The Polarized Narrative Machine: A Comparative Validation of LLMs and Linguistic Features for Quantifying Partisan Narrative Strategies and Predicting Virality in U.S. Political Tweets

• In Event: Lightning Talk: Information, Media, and Online Political Behavior

Brief Overview

We quantify narrative intensity via a supervised SBERT/LLM model and confirm polarization's linguistic mechanism. Fixed effects and moderation models show that partisan identity drives high-intensity narratives, asymmetrically amplifying virality.

Author

• Sylvia Dou, University of Chicago