Misinformation and Polarization around COVID-19 vaccines in France, Germany, and Italy

TL;DR

This study analyzes misinformation and polarization around COVID-19 vaccines on Twitter across French, German, and Italian communities, using a large-scale, time-evolving network approach. By constructing country-specific retweet networks at three key temporal snapshots and labeling news sources with NewsGuard scores, the authors quantify echo-chamber formation and shifts in information credibility. They find language-specific polarization trajectories, with German data showing pronounced late-stage echo chambers, while French and Italian discussions exhibit persistent anti-/pro-vaxxer clusters; cross-media vectors like YouTube and Telegram accompany the spread of low-credibility content. The work highlights how polarization interacts with media reliability in a multilingual European context and underscores the importance of cross-platform monitoring for misinformation diffusion and public health communication strategies.

Abstract

The kick-off of vaccination campaigns in Europe, starting in late December 2020, has been followed by the online spread of controversies and conspiracies surrounding vaccine validity and efficacy. We study Twitter discussions in three major European languages (Italian, German, and French) during the vaccination campaign. Moving beyond content analysis to explore the structural aspects of online discussions, our investigation includes an analysis of polarization and the potential formation of echo chambers, revealing nuanced behavioral and topical differences in user interactions across the analyzed countries. Notably, we identify strong anti- and pro-vaccine factions exhibiting heterogeneous temporal polarization patterns in different countries. Through a detailed examination of news-sharing sources, we uncover the widespread use of other media platforms like Telegram and YouTube for disseminating low-credibility information, indicating a concerning trend of diminishing news credibility over time. Our findings on Twitter discussions during the COVID-19 vaccination campaign in major European languages expose nuanced behavioral distinctions, revealing the profound impact of polarization and the emergence of distinct anti-vaccine and pro-vaccine advocates over time.

Figures (14)

• Figure 1: Distribution of different kinds of tweets shared for French data (FR), German data (DE), and Italian data (IT).
• Figure 2: Top-10 most frequent hashtags by country.
• Figure 3: Density distribution of reliability ratings of tweets containing a domain classified by NewsGuard, estimated by Kernel Density Estimation (KDE), for each country.
• Figure 4: Top-15 shared domains in each dataset colored according to their NewsGuard rating.
• Figure 5: Moving average of the daily number of tweets shared in each dataset. We highlight periods in which we constructed three different retweeting networks with different colors.