SESSION 13. Behaviour-based personalisation in contemporary insurance markets

13. Behaviour-based personalisation in contemporary insurance markets.

Gert Meyers (KU Leuven)
Ine Van Hoyweghen (KU Leuven)
Liz McFall (Open University)
Hugo Jeanningros (Université Paris-Sorbonne)


Big Data is promising a revolution in different societal spheres such as security, health and (online) shopping (Mayer-Schönberger & Cukier, 2013). Massive amounts of personal data (genetic information, shared information from wearable devices, internet behavior information) will become manageable in real-time now or in the near future. In the field of insurance, the best known example of Big Data is usage-based car insurance (Car UBI). Big Data – in the form of predictive modelling, Machine Learning (ML), Artificial Intelligence (AI), Internet of Things (IoT), and other technologies that enable the treatment of ‘personalised data’ – is considered to be a ‘disruptive technology’ (FINEOS, 2014; McKinsey Global Institute, 2013), altering the stabilised insurance practices of risk selection through the introduction of predictive data and modelling and the personalisation of risk.

Hypes and fears of the ‘disruptive’ potentials of predictive modelling in insurance abound. Big Data comes with the promise of reducing insurance costs, more accurate pricing and personalising risk, to support healthy lifestyles, make clients accountable, and/or secure their responsible behaviour celebrating predictive modelling solutions as the ‘new way to be smart’ (Ayres, 2007) or as a desirable shift because ‘the ongoing trends towards real-time risk assessment, product and process simplification and automation could accelerate moves towards more radical business models in insurance’ (Swiss Re 2017, 25). Others fear that predictive data and modelling in insurance would increase inequality and discrimination (O’Neil, 2016), resulting in ‘social sorting’ (Minty, 2014), ‘the end of solidarity’ that characterises European insurance markets (Gayant, 2015) as well as the end of insurance-‘as-we-know-it’ (Llull, 2016).

Never mind their respective intuitive merits, such claims on hypes and fears popularise the idea of big data as a paradigm shift. As such, they often neglect the way predictive data and modelling concretely affect, transform, disrupt or reinforce existing practices. In this session, we want to pay attention to contemporary insurance practices, and more specifically to experimental practices of behaviour-based personalisation. Our hypothesis is that behaviour-based personalisation, as a process driving the domain of insurance, does not simply increase the amount of available data and optimise the processes it is applied to, but also changes our ways of knowing, our ways of social ordering and the way we make decisions (Ewald 1991, 2012, Baker 2002, Meyers & Van Hoyweghen 2017). The practices of traditional insurance are being challenged by new forms and uses of data.

This session will accept contributions presenting research on the challenges surrounding behaviour-based personalisation in insurance, triggered by, but not restricted to, the following general question: how does behaviour-based personalisation affect existing insurance practices and institutions?

More precisely the following specified research questions will be discussed:

  • What is behaviour-based personalisation insurance?
  • Which are the technical, legal and social conditions and considerations for behaviour-based personalisation in insurance practices? How is ‘risk’ enacted as ‘personalised’ in insurance experiments?
  • Which are the technical, legal and social consequences of behaviour-based personalisation in insurance practices?
  • How does the emergence of behaviour-based personalisation in insurance reconfigure the role and responsibilities of the insurer and the insured, as well as the idea of ‘risk’? Can this be considered as a shift away from insurance-‘as-we-know-it’?
  • How does the emergence of behaviour-based personalisation in insurance reconfigure the role of insurance as an institution generating solidarity?

Main objectives

The session has four objectives. It will:

  • Empirically document current involved actors and discourses on behaviour-based personalisation in European insurance markets, as well as investigate its national and supranational regulatory contexts;
  • Empirically study behaviour-based personalisation – processes of data collection, curation, analysis and prediction, including their usage in insurance risk assessment processes – in insurance practices, in terms of its actuarial, legal and social aspects and consequences;
  • Reflect more broadly on the future role of insurance in an era of Big Data and its implications for discrimination, solidarity, and fairness.


Paper 1

Distributed autonomy: smart insurance as a technological imaginary.

Maiju Tanninen (University of Tampere)
Turo-Kimmo Lehtonen (University of Tampere)
Minna Ruckenstein (University of Helsinki)

In this paper, we study the so-called “smart insurance” that combines self-tracking technologies with life insurance. Our main question is how smart insurance enacts futures in relation to insurance practices and relationships. Furthermore, we study how the actual and potential consumers of smart insurance products relate to these new technologies and how they perform and negotiate their degrees of autonomy in self-tracking practices. Through the examination of these practices and both the insurance companies’ and the consumers’ imaginaries related to this technology, we develop the concept of distributed autonomy. The paper is based on interviews with active and potential smart insurance policy-holders and product designers of a Finnish insurance company, conducted in Finland in 2017–2018.

Smart insurance products aim to track and manipulate the insured’s behaviour by utilizing activity wristbands and other sensory devices. The policyholders are encouraged to share their acts of self-determination in health decisions with the devices while the insurance company gathers data on the insured’s activities. The policyholders, for their part, receive free data services; also, the self-tracking devices are meant to aid them in their efforts to have a healthy lifestyle. Our findings demonstrate tensions and possibilities around smart insurance, discussing relationalities that it opens. While asymmetries as regards information flows and control are evident, the case study does not give grounds for a linear story of a disciplinary mechanism. Instead, imaginaries of surveillance, control, self-determination, and freedom are distributed in multiple ways, and in various degrees. The empirical materials make evident that personal autonomy is not a question of either/or, or on/off. It is not as if either the control is externalized and given up to institutions and gadgets, or it is retained by the subject. Rather, we emphasize the wide variety of the degrees of autonomy evident in the practices and in the fantasies related to the scope of possibilities that the new technologies provide.


Paper 2

Who, or what, is the person in insurtech personalizing? Persons, property and the historical classifications of risk.

Liz McFall (University of Edinburgh)

A 2017 TV dramatization, loosely based on a Philip K Dick story, restages the classic double indemnity bind with a non-human ‘Jill’, in the femme fatale role, enticing Ed, a human, to insure his life, creating a motive for his wife to murder him and pair with Jill. Insurance futures – and histories – are replete with such strange arrangements. The human and non-humans insured – ships, enslaved people, houses, plate-glass, elevators, mobile phones, Cyd Charisse’s legs, telematically enhanced young drivers – are arranged and classified as risk bearing categories. In this paper, I use insurtech health and life sector innovations – behaviour-based health insurance, peer to peer insurance, robo-advisors and non-human underwriting using selfies and passive data – to explore how insurtech is speculatively reconfiguring and re-classifying aspects, elements and combinations of human and artificial things. I set these developments against a longer history of risk classification in which the uncertainties surrounding persons, part-persons, property, events were transformed into marketable risks.


Paper 3

Reconsidering the Fairness of the Actuarial Fairness: a Historical Approach.

Antonio J. Heras (Universida Complutense de Madrid)
David Teira (Universidad Nacional de Education a Distancia)
Pierre-Charles Pradier (Université Paris 1 Panthéon-Sorbonne)

The concept of actuarial fairness stems from an Aristotelian tradition in which fairness requires equality in exchange. When dealing with aleatory contracts, this principle evolved, among medieval scholars, into equality in risk. The formalization of this principle gave rise to the concept of mathematical expectation, quantifying the fair price of aleatory contracts. Among these, the quantification of equal risks in annuities and life insurance led to the development of mortality tables, upon which it was possible to calculate actuarial fair prices. Yet, in the two following centuries, we find no agreement about the proper quantification of the risks associated with age. Among the obstacles, we highlight the early awareness of the possibility of adverse selection. When buyers and sellers can manipulate the risk assessment for their own private interests, the actuarial fairness collapses. If there was no objective assessment of individual risks through universal mortality tables, it did not make any sense to hold a standard of fairness based on risk equality. Rather than fair, we find mutually convenient agreements at most.


Paper 4

On the Distribution of the Costs and Benefits of Cooperation: The Case of Insurance Mechanisms.

Xavier Landes (Stockholm School of Economics in Riga)

Social cooperation characterizes situations where individuals interact in a mutually beneficial way, e.g. on markets through Pareto-enhancing exchanges, in companies through division of labor and specialization and within insurance schemes through risk pooling (Heath, 2006). Social cooperation produces benefits that are distributed among participants such as better satisfaction of individual preferences, increased productivity (and therefore increase of the social output), lower expected losses, and so forth. Cooperation also produces costs.

There is a strong view shared among economists and political theorists that the only justified redistribution of costs and benefits of cooperation should take place within the cooperative setting under consideration, e.g. only among policyholders within an insurance scheme. Any external distribution of costs and benefits would therefore be in need of strong justification. This is actually a serious obstacle to any redistributive theory: justifying reallocating resources to individuals who, at first sight, did not contribute to the production of these resources (e.g. unemployed people).
This problem is pervasive in the literature in political theory, John Rawls for instance faces it in the framework of his theory of justice. In this paper, I am offering to contrast what I am labelling (for the moment) internal and external approaches to redistribution of cooperative burdens and benefits. More precisely, I will present the reasons that could support limiting such redistribution to individuals who actively cooperate. Then, I will discuss potential objections to this view. All this discussion will be applied to insurance mechanisms in general, and public insurance in particular.


Paper 5

Epistemic authority and the individualisation of financial risk in UK life insurance.

Arjen van der Heide (University of Edinburgh)

This paper discusses the shift towards the individualisation of financial risk in the context of UK life insurance that has gradually occurred since the 1950s, most notably with the introduction of unit-linked insurance (the insurance equivalent of defined contribution pensions). Drawing on a large volume of documentary material and more than thirty semi-structured qualitative interviews, the paper examines the conditions that enabled and constrained the shift of financial risk embedded in insurance contracts from a ‘collective’ to the individual. At the face of it, unit-linked insurance, which is now the dominant form of life insurance in the UK, fully realises this shift. However, while the introduction of unit-linked insurance promised to completely eradicate the social conflicts baked into traditional forms of insurance (that were to be mediated by actuaries), policyholder demand for something less than a full-scale embrace of financial risk meant that new types of ‘hybrid’ policies were invented in the 1980s and 90s. Regardless, financial risk has increasingly been pushed towards the individual, reducing the need for actuaries to mediate between different groups of policyholders. Subsequently, actuarial responsibilities technical experts, skilled at performing the onerous calculations required to comply with solvency regulation. After having considered the political economy of financial risk in UK life insurance ‒ as defined by Lasswell as revolving around the question of who gets what, when and how ¬‒ this paper concludes by considering the similarities and differences between processes of individualisation and personalisation, and questions whether it is useful to distinguish between the two.


Paper 6

Fairness and Accountability in behavior-based personalization in insurance.

Gert Meyers (KU Leuven)
Katrien Antonio (KU Leuven)
Caroline Van Schoubroeck (KU Leuven)
Ine Van Hoyweghen (KU Leuven)

As in many other domains of social life, hypes and fears on the ‘disruptive’ potential of big data in insurance abound. The use of fine-grained data and machine learning in insurance can result in the ‘personalization of risk’, posing important societal and regulatory challenges on issues of discrimination, privacy, accountability and fairness. To move beyond these hypes and fears, robust empirical research is urgently needed. This paper aims to fill this gap and develops a strong interdisciplinary research program to investigate big data-enabled ‘personalization of risk’ in insurance and study the societal dimensions of big data technologies. To go beyond the idea of a simple, unidirectional ‘adoption’ of big data in insurance, we draw upon conceptual resources stemming from the interdisciplinary field of science and technology studies (STS). The paper investigates real-time experimental practices of big data-enabled personalization of risk in car and health insurance, drawing upon multiple research methods. The interdisciplinary approach is original in bringing out the context-specificity of big data in the practices of insurance. This has so far received little attention in the literature and the collaboration between actuarial scientists, humanities and law scholars adds a unique opportunity to detect and frame how big data transforms the way we know and are acted upon in insurance.


Paper 7

The fairness of Telematics: challenging the use of Personal data for Price discrimination.

Freyja van den Boom, PhD Candidate (Bournemouth University)

As vehicles are increasingly equipped with camera’s and sensor technology, they generate vast amounts of data. This data has proven valuable not only for the car manufacturer to better understand how their vehicles perform but other industries too. Telematics insurance is a good example of an innovative product based on the collection and analysis of vehicle data to make a more accurate assessment of the risk the individual driver poses, based on which insurers can offer a more personalised premium. Telematics as a form of use-based Insurance is presented as being fairer than the more traditional approach of risk pooling where a driver is placed in a larger pool based on more general statistics. Although it may be fairer to some who can benefit from lower premiums not everyone agrees that more personalised pricing will indeed be fair for all parties involved. Concerns have been raised about undesirable market outcomes such as high prices, unavailability of insurance coverage and discrimination, which raises the question whose fairness telematics insurers have in mind when they justify their use of vehicle data for their risk analysis and pricing. This paper examines the notion of fairness used by insurers to promote telematics insurance and use of personal data gathered from vehicles. To contribute to the EU debate on access to in-vehicle data and resources as well as price discrimination, it looks specifically at whether vehicle data processing falls under the purpose for which personal data can be lawfully processed under the General Data Protection Regulation and to what extend consumers have control over ‘their’ vehicle data being used for what is considered ‘fair’ insurance pricing.