Peng-Ting Kuo1, Gaeun Jung1, and Hari Acharya2
1Department of Government, University of Texas at Austin | 2Department of Computer Science, University of Rochester
Abstract
This study examines how identity-based rhetoric employed by fictional characters shapes audience emotional responses in naturalistic online discourse. Using Attack on Titan as a case study, we construct an original dataset combining episode-level character dialogue with Reddit discussion threads and code rhetorical features — including threat framing, boundary marking, and solidarity appeals, using Gemini 2.5-pro. Employing a two-way fixed effects model with character and season fixed effects, we find that identity-based rhetorical intensity amplifies audience sentiment polarization and that solidarity appeals generate positive audience responses only when delivered by protagonist characters, consistent with parasocial identity reinforcement. Character-level fixed characteristics, however, account for the majority of variance in emotional responses, suggesting that narrative positioning conditions audience reception more strongly than episodic rhetorical content. Our findings contribute to the literature on narrative persuasion and Social Identity Theory, with implications for how fictional storytelling shapes audience moral alignment with political violence.
Technical Skills: Python, LLM, Two-Way Fixed Effects, Factor Analysis