Unravelling Linguistic Politicization on Social Media: A Comparative Process Tracing

• In Event: Communication, Information Technologies, and Media Sociology Section Roundtables
  In Refereed Roundtable (60 minute): 6. Presentation, propaganda, & politicization

Abstract

In a polarized era, even seemingly everyday language has been increasingly politicized in media and public discourse. While prior research has offered rich accounts of how political elites, radical actors, and news media draw specific terms into ideological conflict, existing theories remain difficult to systematically test due to the predominance of single-case studies, and the strong focus on competitive and polarized political systems. This study examines linguistic politicization in online public discourse in China using a multi-case, mixed-methods design. Drawing on over 183 million Weibo posts from 2012 to 2022, I employ word embeddings to trace quarterly semantic changes for more than 50,000 Chinese words. Politicization is operationalized as a dynamic reduction in semantic distance between a given term and political axes associated with statism or nationalism. Based on these trajectories, I classify words into abrupt, gradual, and stable or reverting politicization patterns, enabling systematic selection of a few positive and negative cases. I then conduct multi-case process tracing using LLM-assisted computational analysis combined with qualitative analysis to reconstruct how different actors (official media, opinion leaders, radical users, and ordinary users) shape linguistic meaning over time. The study advances media sociology literature by offering a dynamic, comparative account of linguistic politicization and introducing an embedding-based framework for studying language politics at scale.

Author

• Yufan Guo, The Chinese University of Hong Kong