THE IMPACT OF ARTIFICIAL INTELLIGENCE ON CONSUMER CREDIT SCORING MODELS: A CRITICAL REVIEW OF FAIRNESS AND BIAS IN ALGORITHMS

Abstract

The rapid integration of artificial intelligence into consumer credit scoring has transformed traditional risk assessment practices by improving predictive accuracy and operational efficiency. However, growing concerns regarding algorithmic fairness, transparency, and bias have raised critical ethical and regulatory challenges for automated lending systems. This study provides a comprehensive empirical and critical evaluation of AI-driven credit scoring models with a particular focus on fairness and bias across demographic and socio-economic groups. Using a mixed-methods experimental design, the research compares traditional statistical models with advanced machine learning approaches, including ensemble and neural network models, across multiple performance and fairness metrics. Quantitative results demonstrate that while AI-based models consistently outperform conventional approaches in terms of accuracy and discriminatory power, they also exhibit measurable disparities in approval outcomes for protected groups. Fairness metrics such as disparate impact and demographic parity reveal systematic bias amplification in unconstrained models, whereas fairness-aware modeling strategies significantly mitigate these disparities at the cost of marginal reductions in predictive performance. Qualitative analysis further contextualizes these findings within existing ethical AI frameworks and regulatory expectations. Overall, the study highlights the inherent trade-off between predictive efficiency and ethical responsibility in AI-driven credit scoring and underscores the need for balanced, fairness-conscious model deployment in consumer finance.