Why Was My Loan Denied?

This article demonstrates the practical application of ethical principles in artificial intelligence within the context of automated loan approval decisions. It addresses the increasing reliance on predictive models in financial institutions, highlighting the critical concerns of fairness and accountability inherent in such systems. A machine learning model, specifically a Gradient Boosting Classifier, is built using a publicly available loan prediction dataset, encompassing various applicant attributes such as credit history, income, and demographic information. The model's predictions are then audited using two explainable AI (XAI) techniques: LIME (Local Interpretable Model-Agnostic Explanations) and SHAP (SHapley Additive exPlanations). LIME provides case-specific explanations, clarifying why individual loan applications were approved or rejected, enhancing transparency for both customers and regulators. SHAP offers both local and global explanations, providing insights into individual decisions and revealing broader patterns in the model's behavior across the entire dataset, enabling internal audits and governance. The combined use of LIME and SHAP facilitates a comprehensive understanding of the model's decision-making process, promoting fairness, accountability, and compliance with ethical and regulatory standards. The article concludes by emphasizing the importance of integrating these explainability practices into production systems and broader business processes to ensure responsible and ethical use of AI in financial decision-making, aligning technical innovation with legal and ethical imperatives. The study also suggests future research directions, including continuous bias auditing, integration into business workflows, and extending these methods to other high-stakes domains.