1. médialab Sciences Po
  2. Activités
  3. European Polarisation Observatory (EPO)

European Polarisation Observatory (EPO)

Comment mesurer la polarisation dans des systèmes socio-politiques européens multipolaires et multi-thèmatiques ?

Traditionnellement, l'opinion des gens sur les questions de débat public est étudiée par le biais de sondages et d'enquêtes. De récentes avancées dans les méthodes de mise à l'échelle des réseaux ont montré que les traces comportementales numériques (typiquement les réseaux de suiveurs/suiveurs) dans les plateformes de médias sociaux peuvent être utilisées pour extraire les opinions à une échelle massive. Ce projet vise à développer une preuve de concept pour un Observatoire européen de la polarisation des idéologies et des attitudes envers les questions du débat public.

Recherche

Objectives

This project seeks to develop a joint workgroup or unit within CIVICA, dedicated to producing a proof of concept for a European Polarisation Observatory of ideologies and attitudes towards issues of the public debate (e.g., taxation, immigration, European integration, perception of elites). Such a unit would study the ideological dynamics of users and media outlets to measure polarization on various issues, and investigate the role of algorithmic recommendations on user access to ideologically diverse news.

The inference of attitudes of users for a multitude of issues will allow for the ideological positioning and tracking of various groups: social movements, demographic groups, political and partisan groups, and even entities such as media outlets and YouTube channels. These results will be implemented in research cutting across several disciplines of interest for CIVICA: media studies, online social movements, and the analysis of the structure of party systems in European countries, among others. In this project we will specifically aim at two kinds of applications. First, we will provide systematic measurements across Europe of the intensity of polarization on different issues (e.g., left-right economics, or attitudes towards people and elites) in various online settings. Second, we will seek to tackle the question of the role of AI in shaping European socio-political systems through algorithmic recommendation and its possible biases.

In order to develop a proof of concept, this project includes multiple activities: data collection of social networks across Europe, the application of network scaling procedures, the use of political science datasets for attitude inference in conjunction with scaled data, and research in social psychology and algorithmic recommendation. 

Preliminary results

Many activities already show promising preliminary results. 

Legend : Embedding of users, politicians, and parties in latent ideological spaces computed from the French Twitter network. Ideological spatialization provides new means for measuring multi-polar and multi-issue polarisation.
Legend : Embedding of users, politicians, and parties in latent ideological spaces computed from the French Twitter network. Ideological spatialization provides new means for measuring multi-polar and multi-issue polarisation.

Overall, this project will be instrumental in further developing an interdisciplinary European network of research in the computational social sciences.