Yiseul Choi, Jiwon Hong, Eunbeen Lee, Junga Kim, Seongmin Kim, Journal of Information Processing Systems Vol. 21, No. 1, pp. 89-101, Mar. 2025  

https://doi.org/10.3745/JIPS.01.0111
Keywords: AI Fairness, Bias Mitigation, Data Preprocessing, Fairness Metrics, Financial data
Fulltext: PDF Full Text PubReader

Abstract

As artificial intelligence (AI) increasingly drives decision-making in the financial sector, ensuring fairness in machine-learning models has become critical. Bias in AI models can lead to discriminatory practices, undermining public trust and restricting access to essential financial services. While existing financial services leverage AI to enhance efficiency and accuracy, these systems can inadvertently produce unfair outcomes for specific groups defined by sensitive attributes, such as gender and race. This study addresses the challenge of mitigating bias in loan-approval models by applying fairness-aware machine-learning techniques. We investigate two distinct constraint-based strategies for bias mitigation: fairness- and accuracy-constrained models. These strategies are evaluated using logistic regression (LR) and a large-scale, contemporary financial dataset from the Korea Credit Information Services. The results demonstrate that fairness-constrained models achieve a superior balance between fairness and accuracy compared to a conventional LR model. Furthermore, we highlight the importance of tailored data preprocessing and carefully selecting relevant sensitive attributes (e.g., gender, age, nationality) in enhancing fairness outcomes. The findings underscore the necessity of integrating fairness considerations into every stage of the AI model development lifecycle within finance, ensuring equitable outcomes without compromising predictive performance.

Statistics
Cite this article