Fairness challenges in insurance premiums: investigating customer profiling algorithmic biases

Abstract

Insurance premium pricing has been a great concern across multiple stakeholders, including actuaries, insurers, policyholders, the justice system, and society, due to the issue of discrimination. The evolution of the insurance industry, marked by significant transformations over time, is currently transitioning towards automated systems. While actuarial considerations guide the calculation of premiums, ethical concerns arise as certain practices may be perceived as unfair by society and the justice system, even when justified from an actuarial standpoint. Some of the types of discrimination are gender-based, age-based, and preexisting condition discrimination. The aim of this paper is to provide an in-depth analysis of the ethical issue and provide recommendations on building customer profiling algorithms that prioritize fairness and eliminate discrimination. The case study presented in this paper involves the Test Achats case in Belgium, where gender-based pricing in insurance was legally challenged, leading to the removal of gender as a factor in premium calculations since 2012. Additionally, the integration of telematics in auto insurance, exemplified by products like Fairzekering, showcases efforts to monitor driving habits for personalized discounts. To navigate these complexities, a value-sensitive design matrix is proposed, outlining the impact of each value on various stakeholders, supplemented by recommendations derived from literature and case critiques. This holistic approach aims to offer a fair and transparent insurance pricing landscape while addressing the ethical implications of discrimination in customer profiling algorithms.