Navigating the Disagreement Space: A Case Study on Persistent YouTube Users’ Interactions in Immigration-Related Discussions

EasyChair Preprint 16017

Abstract

YouTube comment sections constitute volatile, deindividualized arenas, where low accountability and antagonistic dynamics hinder constructive political discussion. This paper investigate how persistent users, those who repeatedly return to these contentious spaces, navigate immigration-related debates and sustain their stance positioning over time. Drawing on a longitudinal corpus of U.S. immigration-related YouTube videos comments (2020–2024), we introduce a Natural Language Inference (NLI) based stance detection pipeline that scales to YouTube’s conversational structure, enabling fine-grained classification of user issue positioning. Across user clusters, we identify diverse interactive strategies: some reinforce in-group alignment, others deliberately seek cross-cutting arenas, and still others combine antagonism with curiosity in ways that challenge conventional accounts of echo chambers. Our results highlight that, among YouTube persistent users, polarization is neither uniform nor inevitable: while some groups radicalize, others sustain margins of openness through cross-cutting interaction with diverse users. The findings illustrate how issue positioning, communicative style, and strategic engagement choices co-evolve in deindividualized online settings. Together, this paper advances theoretical understanding of online polarization and contributes new resources for computational analysis of controversial debates. Complete results, datasets and scripts available at https://anonymous.4open.science/r/Navigating_Disagreement-E94B

Keyphrases: Controversy Analysis, Social Identity Theory, YouTube, polarization, social network analysis, stance detection

@booklet{EasyChair:16017,
  author    = {Davide Bassi and Martín Pereira-Fariña and Renata Vieira},
  title     = {Navigating the Disagreement Space: A Case Study on Persistent YouTube Users’ Interactions in Immigration-Related Discussions},
  howpublished = {EasyChair Preprint 16017},
  year      = {EasyChair, 2026}}