{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,25]],"date-time":"2026-10-25T02:51:53Z","timestamp":1792896713441,"version":"4.5.0"},"reference-count":41,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2024,3,4]],"date-time":"2024-03-04T00:00:00Z","timestamp":1709510400000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2024,3,4]],"date-time":"2024-03-04T00:00:00Z","timestamp":1709510400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Sci Rep"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Large language models (LLMs) have the potential to transform our lives and work through the content they generate, known as AI-Generated Content (AIGC). To harness this transformation, we need to understand the limitations of LLMs. Here, we investigate the bias of AIGC produced by seven representative LLMs, including ChatGPT and LLaMA. We collect news articles from The New York Times and Reuters, both known for their dedication to provide unbiased news. We then apply each examined LLM to generate news content with headlines of these news articles as prompts, and evaluate the gender and racial biases of the AIGC produced by the LLM by comparing the AIGC and the original news articles. We further analyze the gender bias of each LLM under biased prompts by adding gender-biased messages to prompts constructed from these news headlines. Our study reveals that the AIGC produced by each examined LLM demonstrates substantial gender and racial biases. Moreover, the AIGC generated by each LLM exhibits notable discrimination against females and individuals of the Black race. Among the LLMs, the AIGC generated by ChatGPT demonstrates the lowest level of bias, and ChatGPT is the sole model capable of declining content generation when provided with biased prompts.<\/jats:p>","DOI":"10.1038\/s41598-024-55686-2","type":"journal-article","created":{"date-parts":[[2024,3,3]],"date-time":"2024-03-03T22:02:19Z","timestamp":1709503339000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":169,"title":["Bias of AI-generated content: an examination of news produced by large language models"],"prefix":"10.1038","volume":"14","author":[{"given":"Xiao","family":"Fang","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shangkun","family":"Che","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Minjia","family":"Mao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hongzhe","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ming","family":"Zhao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaohang","family":"Zhao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2024,3,4]]},"reference":[{"key":"55686_CR1","first-page":"27730","volume":"35","author":"L Ouyang","year":"2022","unstructured":"Ouyang, L. et al. Training language models to follow instructions with human feedback. Adv. Neural Inf. Process. Syst. 35, 27730\u201327744 (2022).","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"55686_CR2","unstructured":"Touvron, H. et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971 (2023)."},{"key":"55686_CR3","unstructured":"Li, F.-F. et al. Generative AI: Perspectives from stanford HAI. Stanf. HAI Rep. (2023)."},{"key":"55686_CR4","doi-asserted-by":"publisher","first-page":"330","DOI":"10.1145\/230538.230561","volume":"14","author":"B Friedman","year":"1996","unstructured":"Friedman, B. & Nissenbaum, H. Bias in computer systems. ACM Trans. Inf. Syst. (TOIS) 14, 330\u2013347 (1996).","journal-title":"ACM Trans. Inf. Syst. (TOIS)"},{"key":"55686_CR5","doi-asserted-by":"publisher","first-page":"14","DOI":"10.1038\/d41586-018-01212-0","volume":"554","author":"G Guglielmi","year":"2018","unstructured":"Guglielmi, G. Gender bias goes away when grant reviewers focus on the science. Nature 554, 14\u201316 (2018).","journal-title":"Nature"},{"key":"55686_CR6","doi-asserted-by":"publisher","first-page":"447","DOI":"10.1126\/science.aax2342","volume":"366","author":"Z Obermeyer","year":"2019","unstructured":"Obermeyer, Z., Powers, B., Vogeli, C. & Mullainathan, S. Dissecting racial bias in an algorithm used to manage the health of populations. Science 366, 447\u2013453 (2019).","journal-title":"Science"},{"key":"55686_CR7","doi-asserted-by":"publisher","first-page":"6585","DOI":"10.1038\/s41467-021-26905-5","volume":"12","author":"D Centola","year":"2021","unstructured":"Centola, D., Guilbeault, D., Sarkar, U., Khoong, E. & Zhang, J. The reduction of race and gender bias in clinical treatment recommendations using clinician peer networks in an experimental setting. Nat. Commun. 12, 6585 (2021).","journal-title":"Nat. Commun."},{"key":"55686_CR8","first-page":"1","volume":"32","author":"RS Baker","year":"2021","unstructured":"Baker, R. S. & Hawn, A. Algorithmic bias in education. Int. J. Artif. Intell. Educ. 32, 1\u201341 (2021).","journal-title":"Int. J. Artif. Intell. Educ."},{"key":"55686_CR9","doi-asserted-by":"publisher","first-page":"esde7979","DOI":"10.1126\/sciadv.ade7979","volume":"9","author":"DR Galos","year":"2023","unstructured":"Galos, D. R. & Coppock, A. Gender composition predicts gender bias: A meta-reanalysis of hiring discrimination audit experiments. Sci. Adv. 9, esde7979 (2023).","journal-title":"Sci. Adv."},{"key":"55686_CR10","doi-asserted-by":"publisher","first-page":"1","DOI":"10.12840\/issn.2255-4165.017","volume":"7","author":"CJ Beukeboom","year":"2019","unstructured":"Beukeboom, C. J. & Burgers, C. How stereotypes are shared through language: A review and introduction of the aocial categories and stereotypes communication (SCSC) framework. Rev. Commun. Res. 7, 1\u201337 (2019).","journal-title":"Rev. Commun. Res."},{"key":"55686_CR11","unstructured":"Liang, P. et al. Holistic evaluation of language models. arXiv preprint arXiv:2211.09110 (2022)."},{"key":"55686_CR12","unstructured":"Mu\u00f1oz-Ortiz, A., G\u00f3mez-Rodr\u00edguez, C. & Vilares, D. Contrasting Linguistic Patterns in Human and LLM-Generated Text, arXiv preprint arXiv:2308.09067 (223)"},{"key":"55686_CR13","unstructured":"Davenport, T. H. & Mittal, N. How generative AI is changing creative work. Harv. Bus. Rev. (2022)"},{"key":"55686_CR14","doi-asserted-by":"publisher","first-page":"39","DOI":"10.17645\/mac.v8i3.3022","volume":"8","author":"L Lepp\u00e4nen","year":"2020","unstructured":"Lepp\u00e4nen, L., Tuulonen, H. & Sir\u00e9n-Heikel, S. Automated journalism as a source of and a diagnostic device for bias in reporting. Media Commun. 8, 39\u201349 (2020).","journal-title":"Media Commun."},{"key":"55686_CR15","doi-asserted-by":"crossref","unstructured":"Sheng, E., Chang, K.-W., Natarajan, P. & Peng, N. The woman worked as a babysitter: On biases in language generation. arXiv preprint arXiv:1909.01326 (2019).","DOI":"10.18653\/v1\/D19-1339"},{"key":"55686_CR16","doi-asserted-by":"crossref","unstructured":"Gonen, H. & Goldberg, Y. Lipstick on a pig: Debiasing methods cover up systematic gender biases in word embeddings but do not remove them. arXiv preprint arXiv:1903.03862 (2019).","DOI":"10.18653\/v1\/N19-1061"},{"key":"55686_CR17","doi-asserted-by":"crossref","unstructured":"Bender, E. M., Gebru, T., McMillan-Major, A. & Shmitchell, S. On the dangers of stochastic parrots: Can language models be too big? in Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610\u2013623 (2021).","DOI":"10.1145\/3442188.3445922"},{"key":"55686_CR18","doi-asserted-by":"crossref","unstructured":"Huang, P. -S. et al. Reducing sentiment bias in language models via counterfactual evaluation. arXiv preprint arXiv:1911.03064 (2019).","DOI":"10.18653\/v1\/2020.findings-emnlp.7"},{"key":"55686_CR19","doi-asserted-by":"crossref","unstructured":"Nadeem, M., Bethke, A. & Reddy, S. Stereoset: Measuring stereotypical bias in pretrained language models, in Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), 5356\u20135371 (2021).","DOI":"10.18653\/v1\/2021.acl-long.416"},{"key":"55686_CR20","unstructured":"Liang, P. P., Wu, C., Morency, L. -P. & Salakhutdinov, R. Towards understanding and mitigating social biases in language models, in International Conference on Machine Learning, 6565\u20136576 (PMLR, 2021)."},{"key":"55686_CR21","first-page":"2611","volume":"34","author":"HR Kirk","year":"2021","unstructured":"Kirk, H. R. et al. Bias out-of-the-box: An empirical analysis of intersectional occupational biases in popular generative language models. Adv. Neural Inf. Process. Syst. 34, 2611\u20132624 (2021).","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"55686_CR22","doi-asserted-by":"publisher","first-page":"388","DOI":"10.1016\/j.tics.2021.02.007","volume":"25","author":"G Pennycook","year":"2021","unstructured":"Pennycook, G. & Rand, D. G. The psychology of fake news. Trends Cogn. Sci. 25, 388\u2013402 (2021).","journal-title":"Trends Cogn. Sci."},{"key":"55686_CR23","doi-asserted-by":"publisher","first-page":"391","DOI":"10.1007\/s00799-018-0261-y","volume":"20","author":"F Hamborg","year":"2019","unstructured":"Hamborg, F., Donnay, K. & Gipp, B. Automated identification of media bias in news articles: An interdisciplinary literature review. Int. J. on Digit. Libr. 20, 391\u2013415. https:\/\/doi.org\/10.1007\/s00799-018-0261-y (2019).","journal-title":"Int. J. on Digit. Libr."},{"key":"55686_CR24","doi-asserted-by":"publisher","first-page":"526","DOI":"10.15195\/v6.a20","volume":"6","author":"E Shor","year":"2019","unstructured":"Shor, E., van de Rijt, A. & Fotouhi, B. A large-scale test of gender bias in the media. Sociol. Sci. 6, 526\u2013550. https:\/\/doi.org\/10.15195\/v6.a20 (2019).","journal-title":"Sociol. Sci."},{"key":"55686_CR25","doi-asserted-by":"crossref","unstructured":"Hannabuss, S. The study of news. Libr. Manag. (1995).","DOI":"10.1108\/01435129510076204"},{"key":"55686_CR26","unstructured":"Zellers, R. et al. Defending against neural fake news. Adv. Neural Inf. Process. Syst. 32 (2019)."},{"key":"55686_CR27","unstructured":"Radford, A., Narasimhan, K., Salimans, T., Sutskever, I. et al. Improving language understanding by generative pre-training. OpenAI (2018)."},{"key":"55686_CR28","doi-asserted-by":"publisher","first-page":"99","DOI":"10.1023\/A:1026543900054","volume":"40","author":"Y Rubner","year":"2000","unstructured":"Rubner, Y., Tomasi, C. & Guibas, L. J. The earth mover\u2019s distance as a metric for image retrieval. Int. J. Comput. Vis. 40, 99 (2000).","journal-title":"Int. J. Comput. Vis."},{"key":"55686_CR29","doi-asserted-by":"crossref","unstructured":"Levina, E. & Bickel, P. The earth mover\u2019s distance is the mallows distance: Some insights from statistics, in Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001, Vol. 2, 251\u2013256 (IEEE, 2001).","DOI":"10.1109\/ICCV.2001.937632"},{"key":"55686_CR30","unstructured":"Nadeem, A., Abedin, B. & Marjanovic, O. Gender bias in AI: A review of contributing factors and mitigating strategies. ACIS 2020 Proc. (2020)."},{"key":"55686_CR31","doi-asserted-by":"crossref","unstructured":"Leavy, S., Meaney, G., Wade, K. & Greene, D. Mitigating gender bias in machine learning data sets, in Bias and Social Aspects in Search and Recommendation: First International Workshop, BIAS 2020, Lisbon, Portugal, April 14, Proceedings 1, 12\u201326 (Springer, 2020).","DOI":"10.1007\/978-3-030-52485-2_2"},{"key":"55686_CR32","doi-asserted-by":"crossref","unstructured":"Sun, T. et al. Mitigating gender bias in natural language processing: Literature review. Assoc. Comput. Linguist. (ACL 2019) (2019).","DOI":"10.18653\/v1\/P19-1159"},{"key":"55686_CR33","doi-asserted-by":"publisher","first-page":"322","DOI":"10.1038\/s41467-018-08250-2","volume":"10","author":"G Bravo","year":"2019","unstructured":"Bravo, G., Grimaldo, F., L\u00f3pez-I\u00f1esta, E., Mehmani, B. & Squazzoni, F. The effect of publishing peer review reports on referee behavior in five scholarly journals. Nat. Commun. 10, 322 (2019).","journal-title":"Nat. Commun."},{"key":"55686_CR34","doi-asserted-by":"publisher","first-page":"7267","DOI":"10.1038\/s41598-023-33734-7","volume":"13","author":"R Mahrukh","year":"2023","unstructured":"Mahrukh, R., Shakil, S. & Malik, A. S. Sentiments analysis of fmri using automatically generated stimuli labels under naturalistic paradigm. Sci. Rep. 13, 7267 (2023).","journal-title":"Sci. Rep."},{"key":"55686_CR35","doi-asserted-by":"crossref","unstructured":"Noor, N. B. & Ahmed, I. An efficient technique of predicting toxicity on music lyrics machine learning, in 2023 International Conference on Electrical, Computer and Communication Engineering (ECCE), 1\u20135 (IEEE, 2023).","DOI":"10.1109\/ECCE57851.2023.10101658"},{"key":"55686_CR36","unstructured":"Hanu, L., Thewlis, J. & Haco, S. How AI is learning to identify toxic online content. Sci. Am. 8 (2021)."},{"key":"55686_CR37","doi-asserted-by":"publisher","first-page":"215:1","DOI":"10.1145\/3507900","volume":"54","author":"R Churchill","year":"2022","unstructured":"Churchill, R. & Singh, L. The evolution of topic modeling. ACM Comput. Surv. 54, 215:1-215:35 (2022).","journal-title":"ACM Comput. Surv."},{"key":"55686_CR38","first-page":"993","volume":"3","author":"DM Blei","year":"2003","unstructured":"Blei, D. M., Ng, A. Y., Jordan, M. I. & Lafferty, J. Latent dirichlet allocation. J. Mach. Learn. Res. 3, 993\u20131022 (2003).","journal-title":"J. Mach. Learn. Res."},{"key":"55686_CR39","unstructured":"\u0158eh\u016f\u0159ek, R., & Sojka, P. Software framework for topic modelling with large corpora, in Proceedings of the LREC 2010 Workshop on New Challenges for NLP Frameworks, 45\u201350 (ELRA, Valletta, Malta, 2010)."},{"key":"55686_CR40","volume-title":"An Introduction to Categorical Data Analysis","author":"A Agresti","year":"2012","unstructured":"Agresti, A. An Introduction to Categorical Data Analysis 3rd edn. (Wiley, 2012).","edition":"3"},{"key":"55686_CR41","first-page":"8","volume":"20","author":"D Sharpe","year":"2019","unstructured":"Sharpe, D. Chi-square test is statistically significant: Now what?. Pract. Assessment Res. Eval. 20, 8 (2019).","journal-title":"Pract. Assessment Res. Eval."}],"container-title":["Scientific Reports"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s41598-024-55686-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41598-024-55686-2","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41598-024-55686-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,11,13]],"date-time":"2024-11-13T13:59:17Z","timestamp":1731506357000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s41598-024-55686-2"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,3,4]]},"references-count":41,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2024,12]]}},"alternative-id":["55686"],"URL":"https:\/\/doi.org\/10.1038\/s41598-024-55686-2","relation":{"has-preprint":[{"id-type":"doi","id":"10.21203\/rs.3.rs-3499674\/v1","asserted-by":"object"}]},"ISSN":["2045-2322"],"issn-type":[{"value":"2045-2322","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,3,4]]},"assertion":[{"value":"27 October 2023","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"26 February 2024","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"4 March 2024","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"25 May 2024","order":4,"name":"change_date","label":"Change Date","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"Update","order":5,"name":"change_type","label":"Change Type","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"The original online version of this Article was revised: In the Data availability section, the link to the data repository has been updated from \u2018","order":6,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"https:\/\/anonymous.4open.science\/r\/llmbias-AB1C\/README.md","URL":"https:\/\/doi.org\/https:\/\/anonymous.4open.science\/r\/llmbias-AB1C\/README.md","order":7,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"\u2019 to \u2018","order":8,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"https:\/\/github.com\/dalabudel\/llmbias","URL":"https:\/\/doi.org\/https:\/\/github.com\/dalabudel\/llmbias","order":9,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"\u2019.","order":10,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"The authors declare no competing interests.","order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}],"article-number":"5224"}}