Ethics in Big Data : designing recommendation algorithms avoiding 'filter bubbles'

Details

Supervisors

Jungers, Raphaël M.

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en informatique

Abstract

Recommendation algorithms play a major role in our modern society. They are key actors for companies like Amazon, Facebook or Google. For Eli Pariser, filtering and recommendation algorithms induce a filter bubble which is the unique universe of information you are exposed to. He argues that this filter bubble has inter-related effects which result in an intellectual isolation. However, he gives no real proof, no mathematical model of it and even less a way to measure it. The main goal of our work is to bring some mathematical rigor to this problem, and to propose as well as elaborate a mathematical model of the concept of filter bubble. This master thesis benefits of a collaboration with the RTBF which provided us a share of their data. This allows us to combine theory with practice on real life data.