Abstract
Many social and global challenges in the 21st century, such as climate change, require responses that are rapid and scalable. When centralized decision-making proves insufficient, bottom-up solutions become essential for effective mitigation. The Social Web offers a unique environment for such collec-tive action, as it overcomes geographical boundaries and enables rapid, decentralized, and large-scale
interactions. In this context, communication plays a central role in shaping problem perception, fram-ing issues, and coordinating collective behaviors. Although online platforms make large-scale coor-dination possible, understanding how communication dynamics translate into mobilization remains a challenge.
Existing research on online communication and collective action faces two obstacles: scope and measurement. Much of the existing work is limited in scale, focuses on specific online platforms or collective action issues, or examines only certain stages of mobilization. Furthermore, language use, a core component of online communication, and behaviors associated with collective action are often latent or poorly defined, and the vast amount of textual data produced online complicates their mea-surement at scale. Advances in Natural Language Processing, particularly Large Language Models,
offer new tools for large-scale analysis, but their application to complex and nuanced social concepts remains underexplored.
We address these gaps by examining the relationship between language use and the emergence of collective action behaviors on multiple online platforms (Reddit, TikTok, YouTube, X, and Facebook) and in the context of different collective action problems (climate change, electoral engagement, and labor representation). We tackle the challenge of measuring latent components of this relationship, namely language use and behaviors, by providing both methodological and empirical contributions.
Methodologically, we develop and openly release scalable, theory-driven models to quantify the use of language and behavioral signals. Empirically, we apply these models across platforms and domains to explore how online communicative practices correlate with the emergence of collective action. Our findings advance Computational Social Science and the study of online activism by showing that specific patterns of language use are correlated with collective action behaviors across multiple online domains and issues, addressing gaps in the scope of previous studies. In addition, the opera-tionalization and measurement tools developed in this work can be integrated into analytical pipelines to generate new insights into complex social phenomena at scale. By providing these methodolog-ical and empirical resources, this thesis contributes to building a more systematic understanding of collective action as it unfolds online, supporting future research on social mobilization.
interactions. In this context, communication plays a central role in shaping problem perception, fram-ing issues, and coordinating collective behaviors. Although online platforms make large-scale coor-dination possible, understanding how communication dynamics translate into mobilization remains a challenge.
Existing research on online communication and collective action faces two obstacles: scope and measurement. Much of the existing work is limited in scale, focuses on specific online platforms or collective action issues, or examines only certain stages of mobilization. Furthermore, language use, a core component of online communication, and behaviors associated with collective action are often latent or poorly defined, and the vast amount of textual data produced online complicates their mea-surement at scale. Advances in Natural Language Processing, particularly Large Language Models,
offer new tools for large-scale analysis, but their application to complex and nuanced social concepts remains underexplored.
We address these gaps by examining the relationship between language use and the emergence of collective action behaviors on multiple online platforms (Reddit, TikTok, YouTube, X, and Facebook) and in the context of different collective action problems (climate change, electoral engagement, and labor representation). We tackle the challenge of measuring latent components of this relationship, namely language use and behaviors, by providing both methodological and empirical contributions.
Methodologically, we develop and openly release scalable, theory-driven models to quantify the use of language and behavioral signals. Empirically, we apply these models across platforms and domains to explore how online communicative practices correlate with the emergence of collective action. Our findings advance Computational Social Science and the study of online activism by showing that specific patterns of language use are correlated with collective action behaviors across multiple online domains and issues, addressing gaps in the scope of previous studies. In addition, the opera-tionalization and measurement tools developed in this work can be integrated into analytical pipelines to generate new insights into complex social phenomena at scale. By providing these methodolog-ical and empirical resources, this thesis contributes to building a more systematic understanding of collective action as it unfolds online, supporting future research on social mobilization.
| Originalsprog | Engelsk |
|---|---|
| Vejleder(e) |
|
| Status | Udgivet - 2026 |