A Plug-and-Play Explainable AI Framework in Credit Score Assessment

Credit scoring is fundamental in enabling institutions to distinguish between creditworthy and high-risk borrowers. Thus, robust credit scoring models are essential for effective financial risk management. Artificial intelligence and machine learning have significantly improved the precision of credit decision support systems, however, these models often operate as black boxes, concealing the underlying logic of credit denials or approvals. To address this challenge and meet the demand for financial transparency, explainable AI (XAI) has emerged as a crucial mechanism for generating decisions that are interpretable to both financial institutions and clients. This paper researched contemporary decision support systems (DSS) used in credit risk assessment and proposes a modular, plug-and-play framework with integrated explainability. Six requirements were derived for a robust system design based on a review of existing XAI frameworks. The framework consists of five interconnected components such as data ingestion layer, feature processing module, a supervised predictive model, an explanation layer, and a decision endpoint. Existing studies often focus on individual predictive models or explanation methods, leaving their integration and component interchangeability unspecified. The proposed framework addresses this gap through explicit interfaces that allow predictive models and explanation methods to be replaced independently. The system will operate with both traditional credit data, alternative data, or both. Thus, feature processing is incorporated to distinguish the importance and relevance of the data to credit assessment decisions. Any irrelevant redundant data is not be considered by the model. The contribution of a study is a reusable architectural specification. The proposed XAI framework is designed to serve as a universal template for researchers and financial institutions developing transparent, effective credit scoring systems.