Machine learning in bankruptcy risk assessment: Insights from risk management experts in Poland

Abstract

This paper presents the results of a qualitative study based on in-depth interviews with certified risk management experts — Financial Risk Manager (FRM), Professional Risk Manager (PRM), and Credit and Counterparty Risk Manager (CCRM) from Poland. The findings reveal that despite an increased focus on machine learning (ML) methods in research, classic statistical methods, such as discriminant analysis and logistic regression, remain in use. The reason lies in the structural barriers that inhibit the adoption of ML models in bankruptcy assessment. Key limitations include the lack of transparency and difficulty in explaining and justifying model decisions to regulators and courts (‘black box’ effect1), insufficient availability of high-quality data, and complex implementation requirements involving IT systems,2 supervision processes, and governance frameworks. Experts also report limited trust in algorithmic decision making and perceive ML as a technology often driven by marketing narratives rather than demonstrated added value in insolvency prediction. This study offers a critical insight into the predictive superiority, which, when considered independently, is insufficient for adoption. The effectiveness of ML in financial risk assessment depends on its alignment with regulatory expectations for explainability, interpretability, legal accountability, and organisational trust. By incorporating certified expert perspectives, this paper contributes a missing dimension, outlining the conditions under which ML can be responsibly and realistically applied to advance bankruptcy risk management in Poland. This article is also included in The Business & Management Collection which can be accessed at https://hstalks.com/business/.