentMoralGeometryinLargeLanguageModelsAlignswithHumanNature.docx","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/4a7cd3745f0d91dffbd7adbc.docx"},{"id":97424099,"identity":"7c72281d-8f28-4f80-b70b-1be111b963e4","added_by":"auto","created_at":"2025-12-04 08:54:03","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342,"visible":true,"origin":"","legend":"","description":"","filename":"rs8270539.json","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/9b678732951e61d61fbb6e92.json"},{"id":97424334,"identity":"09e82638-739f-4058-8ae2-9d135fec57e4","added_by":"auto","created_at":"2025-12-04 08:55:34","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342668,"visible":true,"origin":"","legend":"","description":"","filename":"rs82705390enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/26ebca9f7f3368c168d21e40.xml"},{"id":97424074,"identity":"4dcb17c6-1b96-4ebf-aed0-1609c9defd9d","added_by":"auto","created_at":"2025-12-04 08:54:00","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":315559,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/2bf6ee25dc54047aa92e9307.png"},{"id":97424098,"identity":"2fcfa2dd-4ceb-43df-9d9c-bd6db219633b","added_by":"auto","created_at":"2025-12-04 08:54:03","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":240985,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/ab7a10376da16aa54d0a5f4d.png"},{"id":97424117,"identity":"c3212ce8-7ebd-4f46-b9c1-91ff31c6c8bd","added_by":"auto","created_at":"2025-12-04 08:54:04","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":922911,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/eca304d66fd5961dfbd882e5.jpeg"},{"id":97424092,"identity":"6e9defc9-3d18-458e-96ff-19a19fbf6382","added_by":"auto","created_at":"2025-12-04 08:54:02","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":270101,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/e7e97b4466a5b060de67a45f.jpeg"},{"id":97424090,"identity":"bff9cf20-b6e7-4d3d-9c3a-b3ed7b61c452","added_by":"auto","created_at":"2025-12-04 08:54:02","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":65028,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/813f01ae74e4137067e16ecb.png"},{"id":97424059,"identity":"1e0de931-3125-4396-8ad1-3d520f9d39e0","added_by":"auto","created_at":"2025-12-04 08:53:58","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":56256,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/4cc68e4c5d8f5f4890bf3aa3.png"},{"id":97424088,"identity":"67ca9c88-34dc-4f73-8e36-98445ed05d0b","added_by":"auto","created_at":"2025-12-04 08:54:01","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":170229,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/4e8e178b0f59927dab632cf3.png"},{"id":97424116,"identity":"4c4ba4f5-aaa1-4fa6-8cd3-f81530219f14","added_by":"auto","created_at":"2025-12-04 08:54:04","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":51548,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/8e1d6a7250188f2c8e6b9250.png"},{"id":97424126,"identity":"21b294b7-9462-482e-bae2-4af4746e5997","added_by":"auto","created_at":"2025-12-04 08:54:06","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":338917,"visible":true,"origin":"","legend":"","description":"","filename":"rs82705390structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/c50d83af5b6b910d16e1af38.xml"},{"id":97424089,"identity":"ad863432-1f69-427b-9381-bf1d38b6e6ee","added_by":"auto","created_at":"2025-12-04 08:54:02","extension":"html","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":362297,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/f508a5b4e34c4a3947eb94b6.html"},{"id":97666887,"identity":"9f0c3a25-951f-4937-bcd2-83afb28b4043","added_by":"auto","created_at":"2025-12-08 09:22:20","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1112193,"visible":true,"origin":"","legend":"","description":"","filename":"NMIEmergentMoralGeometryinLargeLanguageModelsAlignswithHumanNature.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1_covered_cfa99a02-3060-434b-876e-11ae8a5a211d.pdf"},{"id":97424121,"identity":"810a7bb2-9e19-4880-af7d-2f278caf69b1","added_by":"auto","created_at":"2025-12-04 08:54:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":224454,"visible":true,"origin":"","legend":"","description":"","filename":"NMISupplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/3861e8751be4e5d2ee27f4ff.docx"},{"id":97424119,"identity":"716b0eec-8e82-4d86-a523-9d6f61845243","added_by":"auto","created_at":"2025-12-04 08:54:06","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":13798467,"visible":true,"origin":"","legend":"","description":"","filename":"NMISupplementaryExtendedFigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-8270539/v1/7ded2cf45fe00c3d341eead4.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eEmergent Moral Representations in Large Language Models Aligns with Human Conceptual, Neural, and Behavioral Moral Structure\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Tehran","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"NeuroAI, NeuroConnectionism, Moral Foundations Theory, NeuroCogntive Model, Large Language Models ","lastPublishedDoi":"10.21203/rs.3.rs-8270539/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8270539/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eLarge language models (LLMs) increasingly operate in ethically sensitive settings, yet it remains unclear whether they internally encode structured representations of morality. Here we examine the activation space representations of several mid-sized LLMs to test whether statistical learning over text gives rise to moral distinctions that parallel human conceptual, behavioral, and neural organization. Using multivariate decoding, representational similarity analysis, partial least squares correlation, and behavioral prediction, we show that moral foundations are linearly decodable from hidden activations, with peak discriminability in mid layers. Representational similarity analysis uncovered a hierarchical moral geometry consistent with Moral Foundations Theory, with the posterior cingulate cortex (PCC) showing robust multivariate decoding and a representational structure that aligned most strongly with mid-layer LLM activations. These same mid-layers also predicted human wrongness judgments, indicating a shared computational substrate for moral evaluation. Partial least squares correlation further revealed orthogonal activation dimensions corresponding to individual foundations and their higher-order abstractions, yielding interpretable axes along which moral meaning is encoded. These results reveal a striking convergence across conceptual, behavioral, neural, and model representations, positioning LLMs as emerging neurocognitive models of moral reasoning and offering a window into the internal mechanisms that shape their behavior in sensitive domains. Such alignment may enable more explainable and transparent AI systems and support future efforts to ground LLMs in human values.\u003c/p\u003e","manuscriptTitle":"Emergent Moral Representations in Large Language Models Aligns with Human Conceptual, Neural, and Behavioral Moral Structure","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-04 08:53:20","doi":"10.21203/rs.3.rs-8270539/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b651f288-842e-463a-8463-960de3946242","owner":[],"postedDate":"December 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":59031636,"name":"Computational Neuroscience"}],"tags":[],"updatedAt":"2025-12-04T08:53:20+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-04 08:53:20","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8270539","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8270539","identity":"rs-8270539","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}