Stakeholder Analysis of Explainability Requirements in Automated Credit Scoring Systems

Abstract

Automated credit scoring systems increasingly rely on machine learning models that improve risk prediction but create explainability challenges for stakeholders affected by or responsible for lending decisions. This study analyzes stakeholder-specific explainability requirements in automated credit scoring systems using a governance-oriented stakeholder analysis framework. The analysis involved seven stakeholder groups: credit applicants, loan officers, data scientists, compliance officers, regulators, auditors, and consumer advocates. Results show that regulators obtained the highest overall salience, with power, legitimacy, urgency, and risk exposure scores of 9.3, 9.5, 8.5, and 6.8, respectively. Credit applicants also showed high salience, driven by urgency of 9.1 and risk exposure of 9.4, despite a lower power score of 3.8. Requirement intensity analysis showed that applicants prioritized actionability at 4.9 and contestability at 4.8, while regulators assigned the highest scores to fairness justification and auditability, both reaching 5.0. Compliance officers similarly prioritized auditability at 4.9 and fairness at 4.8. The prioritization results identified reason codes, appeal mechanisms, decision logs, fairness monitoring, and adverse action notices as the most critical requirements. Reason codes achieved governance value of 9.1 and feasibility of 8.8, while fairness monitoring achieved the highest governance value of 9.4 but lower feasibility of 6.3. Validation results confirmed the methodological robustness of the framework, with expert relevance scoring 0.92, stakeholder confirmability 0.88, intercoder agreement 0.86, documentary consistency 0.84, and framework completeness 0.89. These findings demonstrate that explainability in credit scoring must be designed as a layered governance mechanism rather than a single technical output. The study contributes a stakeholder-sensitive requirement framework for aligning automated credit scoring with transparency, fairness, auditability, and procedural accountability.