Les algorithmes sont-ils racistes ?

Abstract

Following a sociological perspective, this article offers a documented overview of issues of racism and discrimination in the digital context, particularly since the emergence in the 2010s of a series of advanced technologies related to artificial intelligence. The functioning of certain algorithms (those using machine learning) is questioned, and a reflection on the training data, the algorithmic biases and the social choices behind the development of these technical tools is carried out. Particular attention is paid to the content recommendation algorithms of social media platforms, considered as the tools of artificial intelligence posing the most important social challenges today. This work was carried out in light of the productions and definitions established by the sociologies of racism, discrimination and race. We also exploit the technical literature in security and reliability of machine learning.

Keywords

• racism
• big data
• discriminations
• algorithms
• artificial intelligence
• recommendation algorithms