The last few decades have seen a growing interest in the impact of social media. One of the reasons for this is their increasing impact on social and political dynamics, which raises concerns about several hypothetical phenomena : echo chambers, polarization, filter bubbles, extremism, etc. Furthermore, research indicates that the recommendation algorithms might play a role in some of those phenomena.

While past research has focused on the outputs of the recommendation algorithms, our work aims to focus on the algorithm itself and on the process of computation used to create the recommendations. By doing this we integrate our work in the growing literature around algorithmic interpretability and explainability. And we put the basis for further work on algorithmic audit and transparency highly researched by the Europeans public institutions.