Reducing Biases towards Minoritized Populations in Medical Curricular Content via Artificial Intelligence for Fairer Health Outcomes

Chiman Salavati, Shannon Song, Willmar Sosa Diaz, Scott A. Hale, Roberto Montenegro, Fabricio Murai, Shiri Dori-Hacohen

Abstract: Biased information (recently termed bisinformation) contin- ues to be taught in medical curricula, often long after having been debunked. In this paper, we introduce BRICC, a first- in-class initiative that seeks to mitigate medical bisinforma- tion using machine learning to systematically identify and flag text with potential biases, for subsequent review in an expert-in-the-loop fashion, thus greatly accelerating an other- wise labor-intensive process. A gold-standard BRICC dataset was developed throughout several years, and contains over 12K pages of instructional materials. Medical experts metic- ulously annotated these documents for bias according to com- prehensive coding guidelines, emphasizing gender, sex, age, geography, ethnicity, and race. Using this labeled dataset, we trained, validated, and tested medical bias classifiers. We test three classifier approaches: a binary type-specific clas- sifier, a general bias classifier; an ensemble combining bias type-specific classifiers independently-trained; and a multi- task learning (MTL) model tasked with predicting both gen- eral and type-specific biases. While MTL led to some im- provement on race bias detection in terms of F1-score, it did not outperform binary classifiers trained specifically on each task. On general bias detection, the binary classifier achieves up to 0.923 of AUC, a 27.8% improvement over the base- line. This work lays the foundations for debiasing medical curricula by exploring a novel dataset and evaluating differ- ent training model strategies. Hence, it offers new pathways for more nuanced and effective mitigation of bisinformation.