Balancing Fairness and Performance in Healthcare AI: A Gradient Reconciliation Approach

Xiaoyang Wang; Christopher C. Yang

doi:10.1007/978-3-031-95838-0_47

Back

Balancing Fairness and Performance in Healthcare AI: A Gradient Reconciliation Approach

Book chapter

Open access

Balancing Fairness and Performance in Healthcare AI: A Gradient Reconciliation Approach

Xiaoyang Wang and Christopher C. Yang

Artificial Intelligence in Medicine (AIME 2025), v 15734, pp 480-489

23 Jun 2025

DOI: https://doi.org/10.1007/978-3-031-95838-0_47

Files and links (1)

url

https://doi.org/10.48550/arXiv.2504.14388View

Preprint (Author's original) Open CC BY-NC V4.0

Abstract

Conflict Gradient

Healthcare AI

Multi-attribute Fairness

Sepsis mortality prediction

Substance Use Disorder

The rapid growth of healthcare data and advances in computational power have accelerated the adoption of artificial intelligence (AI) in medicine. However, AI systems deployed without explicit fairness considerations risk exacerbating existing healthcare disparities, potentially leading to inequitable resource allocation and diagnostic disparities across demographic subgroups. To address this challenge, we propose FairGrad, a novel gradient reconciliation framework that automatically balances predictive performance and multi-attribute fairness optimization in healthcare AI models. Our method resolves conflicting optimization objectives by projecting each gradient vector onto the orthogonal plane of the others, thereby regularizing the optimization trajectory to ensure equitable consideration of all objectives. Evaluated on diverse real-world healthcare datasets and predictive tasks—including Substance Use Disorder (SUD) treatment and sepsis mortality—FairGrad achieved statistically significant improvements in multi-attribute fairness metrics (e.g., equalized odds) while maintaining competitive predictive accuracy. These results demonstrate the viability of harmonizing fairness and utility in mission-critical medical AI applications.

Metrics

10 Record Views

Details

Title: Balancing Fairness and Performance in Healthcare AI: A Gradient Reconciliation Approach
Creators: Xiaoyang Wang - Drexel University
Christopher C. Yang - Drexel University
Contributors: Riccardo Bellazzi (Editor)
José Manuel Juarez Herrero (Editor)
Lucia Sacchi (Editor)
Blaž Zupan (Editor)
Publication Details: Artificial Intelligence in Medicine (AIME 2025), v 15734, pp 480-489
Conference: 23rd International Conference, AIME 2025, 23rd (Pavia, Italy, 23 Jun 2025–26 Jun 2025)
Series: Lecture Notes in Computer Science; 15734
Publisher: Springer Nature Switzerland; Cham
Number of pages: 10
Grant note: National Science FoundationDepartment of Defense: DoD W91XWH-05-1-023
This work was supported in part by the National Science Foundation under the Grants IIS-1741306 and IIS-2235548, and by the Department of Defense under the Grant DoD W91XWH-05-1-023. This material is based upon work supported by (while serving at) the National Science Foundation. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
Resource Type: Book chapter
Language: English
Academic Unit: Information Science; School of Computer and Information Sciences
Web of Science ID: WOS:001553200500047
Scopus ID: 2-s2.0-105009763956
Other Identifier: 991022059873304721

InCites Highlights

Data related to this publication, from InCites Benchmarking & Analytics tool:

Web of Science research areas: Computer Science, Artificial Intelligence; Computer Science, Interdisciplinary Applications; Medical Informatics