Enhancing Credit Decision Transparency for Small Business Owners: An Explainable AI Approach to Mitigate Algorithmic Bias in Micro-lending

Abstract

The proliferation of artificial intelligence in small business lending has created unprecedented challenges regarding algorithmic transparency and fairness. Traditional credit assessment models exhibit significant opacity, preventing small business owners from understanding rejection rationales and potentially perpetuating discriminatory practices. This research presents a novel probabilistic explainability framework that integrates SHAP-enhanced feature attribution with specialized bias detection metrics tailored for micro-lending contexts. Our approach addresses the critical gap between high-performing machine learning models and regulatory compliance requirements for transparent financial decision-making. The proposed methodology combines advanced interpretability techniques with multi-objective optimization to maintain predictive accuracy while ensuring algorithmic fairness. Experimental validation demonstrates substantial improvements in explanation quality and bias reduction across diverse small business lending scenarios. The framework provides actionable insights for loan officers while enhancing trust among small business applicants. This work contributes to the emerging field of responsible AI in financial services by establishing technical standards for explainable credit assessment. The research implications extend beyond individual lending decisions to inform broader policy discussions regarding algorithmic accountability in financial inclusion initiatives.