Transparency and Explainability in AI Credit Scoring Systems: Case Study of Interpretable ML Design in ScoreTech's Five-Layer Model

This study examines the challenges of transparency and explainability in artificial intelligence-based credit scoring by analyzing ScoreTech's peer-to-peer credit intelligence platform in Uzbekistan. As machine learning models increasingly replace traditional credit assessment methods, concerns regarding algorithmic opacity, regulatory compliance, and consumer protection have intensified, particularly in emerging markets with limited formal credit histories. The analysis centers on ScoreTech's five-layer architecture: Identity Confidence Score, Income Integrity, Network Behavior, Machine Learning Risk, and Dynamic Adjustment. This architecture serves as a model for interpretable machine learning design, striking a balance between predictive accuracy and transparency requirements. Using case study methodology, technical analysis, and regulatory review, the study demonstrates that modular credit scoring architectures can achieve high discrimination performance (an area under the curve of 89%) while maintaining interpretability comparable to that of traditional scorecards. The findings suggest that decomposable scoring systems enable feature-level explainability, improve regulatory compliance in emerging markets, and build stakeholder trust through transparent decision-making processes. This research provides empirical evidence that architectural design can reduce the trade-off between accuracy and interpretability in financial technology. The discussion addresses implications for financial inclusion, regulatory policy in developing markets, and the broader adoption of explainable artificial intelligence in high-stakes financial decision-making.