Algorithmic Fairness and Democracy

Abstract

It is becoming common, in many spheres of our everyday life, to bump into algorithms that are used to take decisions in our place or to help us take decisions. To prevent these technologies from generating injustices, a new branch of computer science is taking place, called algorithmic fairness. The idea is that a certain vision of justice must be formalized into some statistical criteria, that are put at the core of the computer tools we recur to. Doing so, we can be sure that those instruments will automatically be fair, every time we use them. However, if we study the most common algorithmic fairness criteria, we find out that they are generally based on a very specific vision of justice, which is liberal, in a sense that will be described. Only a few of them derive from another concept of justice. This may have some effects on the quality of our democratic life, as the algorithms that are projected coherently with those principles are used to decide whether to free people from jail or not, to give them a loan, a job, the possibility to study in a university or in another, etc. This article makes an overview of the field of algorithmic fairness and raises some philosophical problems about the kind of democracies we are building, by recurring to the computer tools that are more and more diffused in our societies.