On the Development of ToxicBias-Reasoning for Responsible Multicultural Bias Detection and Explanation

On the Development of ToxicBias-Reasoning for Responsible Multicultural Bias Detection and Explanation | 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 On the Development of ToxicBias-Reasoning for Responsible Multicultural Bias Detection and Explanation Anuj Kumar, Mahendra Kumar Gurve, Satyadev Ahlawat, Yamuna Prasad, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7505866/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 25 Mar, 2026 Read the published version in Language Resources and Evaluation → Version 1 posted 14 You are reading this latest preprint version Abstract Bias in language models refers to systematic unfairness toward social groups, making it essential to build datasets that capture such bias in diverse and multicultural contexts. Existing bias detection datasets are limited in cultural diversity, do not capture overlapping categories of bias, and rarely provide support for generating human-interpretable reasoning, which restricts their usefulness for responsible AI development. To address these gaps, we introduce the \textit{ToxicBias-Reasoning} dataset with 7,562 statements (5639 biased, 1,923 non-biased), including a new \textit{Caste} category (247 examples) and additional samples reflecting Indian cultural biases. Our key contributions are threefold: (1) we provide a high-quality dataset in which all classification labels are manually annotated, the reasoning test set is entirely manual, and the reasoning annotations for training and validation are generated through a GPT-4o–assisted human-in-the-loop pipeline, ensuring scalability while maintaining quality; (2) we establish strong baselines using transformer-based models (BERT, RoBERTa) under hierarchical and multitask configurations, where a logic-aware loss function is introduced to capture inter-label dependencies in the multilabel category classification task which improves macro-F1 for category-level prediction; and (3) we benchmark reasoning generation using a BART-Large model distilled from GPT-4o outputs, achieving a ROUGE-L of 45.22. These contributions offer the first comprehensive benchmark for interpretable and culturally inclusive bias detection with reasoning. Knowledge Base Responsible AI Bias Detection Multilabel Classification Natural Language Reasoning Social Bias Full Text Additional Declarations No competing interests reported. Ethics Statement This study uses publicly available data from the ToxicBias-Reasoning repository under the Creative Commons Attribution 4.0 International (CC-BY 4.0) license, which is open for research use. The dataset can be accessed at Kumar, A. (2025). ToxicBias-Reasoning dataset [Data set]. Zenodo. https://doi.org/10.5281/zenodo.17007112 Cite Share Download PDF Status: Published Journal Publication published 25 Mar, 2026 Read the published version in Language Resources and Evaluation → Version 1 posted Editorial decision: Revision requested 20 Nov, 2025 Reviews received at journal 17 Nov, 2025 Reviews received at journal 12 Nov, 2025 Reviews received at journal 02 Nov, 2025 Reviewers agreed at journal 22 Oct, 2025 Reviewers agreed at journal 21 Oct, 2025 Reviewers agreed at journal 18 Oct, 2025 Reviewers agreed at journal 17 Oct, 2025 Reviewers agreed at journal 17 Oct, 2025 Reviewers agreed at journal 17 Oct, 2025 Reviewers invited by journal 16 Oct, 2025 Editor assigned by journal 14 Oct, 2025 Submission checks completed at journal 01 Sep, 2025 First submitted to journal 01 Sep, 2025 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-7505866","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":508635800,"identity":"e0caacc4-2676-41d8-9587-ee9ccdce77f6","order_by":0,"name":"Anuj Kumar","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0ElEQVRIiWNgGAWjYJACgw8GEgyMDTAuMxFaDGcYSEiQpoWZh4FBgnhH8c/uMSi2KbCoY24/nbqZh8FOnoGd9wBeLRJ3zhgY54Ac1pO77TYPQ7JhAzNfAn5rbuRAtTSAtTAnAN1pgFeHPEiLBUhL/1uQlnrCWgxAWhhAWmaAbTlMWIvhjbQCwx4DCcnGGW+33ZxjcNywjZAWuRvJ2wx+/KnjN+zP3XbjTUW1PD//GfxagIANrMKwAexOIJeQeiBgfgAi5YlQOQpGwSgYBSMUAABouTnwZR9WfQAAAABJRU5ErkJggg==","orcid":"","institution":"Indian Institute of Technology Jammu","correspondingAuthor":true,"prefix":"","firstName":"Anuj","middleName":"","lastName":"Kumar","suffix":""},{"id":508635801,"identity":"a39d5137-6625-4fda-9f64-78fc1e69945f","order_by":1,"name":"Mahendra Kumar Gurve","email":"","orcid":"","institution":"Indian Institute of Technology Jammu","correspondingAuthor":false,"prefix":"","firstName":"Mahendra","middleName":"Kumar","lastName":"Gurve","suffix":""},{"id":508635802,"identity":"8f2f20e1-c375-4216-ae36-d7251a1bea8c","order_by":2,"name":"Satyadev Ahlawat","email":"","orcid":"","institution":"Indian Institute of Technology Jammu","correspondingAuthor":false,"prefix":"","firstName":"Satyadev","middleName":"","lastName":"Ahlawat","suffix":""},{"id":508635803,"identity":"e39fe459-0979-4c2f-b073-de812645cc2f","order_by":3,"name":"Yamuna Prasad","email":"","orcid":"","institution":"Indian Institute of Technology Jammu","correspondingAuthor":false,"prefix":"","firstName":"Yamuna","middleName":"","lastName":"Prasad","suffix":""},{"id":508635804,"identity":"bb4f63b4-833e-48ea-b017-01cb207a3e70","order_by":4,"name":"Virendra Singh","email":"","orcid":"","institution":"Indian Institute of Technology Bombay","correspondingAuthor":false,"prefix":"","firstName":"Virendra","middleName":"","lastName":"Singh","suffix":""}],"badges":[],"createdAt":"2025-09-01 08:08:42","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7505866/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7505866/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10579-026-09916-w","type":"published","date":"2026-03-25T16:13:17+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":105755983,"identity":"040881c5-7978-4902-a535-ebf7927f281f","added_by":"auto","created_at":"2026-03-30 16:33:46","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1396081,"visible":true,"origin":"","legend":"","description":"","filename":"LREbias.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7505866/v1_covered_8d6163a7-ba60-4001-b369-791e12ec2668.pdf"}],"financialInterests":"\u003cp\u003eNo competing interests reported.\u003c/p\u003e\n\u003cp\u003eEthics Statement\u003c/p\u003e\n\u003cp\u003eThis study uses publicly available data from the ToxicBias-Reasoning repository under the Creative Commons Attribution 4.0 International (CC-BY 4.0) license, which is open for research use. The dataset can be accessed at Kumar, A. (2025). ToxicBias-Reasoning dataset [Data set]. Zenodo. https://doi.org/10.5281/zenodo.17007112\u003c/p\u003e","formattedTitle":"On the Development of ToxicBias-Reasoning for Responsible Multicultural Bias Detection and Explanation","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"language-resources-and-evaluation","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"lrev","sideBox":"Learn more about [Language Resources and Evaluation](http://link.springer.com/journal/10579)","snPcode":"10579","submissionUrl":"https://submission.nature.com/new-submission/10579/3","title":"Language Resources and Evaluation","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Knowledge Base, Responsible AI, Bias Detection, Multilabel Classification, Natural Language Reasoning, Social Bias","lastPublishedDoi":"10.21203/rs.3.rs-7505866/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7505866/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Bias in language models refers to systematic unfairness toward social groups, making it essential to build datasets that capture such bias in diverse and multicultural contexts. Existing bias detection datasets are limited in cultural diversity, do not capture overlapping categories of bias, and rarely provide support for generating human-interpretable reasoning, which restricts their usefulness for responsible AI development. To address these gaps, we introduce the \\textit{ToxicBias-Reasoning} dataset with 7,562 statements (5639 biased, 1,923 non-biased), including a new \\textit{Caste} category (247 examples) and additional samples reflecting Indian cultural biases. Our key contributions are threefold: (1) we provide a high-quality dataset in which all classification labels are manually annotated, the reasoning test set is entirely manual, and the reasoning annotations for training and validation are generated through a GPT-4o–assisted human-in-the-loop pipeline, ensuring scalability while maintaining quality; (2) we establish strong baselines using transformer-based models (BERT, RoBERTa) under hierarchical and multitask configurations, where a logic-aware loss function is introduced to capture inter-label dependencies in the multilabel category classification task which improves macro-F1 for category-level prediction; and (3) we benchmark reasoning generation using a BART-Large model distilled from GPT-4o outputs, achieving a ROUGE-L of 45.22. These contributions offer the first comprehensive benchmark for interpretable and culturally inclusive bias detection with reasoning.","manuscriptTitle":"On the Development of ToxicBias-Reasoning for Responsible Multicultural Bias Detection and Explanation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-05 04:29:19","doi":"10.21203/rs.3.rs-7505866/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-20T10:22:02+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-17T11:43:56+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-12T13:22:29+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-02T14:06:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"147580533351307045243801636273335725142","date":"2025-10-22T09:05:36+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"10660604751650835152047639709837059206","date":"2025-10-21T15:09:29+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"6448693650653761412269243215973503103","date":"2025-10-18T11:27:02+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"90512516961298649447497329598691440741","date":"2025-10-17T09:57:05+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"37947953216490851591712905116147459422","date":"2025-10-17T09:48:57+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"44478542888716540709627845118476827933","date":"2025-10-17T07:12:29+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-16T15:06:01+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-14T15:18:35+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-01T15:22:03+00:00","index":"","fulltext":""},{"type":"submitted","content":"Language Resources and Evaluation","date":"2025-09-01T08:04:13+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"language-resources-and-evaluation","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"lrev","sideBox":"Learn more about [Language Resources and Evaluation](http://link.springer.com/journal/10579)","snPcode":"10579","submissionUrl":"https://submission.nature.com/new-submission/10579/3","title":"Language Resources and Evaluation","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"ff8a2fa7-a259-46d1-b826-7a3a0b26cb3a","owner":[],"postedDate":"September 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-30T16:29:06+00:00","versionOfRecord":{"articleIdentity":"rs-7505866","link":"https://doi.org/10.1007/s10579-026-09916-w","journal":{"identity":"language-resources-and-evaluation","isVorOnly":false,"title":"Language Resources and Evaluation"},"publishedOn":"2026-03-25 16:13:17","publishedOnDateReadable":"March 25th, 2026"},"versionCreatedAt":"2025-09-05 04:29:19","video":"","vorDoi":"10.1007/s10579-026-09916-w","vorDoiUrl":"https://doi.org/10.1007/s10579-026-09916-w","workflowStages":[]},"version":"v1","identity":"rs-7505866","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7505866","identity":"rs-7505866","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}