Can AI Enable More Affordable Personalized Insurance Premiums? Here’s Why It’s a Risky Path

Insurance is based on a principle of solidarity, but that is now being undermined by the algorithms used to build our profiles.

As these algorithms grow more sophisticated and precise, premiums become increasingly personalized. That means “high-risk” profiles may end up excluded from insurance schemes altogether because the costs become prohibitive. Personalizing has some legitimacy, but it must be balanced with fair access to insurance.

But first, it’s important to understand that insurance is built on a fundamental paradox. On one hand, its very principles rest on a collective mechanism in which everyone contributes according to their means and benefits from solidarity in the event of a loss. On the other hand, technological advances, the growing abundance of data and increasingly refined actuarial methods are pushing toward ever more individualized pricing.

Adding to this tension is an increasingly demanding legal framework that prohibits any form of discrimination based on sensitive data, which is sometimes correlated with relevant risk factors.

As a professor of mathematics at the Université du Québec à Montréal (UQAM), I am co-author of the Manuel d’Assurance and author of the recently released book Insurance, Biases, Discrimination and Fairness. I want to examine the challenge of reconciling the solidarity-based pooling, which is the foundation of insurance, with the hyper-segmentated pricing made possible by big data — without excluding or discriminating against policyholders.

Price segmentation

Insurance companies have long used classification as a pillar of their business model: age, gender, occupation, geographical area, claims history, etc.

In 1662, English statistician John Graunt published the Bills of Mortality, the first statistical analysis of London’s death records. In 1693, English astronomer Edmund Halley developed the first mortality table, which made it possible to calculate life expectancy at each age.

This work laid the foundations for differentiated pricing based on age and gender, which for a long time remained the two main criteria for segmentation in life and death insurance.

At the same time, after the Great Fire of London in 1666, the first fire insurance contracts appeared: companies collected data on the nature of building materials and urban density. In the 18th and 19th centuries, rates were segmented according to the proximity of neighbouring buildings and the presence of firefighting services, giving rise to the first “high-risk areas” and “low-risk areas.”

With the rise of the automobile in the 1910s and 1920s, American insurers began systematically recording the number of claims, the age and the gender of drivers. By the 1920s, several “classes” of rates were established: young drivers, female drivers and experienced drivers, which allowed premiums to be set according to specific profiles.

Today, actuaries have access to sophisticated algorithms, machine learning tools and a flood of data: onboard telematics, connected devices, geolocation, driving or lifestyle behaviours, and more. For insurers, refining segmentation allows them to bill each policyholder “at their true level of risk,” reducing the effects of cross-subsidization from good risks to bad risks, while improving overall profitability.

However, overly precise pricing reduces mutualization and can make insurance very expensive or even inaccessible for certain high-risk segments. As a result, actuaries today are seeking a subtle balance between capturing the right information to differentiate profiles and preserving the viability of the insured community.

The illusion of winning personalization

In Europe, the Financial Data Access Framework (FIDA) legislative proposal would give insurers regulated access to individuals’ financial data. Its aim is to refine knowledge of spending and repayment behaviour. In this context, the promise of ultra-personalized pricing raises hopes of lower premiums but also fears of excessive profiling and significant exclusions.

Faced with this new influx of data, many customers see personalization as a win-win approach: if I manage my budget better, I will benefit from a discount; if my savings and repayment habits are deemed virtuous, my health insurance premium will decrease; if my financial profile improves, my home insurance will become cheaper.

This logic of “pay-as-you-live” or “pay-how-you-drive” is appealing: individuals believe they are in control of their insurance costs through their lifestyle choices.

However, several points are worth highlighting.

• The principle of mutualization is not neutralized; those who cannot adopt the most virtuous behaviours remain dependent on the solidarity of others. Even if the most at-risk individuals pay more individually, those who are less at risk continue to bear a share of the costs thanks to the principle of mutualization.

• Information asymmetry is reinforced, as the insurer knows the statistics better than the customer. Personalized offers are often based on correlations, sometimes tenuous, whose significance is unknown to the customer.

• Highly personalized products may force those most at risk to overinsure or, conversely, to forego insurance altogether, thereby undermining mutuality.

That means even when enhanced by access to financial data, “personalization” is not necessarily synonymous with “empowerment” for the consumer.

The legal framework

The development of big data in insurance raises important ethical and legal questions: to what extent can sensitive variables be used to predict risk?

Read more:
IA et crédit, comment éviter que les machines ne contribuent à reproduire nos biais sociaux ?

In France and the European Union, legislation explicitly prohibits discrimination based on protected criteria such as ethnic origin, gender, sexual orientation, disability, religious beliefs, and more. The Solvency II Directive (EU) requires insurers to use risk models that are “transparent” and non-discriminatory.

Unlike the European Union, which prohibits differentiated pricing based on protected criteria (gender, origin, disability), the Quebec model offers a more permissive framework. While the Quebec Charter of Human Rights and Freedoms also prohibits discrimination, it provides exemptions for insurers: they may, when a factor is statistically relevant, base pricing on age, gender, or marital status.

This practice, authorized solely on the basis of correlation, raises questions.

Ethics,social responsibility of insurers

Beyond mere legal compliance, the ethical practices and social responsibility of insurers are increasingly being scrutinized by consumer associations and the media, which report incidents of algorithmic discrimination and exert reputational pressure.

As a result, in recent years insurers have had to ask themselves, collectively, how to guarantee fair access to their products for vulnerable populations without sacrificing the financial viability of their portfolios. To avoid exclusion, some innovative models offer “solidarity” formulas or capped rates.

Insurers are facing ever-increasing transparency requirements. To avoid perceptions of arbitrariness, they must clearly explain their pricing criteria and make their calculation methods accessible. Finally, they must integrate data protection and privacy into the design of their products (“Privacy by Design”) in order to maintain trust.

Insurers that are able to reconcile personalization, fairness and inclusion will become the benchmark for ethically minded customers.

Reconciling solidarity and data

The challenge, as we can see, is considerable.

It requires nothing less than reconciling actuarial precision with the values of redistribution and solidarity that have underpinned the insurance profession.

The future of insurance will be decided only by resolving this tension. There can be neither pure price discrimination nor simple illusory personalization. The insurance industry will, instead, need to balance the two to allow everyone to contribute according to their risk and benefit fairly from the mutualization of life’s uncertainties.