Previous research

Over the past decade, sociological research into the personalisation strategies implemented by insurance companies—and specifically into life and health insurance based on the use of self-tracking technologies and behavioural data processing techniques—has intensified. In this section, we engage with the relevant literature and briefly outline its main findings.

The personalisation of insurance coverage has been the subject of extensive research from both a technical–actuarial and a sociological viewpoint. The use of predictive algorithms to process behavioural data generated by self-tracking technologies has led many observers to predict a disruptive change in the insurance industry, bringing with it both opportunities and significant risks [Braun and Schreiber Reference Braun and Schreiber2017]. Sociological research shows that the aspiration to personalise insurance products on the basis of increasingly granular risk profiles is by no means new within the sector [Krippner and Hirschman Reference Krippner and Hirschman2022; Krippner Reference Krippner2023]. At the same time, scholars have acknowledged that the potential inherent in digital technologies—and the corresponding access to biometric and behavioural data that can be processed through sophisticated computer-based actuarial procedures—represents a genuine novelty that cannot be overlooked, and whose possible consequences remain largely unexplored [Corlosquet-Habart and Janssen Reference Corlosquet-Habart and Janssen2018; Frary Reference Frary2019; Cevolini and Esposito Reference Cevolini and Esposito2022; Charpentier and Vamparys Reference Charpentier and Vamparys2025].

Important studies have focused on the relationship between the personalisation of premiums and the role of financial mutuality that constitutes the crucial solidarity function of insurance [Braun and Nürnberg Reference Braun and Nürnberg2015; Cevolini and Esposito Reference Cevolini and Esposito2020]. This issue is closely related to a lively and complex discussion regarding the problem of fairness in behavioural insurance policies [Meyers and Van Hoyweghen Reference Meyers and Van Hoyweghen2018; Barry Reference Barry2020]. Personalisation, in fact, entails discrimination, and a premium that may be fair from a behavioural perspective is not necessarily fair from an actuarial standpoint. The findings of this increasing body of research have not led to a shared consensus.

On one side, many sociologists and actuarial mathematicians have argued that the promise of personalising insurance policies has not been fulfilled. Rate-making techniques remain largely based on traditional actuarial procedures, and the much-anticipated “revolution” in the insurance field has never materialised [Meyers Reference Meyers2018; Barry and Charpentier Reference Barry and Charpentier2020; François and Voldoire Reference François and Voldoire2022; Charpentier and Vamparys Reference Charpentier and Vamparys2025]. On the other side, other researchers have pointed out that insurance companies may in fact be less interested in the personalisation of premiums—and therefore in actuarial procedures producing discriminatory outcomes—than is commonly assumed. The real novelty, they suggest, may lie in the personalisation of the relationship between insurers and insured persons for preventive purposes, and in the resulting difficulty to balance insurance protection and the responsibilisation of policyholders [Guillén and Cevolini Reference Guillén and Cevolini2021; Presset and Tanninen Reference Presset and Tanninen2025; Cevolini and Esposito Reference Cevolini and Esposito2026].

With regard to the use of self-tracking technologies in behavioural life and health insurance, the research conducted by Hugo Jeanningros and Liz McFall was particularly instructive in the early stages of our work. These studies take as their empirical reference model the behavioural insurance programme developed by the South African insurance company, Discovery Limited, with its Vitality policy. The structure of this programme and its penetration into the French insurance market were investigated in depth by Hugo Jeanningros [Reference Jeanningros2020]. His analysis shows that one of the crucial problems still concerns the level of engagement among policyholders, which is clearly disappointing compared with the company’s expectations, and does not exceed the threshold of 30% of eligible customers [see also Jeanningros Reference Jeanningros2021].

The same insurance programme was analysed in collaboration with Liz McFall in a comparative perspective with the UK market, also taking into account the different regulatory constraints across countries. The central thesis advanced by Jeanningros and McFall [Reference Jeanningros and McFall2020, esp.: 68 and 70] is that insurance companies’ investments in behavioural programmes are explained less by an effective improvement in the lifestyle—and thus the health—of their policyholders than by an alternative form of brand promotion in an increasingly competitive insurance market.

This thesis is also endorsed by Jathan Sadowski et al. [Reference Sadowski2024]. Like Jeanningros and McFall, Sadowski et al. [Reference Sadowski2024, esp.: 230ff.] focus on the Vitality policy. Their empirical research is based on ten formal interviews and more than one hundred informal conversations conducted between 2022 and 2023 at major industry conferences during ethnographic fieldwork. The authors’ analysis is rather critical: Sadowski et al. [Reference Sadowski2024: 233] share the view that practices of personalisation and incentive programmes contingent on policyholders’ behavioural change are “primarily branding tactics”, while the supposed hyper-personalisation and actuarial discrimination “is still more aspirational than actual”. The entire prevention programme might therefore be “an ideological façade” designed to increase the company’s market share, although this does not rule out the possibility that behavioural interventions and tracking technologies may become more effective in the future [Sadowski et al. Reference Sadowski2024: 246; see also Sadowski Reference Sadowski2024].

Particularly relevant to our topic is the research conducted by Tanninen et al. [Reference Tanninen2021] on two Finnish insurance companies offering self-tracking technologies and coaching programmes within life insurance policies. Their analysis largely deflates the narrative of the “digital disruption” of the insurance sector and instead points to the technologically mediated relationship between insurance companies and their policyholders. Self-tracking initiatives and access to behavioural data, they argue, may lead to discriminatory practices in rate-making technologies and to a form of “dataveillance” (i.e. surveillance based on data) to the detriment of personal autonomy and self-determination [see also Tanninen et al. Reference Tanninen2022a]. That this risk does not yet appear particularly alarming depends largely on the limited success of the behavioural life and health insurance policies offered by the companies examined.

Our work connects this strand of research on insurance with recent work in organisation studies and science and technology studies that analyse the organisational effects of digital technologies as ongoing processes of adaptation, negotiation, and situated use [Barley Reference Barley1986; Orlikowski Reference Orlikowski1992; Reference Orlikowski2000]. Digital technologies reshape organisational routines, but they are simultaneously reshaped by the practices, expectations, and heterogeneous responses of their users.

Our findings extend this perspective in an important direction. Rather than simply analysing the implementation of digital technologies within insurance, we focus on proactivity by examining the implementation of an algorithmic feedback loop between organisations and their customers. Insurance companies know that meaningful behavioural change requires sustained long-term engagement. At the same time, they are aware that such engagement depends on users’ familiarity with the app and their motivation to continuously use the self-tracking devices associated with it. As a substantial body of research has shown [Ledger and McCaffrey Reference Ledger and McCaffrey2017; Kangovi and Asch Reference Kangovi and Asch2018; Spender et al. Reference Spender2019; Hovard Reference Hovard2023; Cevolini et al. Reference Cevolini2025], the main challenge is to counteract the high rate of customer drop-off in the use of self-tracking devices and the consequent decline in engagement with the app. The challenge, in other terms, is to achieve sustained long-term engagement in behavioural improvement by fostering both in-app engagement and the regular use of self-tracking devices.

Because different users interact differently with the same app and have a different degree of familiarity with self-tracking devices, the effective implementation of these technologies depends on a continuous recalibration of the interaction between predictive algorithms and policyholders. As our empirical research shows, insurance companies do not merely deploy proactive technologies; they continually redesign thresholds, incentives, communication strategies, and recommendations in response to the behaviour generated by the very systems they have introduced. Proactivity thus emerges as an ongoing process of (re)calibration, in which the programs of insurance companies and the practices of policyholders recursively reshape one another. The key concept underpinning this process of continuous recalibration is “personalisation”.

Data and methods

Our analysis draws on scholarship in the sociology of insurance and of quantification, STS, actuarial studies, on close reading of scientific publications and insurance industry reports, and on semi-structured interviews with selected representatives of our field of inquiry. After a thorough investigation of the sector and several preliminary meetings, we contacted a diverse range of stakeholders, who provided information on various relevant aspects of insurance innovation.

Between June 2023 and April 2024, we conducted 12 interviews with leading figures in the field. In more detail, we were able to meet 1 CEO and co-founder of a North American software provider, 1 head of digital innovation at a European software provider, 1 chartered insurance practitioner, 2 insurance business researchers, 1 academic scholar, 1 co-CEO of a world-leading private insurance company, 1 actuary of the same world-leading private insurance company, 2 heads of digital innovation from European public sickness funds,Footnote ⁴ and 2 heads of digital innovation from European private insurance companies. Only two software providers declined to be interviewed. Examining our transcripts, we judged the material to be sufficiently comprehensive to capture the breadth of perspectives on the phenomenon. Additional interviews would likely have resulted in redundancy rather than novel insights.

We used two different questionnaires, one more exploratory, the other more detailed, which we adapted to the professional profile of the person interviewed (for example, the questions we asked software providers were significantly different from the questions we asked insurance companies). Each questionnaire contained four or five questions, and the interview was scheduled to last 60 minutes. The interviews were recorded with the consent of the interviewees, then transcribed and thoroughly scrutinised by the authors of this article.

We opted for a qualitative methodology without relying on dedicated software. Our analysis followed a thematic and comparative logic, enabling a systematic and rigorous approach to the interpretation of the data. The first stage consisted of in-depth reading of the interviewees’ responses, aimed at familiarising ourselves with the content of the transcripts and identifying recurring themes and concepts. In the second stage, we developed a set of analytical categories (e.g. “lifestyle”, “behavioural change”, “engagement”) that were both grounded in the empirical material and connected to the theoretical framework developed in this study. The third stage involved linking these categories to the identification of recurring thematic patterns (e.g. “engagement” is closely connected to “behavioural change”) and comparing these patterns across the different interviewees, in order to identify convergences and divergences among the perspectives that emerged during the interviews. In the final stage, we synthesised the results of our thematic and comparative exploration, considering not only the internal coherence of the arguments advanced by the interviewees, but also their external coherence with the current state of the market for proactive life and health insurance policies. The project received ethics approval from the Bielefeld University Ethics Committee and the European Research Council.

Our empirical investigation takes the insurers’ standpoint as a reference and is focused on the main players engaged in proactive life and health insurance today: software providers, public sickness funds, and private insurance companies. We did not examine the complementary perspective of policyholders. While such research would be highly informative, it requires a substantially different design, both in scale and in methodological approach. We hope that in the near future our findings can be set alongside those of empirical studies that focus specifically on the standpoint of policyholders.

The structure of proactive life and health insurance policies

Our interviewees provided us with a wealth of information on the complex structure of proactive life and health insurance policies, which, combined with the data available in the literature and public insurance reports and white papers, gave us a comprehensive picture of how healthcare engagement programmes work.Footnote ⁵ In this section, we describe in detail the tools adopted, the data collected and their use, the feedback loop between insurers and policyholders, the various types of incentives, and the regulatory framework.Footnote ⁶

The policyholder is supposed to download an app and, in some cases, pay a monthly fee for it. The app performs several functions: it tracks the policyholders’ biometric values and daily behaviour (e.g. how many steps they walk, how many calories they consume, how many hours they sleep), and records important aspects of their lifestyle (e.g. gym membership, shopping in healthy grocery stores) and the health checks they undergo throughout the year (e.g. dental health check, mammogram, flu vaccine, etc.). For each of these elements, policyholders earn points which are added up to generate a score. The score defines the status of the insured, from four to five depending on the software providers: the higher the score, the healthier the customer’s lifestyle is assumed to be. The app also notifies policyholders as to how many and what rewards they may enjoy based on their actions, their score and status, and sends recommendations to improve their lifestyle (more on this later).

A kind of feedback loop is thus established between policyholders and the insurer: policyholders consensually disclose their data to the insurance company, the insurance company in turn extracts information from these data and feeds it back to the policyholders to prompt them to improve their lifestyle (and possibly extend their life expectancy). The feedback loop, in other words, is used to build “actionable insights” [Tanninen et al. Reference Tanninen and Booth2022b: 195] that are crucial for making proactivity effective.

However, this feedback loop only works if policyholders are engaged. Engagement can be understood in a broad sense or in a narrow sense [Cevolini et al. Reference Cevolini2025]. In the first case, it refers to the policyholders’ willingness to participate in the well-being programme, for example by attending a gym or carrying out regular health checks. In the second case, engagement concerns the regular interaction of the insured with the app. This is referred to as in-app engagement. As we will see below, both of these forms of engagement are difficult to achieve: they require significant resource investment and the application of behavioural science by insurance companies.

Within the insurance industry there is also consensus that rewards, particularly financial incentives, are a “vital mechanism for driving programme participation and ongoing engagement” [Vitality 2022: 10]. Among the rewards, many insurance companies include the offer of a fitness device (e.g. Apple watch, FitBit watch, Garmin watch, etc.) that must be paired with the app. The fitness device is usually offered free of charge as long as the insured meet certain monthly targets (e.g. taking a certain number of steps each day). The pairing of the tracking device with the app is crucial for generating behavioural and biometric data that help insurance companies improve the policyholders’ risk profile.

The structure of incentives is relatively complex. Insurance companies distinguish between enabling rewards (e.g. the fitness device we mentioned, or a substantial discount on gym membership) and contingent rewards (those which depend on the achievement of goals and, consequently, on the policyholders’ score). They also distinguish between short-term and long-term benefits. Policyholders may receive a small reward (e.g. a Caffè Nero coupon by Vitality) if they meet their weekly goals [Hajat et al. Reference Hajat2019]. At the end of the month or the end of the year, they may also receive more substantial financial rewards if they achieve or maintain a high score.Footnote ⁷

In addition to these technical and financial factors, which are shared with some variation by the majority of proactive policies, there is a multiplicity of regulatory and operational issues that make the global landscape highly differentiated. First of all, a distinction must be made between public sickness funds and private insurance companies. The former are embedded in national social security systems and must consequently meet much more stringent regulatory constraints. This translates concretely into the absence of underwriting freedom.Footnote ⁸ Private insurance companies doing business in community rated environmentsFootnote ⁹ face similar regulatory constraints, but in countries such as the UK, where there is no subsidizing mutuality requirement, these companies are allowed to underwrite. In these cases, behavioural data can be used to adjust the insurance rate to the individual risk profile.

It is also important to consider the regulatory framework, which varies significantly regarding the use of personal data. Whereas in European countries, since 2018, the General Data Protection Regulation (GDPR) has become effective and enforceable by law, in non-European countries such as the US the use of personal data is more liberalised, making a huge difference for companies that can access and use these data. Another crucial factor is the cultural attitude of individuals, who may be more or less reluctant to share their personal data, even when doing so could be to their benefit.

A further important distinction, again with regard to data, is that between clinical data and claims data. The availability of clinical data plays a key role in predictive practices, but not all clinical data are immediately and preemptively accessible. Insurance companies usually only access clinical data when the insured files a claim, to ensure that it is not fraudulent, and must confine themselves to drawing retrospective inferences from that data. Despite their high predictive value, clinical data remain difficult to access and do not directly impact proactive programmes.Footnote ¹⁰

Personalisation, personalisation everywhere

What is going on in the insurance industry selling proactive life and health insurance policies? The increasing use of tracking devices and access to behavioural data have sparked enormous enthusiasm in the insurance industry, primarily because they suggest the possibility of using personalised predictions to trigger positive behaviour change and make people healthier. Rather than simply calculating the loss probability of pools based on statistical variables such as gender and age, insurance companies with access to biometric and behavioural data can gain insight into the lifestyle of policyholders to profile them and thus better assess their risk exposure.

One interviewee illustrates this profiling practice by comparing actuaries to someone aiming an arrow at a target. The target has many rings, and when shooting, you can be “on-target” or “off-target”. The largest ring represents mortality, which is easy to hit because everyone dies. Smaller rings represent causes of death (e.g. diabetes, cardio-vascular diseases, stroke, etc.), which are much harder to hit.Footnote ¹¹ In this context, profiling helps actuaries “get the directionality” (Interview B.6, 1 March 2024).

Our research shows that personalisation takes place on multiple levels, significantly complexifying the structure of the interaction between insurer and policyholders. In addition to risk assessment, the proactive prevention programme is also implemented differently for individual policyholders. The persons we interviewed told us that prevention programmes are not merely personalised, but ultra-personalised. Whereas the first era of prevention programmes could be described as “one size fits all” (Interview B.9, 12 April 2024), now insurance companies claim to be really

Based on that, insurance companies can offer ultra-personalised proactive coaching programmes and rewards:

Prevention thus becomes doubly personalised, because the insurance company not only targets the insured who are most at risk but also customizes the recommended health checks according to individual profiles.

Personalised recommendations also concern fitness. A case in point is the number of steps policyholders are supposed to walk every day. In this case, the frequency of interactions (touchpoints) between insurer and insured changes: while health-related recommendations can be sent monthly or even quarterly, those concerning physical activity are sent weekly or even daily. Here, the personalisation of prevention becomes extremely complex.

Even if there is strong evidence that exercise, particularly walking, has a beneficial effect on individual health [WHO 2018; McCrea and Farrell Reference McCrea and Farrell2018], there is no solid evidence on the recommended number of daily steps for each individual [Banach et al. Reference Banach2023]. One of our interviewees referred to the “10,000 steps paradox” (Interview B.4, 7 February 2024). This figure was a marketing stunt launched by a Japanese firm selling a pedometer during the 1964 Tokyo Olympic Games [Lee et al. Reference Lee2019]. It was later adopted by insurance companies as a standard benchmark for many years. Recently, however, scholars have discovered that optimal mortality benefits are achieved with an average of 7,500 to 8,000 daily steps [Vitality and London School of Economics 2024]. Beyond that point, benefits diminish. In the segment of over 65 years old, for example, 6,400 average daily steps suffice to reduce mortality by 63%.

As a consequence, insurance companies are experimenting with forms of “dynamic step goal”, “re-adjusted” based on individual characteristics and individual behaviour (Interview B.6, 1 March 2024). This dynamic implies incremental improvements, leading policyholders to gradually reach 7,000 steps—which is regarded as the optimal number for mortality improvement, provided the behaviour is habitual and consistent. This progression could occur, for example, from 2,000 to 5,000 steps, then from 5,000 to 6,000, and finally to 7,000 steps.

An actuary explains, in this respect, that if a person’s average daily steps count increases by more than 800 steps a day in any week (e.g. if they are currently at 3,000, their target becomes 4,000 each day in the following week every), that is too much of an increase, and the drop-off rate is too high (Interview B.11, 23 April 2024). For this reason, their company decided to keep the increment below 800 steps. There is evidence that a 200-increment works very well, and the company had more people responding (Interview B.11, 23 April 2024).

Insurance companies are therefore testing personalised thresholds. Threshold-values have been investigated in risk sociology [Beck Reference Beck1986; Reference Beck1994] as distinctions (e.g. above/below 7,000 average daily steps) that are used not so much to indicate real-world data, but rather to self-produce information that enables observers to guide their decisions [Luhmann Reference Luhmann1991: 177; Reference Luhmann, Hiller and Krücken1997: 197ff.]. Thresholds inevitably need to be set as contingent decisions, because one can never know with certainty which value really makes the difference (7,000 steps, 6,780 steps or 7,400?). By referring to the threshold-value, however, policyholders may know, for example, that they are walking too little and decide to make an effort to walk more. With the threshold-value, the insured can observe themselves—and the more personalised the value, the more effective the observation is.

The threshold-value is thus a fundamental tool for the self-tracking and self-monitoring practice of policyholders, and serves to produce a “safety equivalent” [Luhmann Reference Luhmann, Hiller and Krücken1997: 209]: even if those who walk a lot are not sure of eradicating all cardio-vascular risks, they nevertheless know that they took care to at least mitigate these risks. The threshold-value, then, has a kind of “self-calming effect” [Luhmann Reference Luhmann, Hiller and Krücken1997: 202].

In this respect, the ultra-personalisation that insurance companies talk about could be described as a form of third-order personalisation: not only is the prediction adjusted to the individual risk profile (1st order), but the types of recommendations (2nd order) and the timing of communications (3rd order) are also tailored to individual policyholders and their personal behaviour.

A further crucial personalisation factor is incentives. Here, too, one can speak of ultra-personalisation because the right to receive a reward depends on achieving a goal, and the goal itself depends on the threshold set for each policyholder. This means that two policyholders starting from different thresholds can receive the same reward if they achieve the goal set by the insurance company. One person we interviewed told us that a well-known Japanese company is experimenting with the segmentation of prevention programmes, distinguishing policyholders into age groups (e.g. under 35, 35-65, over 65). For each of these groups, both goals and rewards are tailor-made (Interview A.1, 16 June 2023). Another interviewee pointed out that the way people react to incentives is not consistent across countries. In some markets, some incentives work; in others they do not. For insurance companies, this means they have to customize the reward system to fit the local environment (Interview B.11, 23 April 2024).

In addition to all these structural levels, the customisation of goals and rewards is also crucial for personalising policyholders’ engagement. Engagement is one of the main challenges in proactive prevention programmes. Many scholars show that the engagement rate is rather low, in any case lower than the level sought by insurance companies [Hajat Reference Hajat2016; Ledger and McCaffrey Reference Ledger and McCaffrey2017; Spender et al. Reference Spender2019; Jeanningros Reference Jeanningros2021]. The challenge for companies therefore lies in “routinizing [policyholders’ engagement] practices” [Jeanningros Reference Jeanningros2021: 59] so that they become a new healthy habit [Spender et al. Reference Spender2019: 21].

An interviewee told us that a crucial advancement of their company was implementing

In particular, the same interviewee explained that

To be more precise, one must distinguish dis-engagement from not-engagement. In the first case, we are talking about people who start to engage but then, for whatever reason, drop out. In the second case, we refer to people who do not even start the programme, even though they pay the monthly fee and the rewards are available to them. According to one interviewee, the latter category (i.e. people who never become engaged in the first place) is even bigger than the former and is

Another interviewee confirmed this challenge by pointing out that

It is difficult, on the other hand, to understand why people do not engage. The assumption is that these people face barriers. For example, for someone who is very overweight, exercising may be difficult and even embarrassing (Interview B.11, 23 April 2024). To solve this problem, companies are experimenting with strategies based on gradually increasing rewarded challenges for not-engaged policyholders. The company sets lower thresholds (e.g. not abruptly from 0 to 10,000 steps, but more softly from 0 to 2,000 steps) and then gradually raises the bar by rewarding policyholders each time they achieve the target.Footnote ¹² As one interviewee explains, the assumption is that it

Communication between insurers and policyholders is also important and is being personalised, including by experimenting with the use of AI for targeted preventative notifications (Interview B.1, 19 June 2023). The head of digital innovation of one insurance company told us that

Another interviewee told us that

The ultra-personalisation insurance companies speak of should be understood as the combination of all these forms of personalisation—personalisation of threshold-values, personalisation of challenges and goals, personalisation of recommendations, personalisation of rewards and financial incentives, personalisation of communication. None of these personalising initiatives is in itself hyper-personal––that comes from the way they are systematically integrated. This also means that the interactions between the insurance company and policyholders, and the app itself which makes this interaction possible, look somewhat different for each policyholder. In the words of an interviewee, “all of that gets personalised over time” (Interview B.11, 23 April 2024). Another interviewee speaks of a “dynamic app”: the latter is different for different policyholders and, for the same policyholder, different every month and every week (Interview B.9, 12 April 2024). Proactive life and health policies are based on this practice of personalising predictions—in the twofold sense that predictions are tailored to individuals and that predictions are used to personalise prevention programmes.

Consequences of proactive insurance for insurance companies

What do our data show about the way insurance companies use this complex technical-actuarial personalisation of health policies? What consequences do proactive prevention programmes have on insurance companies and their business models?

Like all technical innovations that have a broad social impact, digital technologies employed for insurance purposes have elicited both enthusiasm and scepticism. In the specific case of self-tracking technologies embedded within life and health insurance policies, many sociologists have argued that their use is driven more by a form of brand promotionFootnote ¹³ than by a genuine commitment to helping policyholders improve their lifestyles (to live longer and healthier lives, as the insurance rhetoric claims). Since the insurance industry is profit-driven, and profit depends on maintaining a sufficiently large pool of policyholders to offset losses (bad cases) with gains (good cases), it is understandable that insurance companies are interested in promoting their brand. We, on the other hand, are interested in understanding why, despite all these socio-technical challenges, insurance companies invest substantial capital in digital technologies. What kind of rationale underlies these investments (commonly referred to as InsurTech)? How might this rationale affect the social role of insurance as a fundamental institution of mutuality in modern society?

Digital technologies are often regarded as a tool that could potentially reverse information asymmetry, the situation where one party in an insurance transaction has more or better information than the other. Typically, this occurs when the insured have more knowledge about their own risk profiles than the insurer [Rothschild and Stiglitz Reference Rothschild and Stiglitz1976]. Techniques for tracking individual behaviour could lead to a situation where the insurer knows more about the insured than the insured know about themselves [Siegelman Reference Siegelman2014; Lasry Reference Lasry2015; Interview B.5, 1 March 2024].

Our findings, however, show that this description overlooks the complexity of the structure of proactive policies, where insurers return to their policyholders the information extracted from their data. Rather than a flip of information asymmetry, what emerges is a feedback loop that circulates information in an unprecedented manner, significantly transforming the relationship underlying the insurance agreement. The issue of engagement, for example, acquires unprecedented relevance, as we have seen above.

Another crucial issue is adverse selection, which occurs when individuals who are at higher risk of making claims are more likely to purchase insurance [Rothschild and Stiglitz Reference Rothschild and Stiglitz1976; Albrecher Reference Albrecher2016]. The lack of information of insurance companies on prospective policyholders leads them to offer coverage also to individuals who are more at risk than actuarially estimated [Baker Reference Baker, Ericson and Doyle2003]. As a result, the company’s loss ratio worsens and policyholders in the same pool have to pay higher premiums.Footnote ¹⁴ Companies try to mitigate adverse selection. In the case of life and health insurance, this means attracting the healthiest individuals. What impact do proactive programmes have on adverse selection? Do behavioural policies mainly attract low-risk individuals, or do they actually help policyholders lower their risk? In other words: is it the healthiest who participate in the programme, or do individuals become healthier by participating in the programme?

It is not easy to provide a definitive answer to these questions. Many interviewees confirm that proactive policies have a positive self-selection effect (Interview B.2, 21 September 2023; Interview B.8, 27 March 2024; Interview B.9, 12 April 2024). One interviewee even speaks of “double [positive] self-selection”: proactive policies are more attractive to young people, and also more appealing to those who take care of their health and are already inclined towards a healthy lifestyle (Interview A.1, 16 June 2023). The result could be merely a cherry-picking effect [Cather Reference Cather2018; Hovard Reference Hovard2023], and one of our interviewees asks whether insurance companies really want healthier people or rather healthier portfolios: is it philanthropic interest that leads them to offer products with rewards or simply a marketing strategy to attract healthy people (Interview B.5, 1 March 2024)?

Insurance companies themselves are interested in understanding whether their proactivity programmes succeed in incentivising behavioural change. An interviewee reports that there is statistical evidence that

Is this true? What evidence do we have to support this? Here we must distinguish between countries such as the UK that allow underwriting [Jeanningros and McFall Reference Jeanningros and McFall2020], and community rated environments, where everyone is entitled to life and health insurance at the same price regardless of their individual risk profile. In the first case, the mechanism of risk pooling and spreading follows the principle of risk transfer: the premium paid by policyholders is proportional to the risk they transfer to the insurance company. Under these conditions, where good cases offset bad cases, the removal of the bad cases would only benefit the good ones, which have little need for insurance protection and can in fact buy it cheaply. The bad cases would be left to their own means. However, that would lead to a considerable shrinking of the insurance business as insurance profits are generated by the sound actuarial management of risks not from their exclusion.

In the second case of community rated environments, bad cases and good cases have equal access to insurance protection. Proactivity programmes based on tracking devices, then, do not have a “depooling effect” [Baker Reference Baker, Ericson and Doyle2003: 263], even though they may be more attractive to young sportspeople in good health.Footnote ¹⁵ Companies rather observe a kind of reverse selection. Since there is more room for improvement in negative cases, proactivity aims not to remove but to improve the offset mechanism. A healthier portfolio leads to more competitive rates; more competitive rates attract more customers, including less healthy ones, who can receive the coverage they need at a lower price while also participating in a programme designed to improve their lifestyle. Under these conditions, healthier people and healthier portfolios are not mutually exclusive goals. If anything, they are mutually reinforcing.

Our research also reveals that the innovative aspects of the logic of proactivity appear to encourage the voicing of an unexpected moral impetus by insurance company representatives. Several interviewees claimed that the goal of insurance was no longer only to select people in good health; it was also to motivate people to take care of their own health, and reward them if they did so. As an actuary told us,

Another interviewee expressed the same concept in similar terms:

The companies we interviewed were also aware of a further ethical issue related to the assessment of risks arising from the behaviour of policyholders. Although risks often depend on individual behaviour, behaviour does not always depend on the individual. Extensive empirical research shows that, underlying unhealthy lifestyles, there are socio-economic and cultural determinants that cannot be directly blamed on individuals. Low income and a low level of education can be just as fatal risk factors for mortality as smoking and hypertension [Stringhini et al. Reference Stringhini2017]. Brute luck can thus deprive individuals of their ability to make reasonable decisions and pursue a healthy lifestyle [Steinberg Reference Steinberg2022].

This is why some insurance companies offer so-called enabling rewards. If an individual has a low income, they propose a substantial discount for gym membership, so as to encourage the person to exercise. If an individual cannot afford an Apple watch, these insurance companies offer it for free as long as that person meets the goals. If an individual cannot spend a lot on healthy food, which is notoriously expensive, the insurance company offers discounts for shopping in partner stores. In this perspective, the use of behavioural data might counter the argument that “having a lower income is likely to result in lacking enough resources to invest in equipment, gyms, or healthier foods” [Steinberg Reference Steinberg2022: 675].

Social consequences of proactive insurance

What are the consequences of proactive practices for the function of insurance within society as a whole? The insurance industry, it has been observed, has always had a “voracious appetite for data” [Swedloff Reference Swedloff2014: 341]. This is the “raw material” [Picard Reference Picard, Corlosquet-Habart and Janssen2018: 89] from which insurance firms extract information used in assessing risk and providing collective protection against future contingencies. How has that been impacted by the spread of tracking devices and the use of sophisticated digital data processing techniques?

The widespread opinion in life and health insurance is that the processing of data obtained through wearables is not a substitute for the statistical calculation of risk underlying actuarial procedures. Rather, the former can supplement the latter and provide insights for enhanced risk segmentation [Quah Reference Quah2018; Corlosquet-Habart and Janssen 2018]. By improving their risk assessment techniques, insurance companies can also improve their loss ratios. But a virtually perfect prediction would undermine the specific form of solidarity that underlies the institution of insurance [Cevolini and Esposito Reference Cevolini and Esposito2020]. Its mechanism, in fact, is based on the fact that it is not possible to know which individuals in a pool will suffer which injury and when. That is why it makes sense for everyone to pay an insurance premium, so that everybody is financially covered when adverse events occur [Wilkie Reference Wilkie1997: 1039]. Complete information, by contrast, would identify exactly which individuals in a pool will suffer which injury and when. But, then, insurance companies would charge a premium tailored to the individual’s predicted lifetime, and insurance would turn into a saving plan for those who can afford it. All others would be left without any financial support.

If digital data management techniques were to remove the opacity surrounding individual fates, the mutuality mechanism would collapse [Ewald Reference Ewald2012: 8], with dysfunctional effects on insurance and society as a whole. Good cases (who knew they are good) would no longer be willing to pay to subsidise bad cases [Picard Reference Picard, Corlosquet-Habart and Janssen2018: 98; Eling and Lehmann Reference Eling and Lehmann2020: 374]. The only form of solidarity still possible would be a paradoxical “personalising solidarity” [McFall Reference McFall2019; Krüger and Ni Bhroin Reference Krüger and Ni Bhroin2020] incompatible with mutuality.

Such extreme scenarios, however, still seem a long way from becoming reality.Footnote ¹⁶ Except in cases where insurance companies are allowed to underwrite, as we have seen, rates cannot be adapted to a policyholder’s risk profile. An interviewee told us that, for reasons of strict regulatory constraints, there was a complete segregation in their company between the office responsible for the actuarial calculation of insurance rates and the office with access to behavioural scores. Employees of the same company did not even have a badge to access the other office (Interview B.2, 21 September 2023).

Another interviewee explained that companies with well-established risk assessment systems are very reluctant to touch their actuarial logic (Interview B.1, 19 June 2023), especially when this has been working well for a long time. For these reasons, the expression pay-as-you-live (PAYL) often used in the insurance literature [Ernst and Young 2015] should only refer to the possible financial incentives that the insured person receives if he or she has a healthy lifestyle, and not to the alleged behavioural tarification that according to some scholars would generate “behavioural tribes” composed of individuals who act the same way and pay a price corresponding to their risk profile [O’Neil Reference O’Neil2016: Ch. 9].

If there is no price reduction, then, why should individuals be willing to provide their data to insurance companies? Why should they give up their privacy, and for what benefits? Companies are well aware of this dilemma. The head of healthcare innovation of a public sickness fund told us that the main challenge was to encourage people to “share their data” (Interview B.7, 15 March 2024). Another person in the same position confirmed that when required to provide more data, people ask:

The insurance industry knows how difficult it is to convince policyholders to share their data. One interviewee told us that people tend to be reluctant because they fear that “the sickness fund might abuse it”, and policyholders believe that “patients should be in control of [their] data” (Interview B.7, 15 March 2024). The resistance encountered by companies is very high and the willingness to adopt wearables for insurance purposes remains low [Wiegard et al. Reference Wiegard2019; Talonen et al. Reference Talonen2021; Neumann et al. Reference Neumann2022].Footnote ¹⁷

Some surveys suggest that policyholders are more willing to give companies access to their behavioural data if these data are used for loss prevention rather than personalised premiums [Avramakis et al. Reference Avramakis2020: 22]. In this case, companies focus on another aspect of data sharing in order to motivate policyholders to give up data. Without data no accurate prediction is possible, and without prediction no targeted prevention initiatives can be launched. It is only on the basis of data, they argue, that insurance companies can understand what really works for people and “how [the fund] can incentivise people to have a healthier life” (Interview B.7, 15 March 2024). They regard prevention as a shared value that affects all policyholders. Therefore, those who share their data should do so not only for themselves but for the entire community, on the basis of an unprecedented form of responsibility.Footnote ¹⁸ This responsibility, in their view, would aim at a societal benefit and not simply at a corporate profit,Footnote ¹⁹ and would imply a new form of solidarity—data solidarity [Prainsack et al. Reference Prainsack2022; Braun and Hummel Reference Braun and Hummel2022; El-Sayed et al. Reference El-Sayed2025].

This argument goes back to a long-standing debate surrounding the mutualistic mechanism underlying insurance, that has often been described as a specific form of social solidarity: damages are suffered by individuals, but the compensation mechanism is collective [Ewald Reference Ewald2020]. This traditional form of solidarity, however, is only financial: expected losses are shared by policyholders belonging to the same pool and paying a premium in advance. Today, by contrast, the claim of insurance companies is that the use of digital technologies and access to behavioural data are introducing another form of data-based insurance solidarity, which is indispensable for the effective implementation of personalised proactive programmes. Companies, which have an interest in mitigating the costs of public healthcare systems,Footnote ²⁰ argue that sharing individual data is also expected to contribute to improving the health of all customers.

The people we interviewed are also aware that policyholders’ willingness to share their data could increase if insurance companies and sickness funds were able to better explain the purpose for which they collect data and the use they intend to make of them. Insurance organisations should clarify

To this purpose, they argue that one of the pillars of data solidarity is that a portion of profits resulting from data usage is transferred back to those individuals who shared their data [Prainsack and El-Sayed Reference Prainsack and El-Sayed2023: 36; Prainsack and Kickbusch Reference Prainsack and Kickbusch2024: 1]. In proactive insurance, as we have seen, this happens primarily through financial incentives that reward those who achieve the set targets. More generally, they claim, proactive prevention benefits all policyholders in the pool, because the improvement of the loss ratio implies a decrease in premium rates. There would be a circular relationship between data solidarity and financial solidarity that could, in principle, lead to reciprocal benefit.

Finally, insurance industry professionals are aware of the importance of reassuring people that data are never associated with the individuals who produced and shared them. One interviewee explained that, in order to produce the ultra-personalised proactive coaching programmes we described above, the personal data collected are first depersonalised, then processed, and finally re-individualised to address the single policyholder. More specifically, the interviewee told us that in recent years their company has made large investments in

Another interviewee confirmed that

In proactive insurance, as in other fields that rely on predictive algorithms (e.g. precision medicine)Footnote ²¹ hyper-personalised actionable insights depend paradoxically on the collection of a huge mass of community data. These data are first depersonalised, then processed, to ultimately draw inferences that no longer concern an average person but specific individuals with their own peculiarities and idiosyncrasies. Personal data are the prerequisite for depersonalised individualization.