Explainable AI for paid-up risk management in life insurance products

Abstract

Explainable artificial intelligence (xAI) provides a better understanding of the decision-making processes and results generated by black-box machine learning (ML) models. Here, we outline several xAI techniques in order to equip risk managers with more explainable ML methods. We illustrate this by describing an application for the more effective management of paid-up risk in insurance savings products. We draw on a database of real universal life policies to fit an initial logistic regression model and several tree-based models. We then use different xAI techniques, including a novel approach that leverages a Kohonen network of Shapley values, to offer valuable perspectives on tree-based models to the end-user. Based on these findings, we show how non-trivial ideas can emerge to improve paid-up risk management.