Search for a command to run...
Over 150 artificial intelligence (AI) applications in financial services caught on within 2020–2025 due to digital transformation during and after the pandemic. Credit scoring systems that use machine learning (ML) to predict credit scores have enhanced predictive accuracy, which can be over 85–90 percent AUC on standard benchmark dataset, but are associated with significant issues of privacy, lack of transparency, and bias. Federated learning (FL) became a privacy-preserving framework, which allows training a model in a decentralized manner that does not require the sharing of raw data. Simultaneously, explainable AI (XAI) methods, including SHAP, LIME, and counterfactual reasoning, became popular in explaining complex models and making them regulator friendly. As legal systems, such as the EU AI Act and fair lending legislation in the United States, have established a set of legal standards to guarantee fairness, the necessity to measure these trade-offs between privacy, interpretability, and equity has become more pronounced. Even with the swift progress, studies in FL, XAI, and fairness were not connected to one another, but did not have unified benchmarks and cross-domain solutions. The paper includes a systematic review of 25 peer-reviewed studies which include 20 peer-reviewed articles and 5 contextual sources articles that have been published in 2020–25 and converge at the intersection of credit scoring. It determines that there are long-standing gaps, such as fairness-conscious aggregation in FL, explainability in privacy limits and bias detection across institutions. To overcome them, we introduce Federated Explainable AI (FedXAI), a conceptual framework that integrates the metrics of fairness (e.g. equal opportunity), explainability instruments (e.g. SHAP fidelity scores), and differential privacy into the federated learning cycle. FedXAI promotes inclusive, auditable, and human-centered credit scoring, which forms the basis of benchmarking in the future, regulatory alignment, and practical implementation.