The work presents a discussion of theoretical issues and of the regulatory and institutional context for the application of Artificial Intelligence and Machine Learning (AI-ML) to credit scoring. The paper also describes the results of a survey of a selection of Italian financial intermediaries on their experience in adopting such models.
The use of AI-ML models in credit scoring is still limited in Italy, though growing. They are currently used to support decision-making processes, the ultimate responsibility for which rests with the analysts. The financial intermediaries surveyed confirmed that these models perform better than traditional models. The evidence from the survey nevertheless indicates that intermediaries should strengthen their analyses to ensure the absence of potential biases, including those possibly leading to discriminatory outcomes, as well as their governance mechanisms, especially in cases in which they outsource the development of their AI-ML models.
