Who Talks to Whom: Quantifying Echo Chamber Effects in Emerging Social Media Platforms

Abstract

This study investigates interaction patterns on three emerging social media platforms in the "post-Twitter era"—BlueSky, Mastodon, and Truth Social—focusing on reply networks as highly interactive, opinion-expressive forms of engagement within a polarization context. We employ a multi-method approach to quantify the echo chamber effect, analyzing (1) the role of influential users in shaping narratives (using PageRank and LexRank), (2) ideological distribution within reply network communities (leveraging the Leiden algorithm for community detection and LLMs to assess ideological leanings), and (3) patterns of connectivity favoring ideologically similar users (via biased random walks). To ensure rigorous cross-platform comparability, we apply strict controls on topic and timeframe. Our findings reveal a relatively weaker echo chamber effect on Mastodon and BlueSky, though the platforms display subtle differences: Mastodon fosters a diverse ideological landscape with frequent cross-ideological exchanges, while discussions on BlueSky are more centered around moderate voices. In contrast, Truth Social exhibits a pronounced echo chamber effect, with the majority of users supporting right-leaning or pro-Trump narratives, in alignment with the platform’s founding intent as a dedicated space for Trump supporters. Despite this homogeneity, we observe occasional instances of critical engagement and fact-checking on Truth Social, suggesting that even within ideologically driven spaces, counter-narratives can emerge, though their influence remains uncertain. These findings underscore the need for further research into the unique dynamics of emerging social media platforms, particularly those shaped by ideological leanings. Understanding the roles echo chambers play in shaping public opinion is essential for gauging their impact on (mis)information diffusion and political polarization in an increasingly fragmented digital landscape.