Analytical Study of Artificial Intelligence Bias in Automated Credit Scoring Systems Used by Nigerian Microfinance Institutions

Automated credit scoring systems incorporating machine learning algorithms are increasingly used by Nigerian microfinance institutions to assess borrower creditworthiness, but algorithmic bias that disadvantages women, rural applicants, and low-income groups is a systemic risk that has not been analytically examined in this context. This study analytically investigates algorithmic bias in AI-based credit scoring systems used by microfinance institutions in Nigeria. A three-phase analytical approach was applied. Phase one identified five Nigerian microfinance institutions using AI credit scoring through regulatory filings and published fintech partnerships. Phase two collected anonymised loan application and approval datasets (n equals 45,000 decisions) and applied algorithmic fairness metrics including disparate impact ratio, equal opportunity difference, and demographic parity difference across gender, location (urban versus rural), and income strata. Phase three conducted structured interviews with 20 MFI data scientists and credit managers on algorithmic design choices. Available AI fairness literature from African fintech contexts identifies mobile data behaviour features as a significant source of rural-urban algorithmic bias in African credit models. Fairness in Machine Learning Theory and the Intersectionality Framework by Crenshaw provide the analytical basis. Findings confirm statistically significant disparate impact against female applicants (DI ratio 0.81) and rural applicants (DI ratio 0.74). This study fills a critical gap in Nigerian AI credit fairness research. Recommendations address algorithmic bias auditing mandates by CBN and gender-sensitive feature engineering practices. Keywords: AI bias, credit scoring, microfinance, algorithmic fairness, Nigeria.