Advocating for the fairness of algorithms.

Summary Algorithms are now part of every individual's daily life: prioritizing the results of a search for information on Google (pagerank), selecting the information that appears on a Facebook wall (Edgerank), optimizing travel, recommending products, detecting pathologies. We are only at the beginning of such a phenomenon, as our lives become data and predictive models are applied to more and more fields. Algorithms raise ethical questions for several reasons. First of all, the very technique of the algorithm can lead to discrimination prohibited by law, due to the reintroduction of profiling, now without the need for prior identification, by simply cross-checking data. The risks of confining Internet users in a personalization of services according to their tastes or in supposed spheres of opinion also pose a problem. This is opposed to the free development of the individual, and participates in a reductive homogenization of information, in opposition to cultural pluralism. The massive use of algorithms also develops the risk of an excessive confidence in the choices recommended by these calculations, based on potentially false or approximate postulates, but likely to orient the choices of individuals without them being really aware of it. Finally, there is the question of the "solutionist" risk of a society that encourages the systematic use of algorithmic solutions that mask the complexity of the socio-economic issues that require other types of intervention. Faced with this situation, it is important to look for ways of thinking about it, such as regulation imposing the loyalty of algorithms, or the implementation of systems that would allow us to verify their operationality, such as rating agencies or a body of experts in algorithms (algorithmists) that can be mobilized at the request of a regulatory authority.