Journal of Economics Intelligence And Technology

Abstract: The application of artificial intelligence and machine learning to credit scoring has transformed traditional financial risk assessment but has raised serious concerns about algorithmic bias, fairness, and transparency. As machine-learning models increasingly determine credit decisions, the ethical considerations and governance structures that underpin them are critical to fintech inclusivity and regulatory compliance. This narrative review condenses recent research on bias detection, fairness methods, and explainability techniques for credit scoring models, while considering the underlying challenges to algorithmic decision-making in the U.S. financial system. Several key findings are highlighted: Recent research demonstrates significant advances in fairness-enhancing interventions, such as pre-processing bias mitigation, in-processing fairness constraints, and post-processing calibration methods. However, persistent challenges remain, including ongoing trade-offs between predictive performance and fairness metrics, providing meaningful explainability for complex ensemble models, and addressing intersectional discrimination. The integration of alternative data sources can increase inclusion, but risks introducing new forms of bias. While technological improvements have advanced bias detection and mitigation in credit scoring, tensions persist among fairness definitions, model performance, and regulatory demands. The review concludes by noting the importance of accounting for the context-dependency of fairness, improving evaluation schemes for real-world use, and establishing governance frameworks for algorithm accountability.