Book chapter
Balancing Fairness and Performance in Healthcare AI: A Gradient Reconciliation Approach
Artificial Intelligence in Medicine, v 15734, pp 480-489
23 Jun 2025
Abstract
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
2 Record Views
Details
- Title
- Balancing Fairness and Performance in Healthcare AI: A Gradient Reconciliation Approach
- Creators
- Xiaoyang WangChristopher C. Yang
- Contributors
- Riccardo Bellazzi (Editor)José Manuel Juarez Herrero (Editor)Lucia Sacchi (Editor)Blaž Zupan (Editor)
- Publication Details
- Artificial Intelligence in Medicine, v 15734, pp 480-489
- Series
- Lecture Notes in Computer Science
- 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 (Informatics); College of Computing and Informatics
- Web of Science ID
- WOS:001553200500047
- Scopus ID
- 2-s2.0-105009763956
- Other Identifier
- 991022059873304721