Explainable Artificial Intelligence in Credit Risk: A Systematic Evidence Map for Regulated Credit Decisions

Machine learning increasingly informs high-stakes credit decisions, yet model opacity complicates transparency and model governance. This review combines a systematic evidence map of explainable artificial intelligence (XAI) in credit risk with a task-oriented narrative synthesis. A Scopus search covering 2017–2026 produced 958 unique records; a frozen title-and-abstract codebook assigned 550 to the core map. SHAP was named in 39.1% of core titles or abstracts, whereas 1.3–1.5% contained strict or broad explanation-quality evaluation signals. A targeted full-text check identified both unsupported signals and evaluation practices missed by metadata coding. The percentages therefore describe reporting visibility rather than full-text practice. The task–artefact–method–validation (TAMV) framework connects explanation outputs to adverse-action reasons, applicant recourse, model review, and documentation, specifying the evidence each use requires. A proposed finance-specific benchmark translates this synthesis into testable priorities. The review was not prospectively registered. Its contribution is a reproducible reporting map and task-specific validation agenda, not a comparative ranking of explanation methods.