FAIM: Fairness-aware interpretable modeling for trustworthy machine learning in healthcare
Mingxuan Liu, Yilin Ning, Yuhe Ke, Yuqing Shang, Bibhas Chakraborty, Marcus Eng Hock Ong, R. S. Vaughan, Nan Liu
Abstract
The escalating integration of machine learning in high-stakes fields such as healthcare raises substantial concerns about model fairness. We propose an interpretable framework, fairness-aware interpretable modeling (FAIM), to improve model fairness without compromising performance, featuring an interactive interface to identify a "fairer" model from a set of high-performing models and promoting the integration of data-driven evidence and clinical expertise to enhance contextualized fairness. We demonstrate FAIM's value in reducing intersectional biases arising from race and sex by predicting hospital admission with two real-world databases, the Medical Information Mart for Intensive Care IV Emergency Department (MIMIC-IV-ED) and the database collected from Singapore General Hospital Emergency Department (SGH-ED). For both datasets, FAIM models not only exhibit satisfactory discriminatory performance but also significantly mitigate biases as measured by well-established fairness metrics, outperforming commonly used bias mitigation methods. Our approach demonstrates the feasibility of improving fairness without sacrificing performance and provides a modeling mode that invites domain experts to engage, fostering a multidisciplinary effort toward tailored AI fairness.