The Fairness of Machine Learning in Insurance: New Rags for an Old Man?

Abstract

Since the beginning of their history, insurers have been known to use data to classify and price risks. As such, they were confronted early on with the problem of fairness and discrimination associated with data. This issue is becoming increasingly important with access to more granular and behavioural data, and is evolving to reflect current technologies and societal concerns. By looking into earlier debates on discrimination, we show that some algorithmic biases are a renewed version of older ones, while others show a reversal of the previous order. Paradoxically, while the insurance practice has not deeply changed nor are most of these biases new, the machine learning era still deeply shakes the conception of insurance fairness.

From bias to black boxes: understanding and managing the risks of AI – an actuarial perspective

“…The fairness of machine learning in insurance: new rags for an old man? (Barry & Charpentier, 2022) The idea of fairness in insurance is fundamentally opposed to a legalistic critique of fairness due to the collective approach taken by insurance. From a legalistic perspective, the necessarily arbitrary reduction of an individual to the data of a class can be seen as a statistical bias.…”

“…In order to adjust correctly, a procedure to adjust a best-estimate price can be used to produce a discrimination-free point estimate or to develop a discrimination-free statistical model where predictive performance is sacrificed to disregard direct and indirect discrimination in an appropriate manner The fairness of machine learning in insurance: new rags for an old man? (Barry & Charpentier, 2022) The idea of fairness in insurance is fundamentally opposed to a legalistic critique of fairness due to the collective approach taken by insurance. From a legalistic perspective, the necessarily arbitrary reduction of an individual to the data of a class can be seen as a statistical bias.…”