Assessing Fairness and Transparency in Credit Scoring Using Explainable AI

As financial institutions rely on machine learning (ML) for credit scoring, ensuring fairness and transparency in loan decisions has become increasingly important. This paper explores how SHapley Additive exPlanations (SHAP) values can help assess the fairness of credit decisions, particularly across different demographic groups. We use the public Lending Club dataset and apply ML models, such as XGBoost, LightGBM, and Random Forest, to evaluate borrowers' creditworthiness. Our analysis shows that SHAP values can serve as a powerful, model-agnostic tool for understanding how individual features contribute to credit decisions and for promoting fairness in these models. By providing insights into both local and global model behavior, SHAP helps the decision-making process to be more transparent. These findings highlight the need for clearer, more equitable credit scoring systems, offering a valuable framework for stakeholders and regulators who are committed to ethical AI practices in the financial sector.