Computational social-science studies have mapped polarized community structure in climate-change debates on Twitter/X and YouTube, and shown that ideologically aligned communities span both platforms: retweets reinforce within-community echo chambers while mentions bridge communities antagonistically, and right-leaning communities share lower-reliability sources. These findings are, however, purely correlational. The open problem is to causally identify the mechanisms that generate and sustain this cross-platform alignment and community isolation — platform recommendation and ranking algorithms, user-side selective exposure and homophily, or coordinated campaigns — and to characterize how the observed structures emerge and evolve over time and across interaction types.

Background

Online climate-change debates take place simultaneously on several platforms, most prominently the microblogging service Twitter (now X) and the video platform YouTube. A central concept for analyzing them is the echo chamber: a self-reinforcing configuration of interactions in which users predominantly exchange information with like-minded peers, so that opposing views rarely reach them. Echo chambers are usually detected from the topology of interaction networks — for example, retweet networks, where a directed link means one user rebroadcasts another's message, are treated as endorsement signals, whereas mention networks (explicitly addressing another user with @) can link users who disagree. Polarization itself is multi-dimensional: structural polarization refers to the segregation of the interaction network into ideologically homogeneous communities, while affective polarization refers to antagonistic feelings toward the opposing group, measurable through negative sentiment in cross-community messages. The political bias and reliability of shared news outlets are commonly quantified with expert labels such as those of MediaBiasFactCheck, allowing communities to be characterized by the average bias and reliability of the sources they share.

A series of results sets the stage. Echo chambers have been measured robustly across several platforms (Cinelli et al. 2021, PNAS), polarization in climate-change discussions on Twitter has grown over time around events such as the COP conferences (COP, the annual UN climate 'Conference of the Parties' summits) and reports of the IPCC (Intergovernmental Panel on Climate Change) (Falkenberg et al. 2022, Nature Climate Change), and experimental work shows that exposure to opposing views can increase rather than decrease polarization (Bail et al. 2018, PNAS). Cross-platform comparisons of interactional, positional, and affective polarization exist (Yarchi, Baden and Kligler-Vilenchik 2021, Political Communication) but remain descriptive. Bassolas, Massachs, Cozzo and Vicens (2025, Royal Society Open Science 12:241974) analyzed four climate debates (COP27, the IPCC Sixth Assessment Report, climate refugees, and the Doñana natural-park dispute) on both Twitter and YouTube, linking Twitter communities to the YouTube communities whose videos they share. They found: (i) retweet networks form echo chambers dominated by ideological communities; (ii) mentions provide substantial inter-group exposure, accompanied by elevated negative sentiment directed across community boundaries, especially from right-leaning communities that share low-reliability sources; and (iii) a topological alignment between the platforms — Twitter communities preferentially reference ideologically matching YouTube communities — indicating that ideological communities span multiple sites.

The authors are explicit that this evidence is correlational: whether the cross-platform alignment is produced or amplified by platform-level mechanisms (notably the recommendation and ranking algorithms that select which content users see), by users' own choices (selective exposure and homophily), or by coordinated campaigns, is not identified by their design. A filter bubble is the hypothesis that such algorithms narrow users' information diet to congenial content; recent experiments have complicated this picture, since studies using counterfactual bots (automated accounts that replay real users' viewing histories to probe what the recommender would otherwise serve) and naturalistic experiments on YouTube's recommender found algorithmic recommendations playing, if anything, a moderating role in partisan consumption (Hosseinmardi et al. 2024, PNAS), and short-term manipulation of YouTube's recommendation feed had limited effects on policy attitudes (Liu et al. 2025, PNAS), while large feed-algorithm experiments on Facebook during the 2020 US election similarly found little effect on attitudes (Guess et al. 2023, Science). These single-platform, short-term, individual-level results sharpen rather than answer the question of why ideologically aligned communities keep appearing across platforms in climate debates.

Problem Statement

The source study established, descriptively, that climate-change discussions on Twitter and YouTube exhibit ideological community structure that aligns across the two platforms, with retweets reinforcing within-community echo chambers, mentions providing antagonistic inter-group exposure, and right-leaning communities sharing lower-reliability sources. Identify, causally, the mechanisms that generate and sustain this cross-platform alignment and community isolation. Specifically, determine which of (i) platform-level algorithmic curation — recommendation and ranking systems that shape what content users are exposed to on each platform; (ii) user-side selective exposure and homophily — users' own choices to follow, share, and comment on ideologically congenial content across sites; and (iii) coordinated activity such as organized campaigns contribute to producing the observed alignment, the echo-chamber structure of retweet networks, and the antagonistic cross-community exposure carried by mentions. The mechanisms need not be mutually exclusive: the observed structure may be produced by any one of them, by several jointly, or by their interaction, and no single dominant driver is presupposed. Causal identification requires evidence that separates algorithmic supply from user demand rather than purely correlational association, and a characterization of how the polarization structure emerges and evolves over time, including the interaction-type dependence of endorsement versus antagonistic bridging.

References

  1. Aleix Bassolas, Joan Massachs, Emanuele Cozzo, Julian Vicens (2025). Multifaceted polarization and information reliability in climate change discussions on social media platforms. Royal Society Open Science 12(11):241974. DOI: 10.1098/rsos.241974. https://doi.org/10.1098/rsos.241974 (arXiv:2410.21187)
  2. Homa Hosseinmardi, Amir Ghasemian, Miguel Rivera-Lanas, Manoel Horta Ribeiro, Robert West, Duncan J. Watts (2024). Causally estimating the effect of YouTube's recommender system using counterfactual bots. Proceedings of the National Academy of Sciences 121(8):e2313377121. DOI: 10.1073/pnas.2313377121. https://doi.org/10.1073/pnas.2313377121
  3. Naijia Liu, Xinlan Emily Hu, Yasemin Savas, Matthew A. Baum, Adam J. Berinsky, Allison J. B. Chaney, Christopher Lucas, Rei Mariman, Justin de Benedictis-Kessner, Andrew M. Guess, Dean Knox, Brandon M. Stewart (2025). Short-term exposure to filter-bubble recommendation systems has limited polarization effects: Naturalistic experiments on YouTube. Proceedings of the National Academy of Sciences 122(8):e2318127122. DOI: 10.1073/pnas.2318127122. https://doi.org/10.1073/pnas.2318127122
  4. Andrew M. Guess, Neil Malhotra, Jennifer Pan, Pablo Barberá, Hunt Allcott, Taylor Brown, Adriana Crespo-Tenorio, Drew Dimmery, Deen Freelon, Matthew Gentzkow, Sandra González-Bailón, Edward Kennedy, Young Mie Kim, David Lazer, Devra Moehler, Brendan Nyhan, Carlos Velasco Rivera, Jaime Settle, Daniel Robert Thomas, Emily Thorson, Rebekah Tromble, Arjun Wilkins, Magdalena Wojcieszak, Beixian Xiong, Chad Kiewiet de Jonge, Annie Franco, Winter Mason, Natalie Jomini Stroud, Joshua A. Tucker (2023). How do social media feed algorithms affect attitudes and behavior in an election campaign? Science 381(6656):398-404. DOI: 10.1126/science.abp9364. https://doi.org/10.1126/science.abp9364
  5. Matteo Cinelli, Gianmarco De Francisci Morales, Alessandro Galeazzi, Walter Quattrociocchi, Michele Starnini (2021). The echo chamber effect on social media. Proceedings of the National Academy of Sciences 118(9):e2023301118. DOI: 10.1073/pnas.2023301118. https://doi.org/10.1073/pnas.2023301118
  6. Max Falkenberg, Alessandro Galeazzi, Maddalena Torricelli, Niccolò Di Marco, Francesca Larosa, Madalina Sas, Amin Mekacher, Warren Pearce, Fabiana Zollo, Walter Quattrociocchi, Andrea Baronchelli (2022). Growing polarization around climate change on social media. Nature Climate Change 12(12):1114-1121. DOI: 10.1038/s41558-022-01527-x. https://doi.org/10.1038/s41558-022-01527-x
  7. Christopher A. Bail, Lisa P. Argyle, Taylor W. Brown, John P. Bumpus, Haohan Chen, M. B. Fallin Hunzaker, Jaemin Lee, Marcus Mann, Friedolin Merhout, Alexander Volfovsky (2018). Exposure to opposing views on social media can increase political polarization. Proceedings of the National Academy of Sciences 115(37):9216-9221. DOI: 10.1073/pnas.1804840115. https://doi.org/10.1073/pnas.1804840115
  8. Moran Yarchi, Christian Baden, Neta Kligler-Vilenchik (2021). Political polarization on the digital sphere: A cross-platform, over-time analysis of interactional, positional, and affective polarization on social media. Political Communication 38(1-2):98-139. DOI: 10.1080/10584609.2020.1785067. https://doi.org/10.1080/10584609.2020.1785067
  9. Edoardo Loru, Alessandro Galeazzi, Anita Bonetti, Emanuele Sangiorgio, Niccolò Di Marco, Matteo Cinelli, Max Falkenberg, Andrea Baronchelli, Walter Quattrociocchi (2025). Ideology and polarization set the agenda on social media. Scientific Reports 15:35816. DOI: 10.1038/s41598-025-19776-z. https://doi.org/10.1038/s41598-025-19776-z
  10. Vincenzo De Leo, Michelangelo Puliga, Martina Erba, Cesare Scalia, Andrea Filetti, Alessandro Chessa (2026). Identifying energy communities of practice on Twitter: A multiplex network analysis using graph traversal techniques. Complexities 2(2):15. DOI: 10.3390/complexities2020015. https://doi.org/10.3390/complexities2020015
  11. Aleix Bassolas, Piero Birello, Julian Vicens (2026). An analytical and experimental study of the energy transition discourse on YouTube. PLOS ONE 21(7):e0352691. DOI: 10.1371/journal.pone.0352691. https://doi.org/10.1371/journal.pone.0352691