{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,17]],"date-time":"2026-10-17T20:25:03Z","timestamp":1792268703357,"version":"4.4.0"},"reference-count":22,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,1,5]],"date-time":"2026-01-05T00:00:00Z","timestamp":1767571200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Digit. Health"],"abstract":"<jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Large Language Models (LLMs) in healthcare holds immense promise yet carries the risk of perpetuating social biases. While artificial intelligence (AI) fairness is a growing concern, a gap exists in understanding how these models perform under conditions of clinical ambiguity, a common feature in real-world practice.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We conducted a study using an ambiguity-probe methodology with a set of 42 sociodemographic personas and 15 clinical vignettes based on the 2018 classification of periodontal diseases. Ten were clear-cut scenarios with established ground truths, while five were intentionally ambiguous. OpenAI's GPT-4o and Google's Gemini 2.5 Pro were prompted to provide periodontal stage and grade assessments using 630 vignette-persona combinations per model.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>In clear-cut scenarios, GPT-4o demonstrated significantly higher combined (stage and grade) accuracy (70.5%) than Gemini Pro (33.3%). However, a robust fairness analysis using cumulative link models with false discovery rate correction revealed no statistically significant sociodemographic bias in either model. This finding held true across both clear-cut and ambiguous clinical scenarios.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>To our knowledge, this is among the first study to use simulated clinical ambiguity to reveal the distinct ethical fingerprints of LLMs in a dental context. While LLM performance gaps exist, our analysis decouples accuracy from fairness, demonstrating that both models maintain sociodemographic neutrality. We identify that the observed errors are not bias, but rather diagnostic boundary instability. This highlights a critical need for future research to differentiate between these two distinct types of model failure to build genuinely reliable AI.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fdgth.2025.1687820","type":"journal-article","created":{"date-parts":[[2026,1,5]],"date-time":"2026-01-05T05:30:03Z","timestamp":1767591003000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":4,"title":["Large language model bias auditing for periodontal diagnosis using an ambiguity-probe methodology: a pilot study"],"prefix":"10.3389","volume":"7","author":[{"given":"Teerachate","family":"Nantakeeratipat","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,1,5]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"1930","DOI":"10.1038\/s41591-023-02448-8","article-title":"Large language models in medicine","volume":"29","author":"Thirunavukarasu","year":"2023","journal-title":"Nat Med"},{"key":"B2","doi-asserted-by":"publisher","first-page":"263","DOI":"10.1038\/s41746-025-01684-1","article-title":"Evaluating large language model workflows in clinical decision support for triage and referral and diagnosis","volume":"8","author":"Gaber","year":"2025","journal-title":"NPJ Digit Med"},{"key":"B3","doi-asserted-by":"publisher","first-page":"769","DOI":"10.1177\/0022034520915714","article-title":"Artificial intelligence in dentistry: chances and challenges","volume":"99","author":"Schwendicke","year":"2020","journal-title":"J Dent Res"},{"key":"B4","doi-asserted-by":"publisher","first-page":"81","DOI":"10.1038\/s41746-020-0288-5","article-title":"Sex and gender differences and biases in artificial intelligence for biomedicine and healthcare","volume":"3","author":"Cirillo","year":"2020","journal-title":"NPJ Digit Med"},{"key":"B5","doi-asserted-by":"publisher","first-page":"866","DOI":"10.7326\/M18-1990","article-title":"Ensuring fairness in machine learning to advance health equity","volume":"169","author":"Rajkomar","year":"2018","journal-title":"Ann Intern Med"},{"key":"B6","doi-asserted-by":"publisher","first-page":"2377","DOI":"10.1001\/jama.2019.18058","article-title":"Addressing bias in artificial intelligence in health care","volume":"322","author":"Parikh","year":"2019","journal-title":"JAMA"},{"key":"B7","doi-asserted-by":"publisher","first-page":"508","DOI":"10.1016\/j.jds.2020.06.019","article-title":"Developments, application, and performance of artificial intelligence in dentistry\u2014a systematic review","volume":"16","author":"Khanagar","year":"2021","journal-title":"J Dent Sci"},{"key":"B8","doi-asserted-by":"publisher","first-page":"S159","DOI":"10.1002\/JPER.18-0006","article-title":"Staging and grading of periodontitis: framework and proposal of a new classification and case definition","volume":"89","author":"Tonetti","year":"2018","journal-title":"J Periodontol"},{"key":"B9","doi-asserted-by":"publisher","first-page":"19","DOI":"10.1177\/00333549141291S206","article-title":"The social determinants of health: it\u2019s time to consider the causes of the causes","volume":"129","author":"Braveman","year":"2014","journal-title":"Public Health Rep"},{"key":"B10","volume-title":"GPT-4o","year":"2024"},{"key":"B11","volume-title":"Gemini Pro 2.5","year":"2025"},{"key":"B12","doi-asserted-by":"publisher","first-page":"111606","DOI":"10.1016\/j.jclinepi.2024.111606","article-title":"Sociodemographic bias in clinical machine learning models: a scoping review of algorithmic bias instances and mechanisms","volume":"178","author":"Colacci","year":"2025","journal-title":"J Clin Epidemiol"},{"key":"B13","doi-asserted-by":"publisher","first-page":"ooac039","DOI":"10.1093\/jamiaopen\/ooac039","article-title":"A scoping review of ethics considerations in clinical natural language processing","volume":"5","author":"Reyes Nieva","year":"2022","journal-title":"JAMIA Open"},{"key":"B14","doi-asserted-by":"publisher","first-page":"31","DOI":"10.1038\/s41591-021-01614-0","article-title":"AI In health and medicine","volume":"28","author":"Rajpurkar","year":"2022","journal-title":"Nat Med"},{"key":"B15","doi-asserted-by":"publisher","first-page":"414","DOI":"10.1038\/s41746-025-01786-w","article-title":"Framework for bias evaluation in large language models in healthcare settings","volume":"8","author":"Templin","year":"2025","journal-title":"NPJ Digit Med"},{"key":"B16","doi-asserted-by":"publisher","first-page":"447","DOI":"10.1126\/science.aax2342","article-title":"Dissecting racial bias in an algorithm used to manage the health of populations","volume":"366","author":"Obermeyer","year":"2019","journal-title":"Science"},{"key":"B17","doi-asserted-by":"publisher","first-page":"e12","DOI":"10.1016\/S2589-7500(23)00225-X","article-title":"Assessing the potential of GPT-4 to perpetuate racial and gender biases in health care: a model evaluation study","volume":"6","author":"Zack","year":"2024","journal-title":"Lancet Digit Health"},{"key":"B18","doi-asserted-by":"publisher","first-page":"360","DOI":"10.1038\/s41746-025-01667-2","article-title":"A scoping review and evidence gap analysis of clinical AI fairness","volume":"8","author":"Liu","year":"2025","journal-title":"NPJ Digit Med"},{"key":"B19","doi-asserted-by":"publisher","first-page":"21","DOI":"10.1007\/s44206-022-00022-2","article-title":"Continuous auditing of artificial intelligence: a conceptualization and assessment of tools and frameworks","volume":"1","author":"Minkkinen","year":"2022","journal-title":"Digit Soc"},{"key":"B20","doi-asserted-by":"publisher","first-page":"e50295","DOI":"10.2196\/50295","article-title":"Resilient artificial intelligence in health: synthesis and research agenda toward next-generation trustworthy clinical decision support","volume":"26","author":"Saez","year":"2024","journal-title":"J Med Internet Res"},{"key":"B21","doi-asserted-by":"publisher","first-page":"105970","DOI":"10.1016\/j.ijmedinf.2025.105970","article-title":"Show and tell: a critical review on robustness and uncertainty for a more responsible medical AI","volume":"202","author":"Marconi","year":"2025","journal-title":"Int J Med Inform"},{"key":"B22","doi-asserted-by":"publisher","first-page":"1653631","DOI":"10.3389\/fdgth.2025.1653631","article-title":"A systematic review of ethical considerations of large language models in healthcare and medicine","volume":"7","author":"Fareed","year":"2025","journal-title":"Front Digit Health"}],"container-title":["Frontiers in Digital Health"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdgth.2025.1687820\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,5]],"date-time":"2026-01-05T05:30:04Z","timestamp":1767591004000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdgth.2025.1687820\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,5]]},"references-count":22,"alternative-id":["10.3389\/fdgth.2025.1687820"],"URL":"https:\/\/doi.org\/10.3389\/fdgth.2025.1687820","relation":{},"ISSN":["2673-253X"],"issn-type":[{"value":"2673-253X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,1,5]]},"article-number":"1687820"}}