Journal of Accounting and Management Vision

Journal of Accounting and Management Vision

The need to use explainable artificial intelligence models in banking

Document Type : Original Article

Author
Master of Accounting, employee of the Development and Cooperative Bank of Bushehr Province
Abstract
The widespread use of artificial intelligence and machine learning algorithms in the banking industry has created significant opportunities for improving credit risk assessment, fraud detection, customer validation, and optimizing financial decisions. However, the increasing complexity of predictive models, especially deep learning-based algorithms and aggregation methods, has made the issue of transparency and explainability of algorithmic decisions one of the fundamental challenges of data-driven banking. In an environment where automated decisions can affect individuals’ access to financial facilities, the cost of financing, the security of assets, and the operational stability of credit institutions, relying on predictive accuracy metrics alone is not sufficient to assess the adequacy of an AI model.The present study aims to explain the necessity of using explainable AI in banking by critically reviewing the scientific literature on algorithmic explainability, model risk management, fairness in credit decision-making, regulatory compliance, and stakeholder trust. Based on this analysis, a conceptual framework is proposed in which explainability contributes to the responsible governance of AI in banks through five main mechanisms: increasing decision transparency, strengthening auditability, improving bias detection, promoting model risk control, and facilitating human oversight. The paper also shows that the use of methods such as SHAP, LIME, and inherently interpretable models, when accompanied by validation of explanations, model documentation, and continuous monitoring, can reduce the gap between predictive performance and operational and regulatory requirements of banking.However, explainability does not by itself guarantee fairness, accuracy, or legal compliance and must be applied within a comprehensive model lifecycle management system. The proposed framework provides a theoretical basis for designing future empirical research and developing policies for the deployment of trustworthy AI in the banking industry
Keywords
Subjects