With their intensive use of data to classify and price risk, insurers have often been confronted with data-related issues of fairness and discrimination. This paper provides a comparative review of discrimination issues raised by traditional statistics versus machine learning in the context of insurance. We first examine historical contestations of insurance classification, showing that it was organized along three types of bias: pure stereotypes, non-causal correlations, or causal effects that a society chooses to protect against, are thus the main sources (...) of dispute. The lens of this typology then allows us to look anew at the potential biases in insurance pricing implied by big data and machine learning, showing that despite utopic claims, social stereotypes continue to plague data, thus threaten to unconsciously reproduce these discriminations in insurance. To counter these effects, algorithmic fairness attempts to define mathematical indicators of non-bias. We argue that this may prove insufficient, since as it assumes the existence of specific protected groups, which could only be made visible through public debate and contestation. These are less likely if the right to explanation is realized through personalized algorithms, which could reinforce the individualized perception of the social that blocks rather than encourages collective mobilization. (shrink)

Algorithmic predictions are promising for insurance companies to develop personalized risk models for determining premiums. In this context, issues of fairness, discrimination, and social injustice might arise: Algorithms for estimating the risk based on personal data may be biased towards specific social groups, leading to systematic disadvantages for those groups. Personalized premiums may thus lead to discrimination and social injustice. It is well known from many application fields that such biases occur frequently and naturally when prediction models are applied to (...) people unless special efforts are made to avoid them. Insurance is no exception. In this paper, we provide a thorough analysis of algorithmic fairness in the case of insurance premiums. We ask what “fairness” might mean in this context and how the fairness of a premium system can be measured. For this, we apply the established fairness frameworks of the fair machine learning literature to the case of insurance premiums and show which of the existing fairness criteria can be applied to assess the fairness of insurance premiums. We argue that two of the often-discussed group fairness criteria, independence (also called statistical parity or demographic parity) and separation (also known as equalized odds), are not normatively appropriate for insurance premiums. Instead, we propose the sufficiency criterion (also known as well-calibration) as a morally defensible alternative that allows us to test for systematic biases in premiums towards certain groups based on the risk they bring to the pool. In addition, we clarify the connection between group fairness and different degrees of personalization. Our findings enable insurers to assess the fairness properties of their risk models, helping them avoid reputation damage resulting from potentially unfair and discriminatory premium systems. (shrink)

The adoption of Big Data analytics in insurance has proved controversial but there has been little analysis specifying how insurance practices are changing. Is insurance passively subject to the forces of disruptive innovation, moving away from the pooling of risk towards its personalisation or individualisation, and what might that mean in practice? This special theme situates disruptive innovations, particularly the experimental practices of behaviour-based personalisation, in the context of the practice and regulation of contemporary insurance. Our contributors argue that behaviour-based (...) personalisation in insurance has different and broader implications than have yet been appreciated. BDAs are changing how insurance governs risk; how it knows, classifies, manages, prices and sells it, in ways that are more opaque and more extensive than the black boxes of in-car telematics. (shrink)

What are the challenges of turning data subjects into research participants—and how can we approach this task responsibly? In this paper, we develop a methodology for studying the lived experiences of people who are subject to automated scoring systems. Unlike most media technologies, automated scoring systems are designed to track and rate specific qualities of people without their active participation. Credit scoring, risk assessments, and predictive policing all operate obliquely in the background long before they come to matter. In doing (...) so, they constitute a problem not only for those subject to these systems but also for researchers who try to study their experience. Specifically, we identify three challenges that are distinct to studying experiences of automated scoring: limited awareness, embeddedness, and ongoing inquiry. Starting from the observation that coming to terms with one's position as a data subject constitutes a form of learning in its own right, we propose a research strategy called critical companionship. Originally articulated in the context of nursing research, critical companionship invites us to accompany a data subject over time, paying critical attention to how the participant's and the researcher's inquiries complicate and constitute each other. We illustrate the strengths and limitations of this methodology with materials from a recent study we conducted about people's credit repair practices and sketch a set of sensibilities for studying contemporary scoring systems from the margins. (shrink)