 |
Esko Kivisaari is Chairperson of the AAE Artificial Intelligence – Data Science Working Group |
|
|
 |
Leonid Zeldin is an actuarial consultant |
|
Innovations in data science and artificial intelligence offer the possibility of giant steps when it comes to predicting the future. Actuaries do analyses that are understood to be actuarially fair, or to represent a fair value. Within actuaries the price of an insurance policy is considered fair if customers bearing the same risk are charged the same price. Similarly, the cost of an annuity is actuarially fair if it is equal to the present value of the payments the purchaser expects to receive. A fair value is a measurement of an asset or a liability that can be agreed between market participants. The users of information are thought to be capable of thoroughly understanding the full meaning of the analyses.
In our quickly developing world it is high time to ask whether our concept of something being fair is truly fair. The concept of fairness is central to topical talk on UN’s Sustainable Development Goals (SDGs). Fairness in this context has a heavy emphasis on the treatment of individuals. These individuals usually do not have what it takes to completely understand the content. Fairness in the SDG world is also very much connected to the objective of avoiding discrimination, either direct or indirect. Discrimination means unjust or prejudicial treatment of people, especially on the grounds of ethnicity, age, sex or disability.
Traditional actuarial thinking addresses the needs of fully knowledgeable actors. Our analyses are certainly fair in that context. But we need to ask how such fairness is related to the requirements of the SDGs. Do we still think that our actuarial fairness is fair?
Philosophy makes a distinction between facts and values. It can also be expressed as the difference between descriptive and evaluative claims. The idea is that one cannot derive how things should be from how things are. How things are describes an observable fact while how things should be is always a statement based on our values.
Actuaries are valuable in telling how things are. But we can easily fall prey to the fallacy that this tells how things should be. The problem with actuarial fairness is that it connects facts and values. The component ‘actuarial’ refers to facts and the descriptive claims on how things are. ‘Fairness’ creates a claim of how things should be, i.e., it is a value proposition. We can easily think that what is actuarially fair is the only possible future.
The vision of the Actuarial Association of Europe says that the AAE ‘is for actuaries throughout Europe to be recognised as the leading quantitative professional advisers in financial services, risk management and social protection, contributing to the well‐being of society’. The AAE apparently does this in an actuarially fair manner, based on its values: concern for the public interest, integrity, independence, collaboration and respect, and transparency and accountability. This is all well, but maybe we should think of adding to this what is good, loving, beautiful, important, and indeed, fair.
Actuarial fairness can be thought of looking at the society in the form of how it is today, and projecting this to the future. It is not totally compatible with the idea of contributing to the well-being of society. It rather continues and sometimes reinforces the injustices experienced by the citizens of our societies.
One concrete example of this is how in insurance pricing models are working. ZIP code is often used as a central factor in pricing traffic insurance. Considering how cities are built this can result into models that penalise the insured based on ethnic or socioeconomic factors. ZIP can here represent a proxy for a prohibited attribute, leading indirectly to discrimination. Proxies can exist in many different situations. It is probable that with more and more complicated models often utilising AI there is a growing danger of new proxies emerging.
Innovations in data science and artificial intelligence create novel ways of doing predictions. Actuaries will hopefully find ways of using these innovations in such a manner that they can rather help marginalised and vulnerable populations to a better life instead of just reinforcing past injustices. It is time for actuaries to move from actuarial fairness to true fairness.
20 November 2024
This blog is written on a personal title.
Read all the AAE Blogs here!
