Content Moderation As a Political Issue: The Twitter Discourse Around Trump's Ban

Meysam Alizadeh, Fabrizio Gilardi, Emma Hoes, K. Jonathan Klüser, Maël Kubli, Nahema MarchalJournal of Quantitative Description Digital Media 2022 · DOI

TL;DR

This paper analyzes the Twitter discourse surrounding content moderation, particularly after Donald Trump's ban, revealing it as a highly politicized issue. The study finds that while left-leaning users initially drove the conversation, "Section 230" became a central, polarizing topic across the political spectrum, highlighting the deep political implications of online speech regulation.

What Problem Does This Paper Address?

The paper tackles the growing political significance of content moderation, moving beyond its technical aspects to understand its role in democratic accountability and civil liberties. Specifically, it investigates how the public discourse on Twitter evolved around this issue, especially following major events like Donald Trump's ban from social media platforms in January 2021. The authors aim to understand how this issue became politically contested, who participated in the discourse, and what specific aspects of content moderation were most debated.

Methods

The researchers employed a multi-faceted approach:

  • Data Collection: They collected over 9 million tweets and retweets from approximately 3.5 million unique users between January 2020 and April 2021, using Twitter's academic API and the R package RTwitterV2.
  • Topic Modeling: Latent Dirichlet Allocation (LDA) was used to identify key themes and sub-topics within the discourse, resulting in six distinct topics.
  • Stance Classification: A fine-tuned transformer model (XLM-T) was employed to determine user stances (support or disapproval) regarding the repeal or maintenance of Section 230.
  • Network Analysis: A retweet network was constructed to identify user communities and their interactions, using the Louvain community detection algorithm.
  • Bot Detection: The Bot-O-Meter API was used to estimate the proportion of bot accounts in the dataset, ensuring findings were not disproportionately driven by automated accounts.
  • Classification: Supervised classifiers (XGBoost) were used to filter for relevant tweets and classify user profiles.

Datasets

  • Twitter Data: Over 9 million tweets and retweets from approximately 3.5 million unique users, collected between January 2020 and April 2021.
  • News Articles: Data collected from English newspapers via LexisNexis API, published between January 1, 2020, and December 31, 2021.

Key Insights and Ideas

  • Content Moderation as a Political Battleground: The study clearly demonstrates that content moderation is not just a technical or operational issue for platforms but a deeply political one, with significant implications for free speech and democratic discourse.
  • Polarization is Key: The retweet network analysis revealed a strong tendency for users to engage primarily within their ideological bubbles (left-leaning and right-leaning communities), with limited interaction between them. This suggests a polarized discourse where different political groups talk past each other on this issue.
  • Section 230: The Central Nexus: The analysis consistently points to Section 230 of the Communications Decency Act as the most frequently discussed topic, acting as a focal point for debates about platform liability, online speech regulation, and the balance between free expression and content control.
  • Event-Driven Salience: The attention to content moderation spiked around major political events, most notably Donald Trump's ban from social media following the January 6th Capitol attack, underscoring how real-world events shape online discourse.

Key Results

  • Salience Peaks: Content moderation discourse saw significant spikes in attention around key events, including Twitter's fact-checking of Trump's tweets (June 2020), the US election, and most prominently, Trump's ban from social media (January 2021).
  • User-Driven Discourse: Non-elite users generated the majority of the political conversation about content moderation.
  • Community Structure: The retweet network analysis identified distinct left-leaning and right-leaning communities, comprising 44% and 37.5% of users respectively, who largely retweeted within their own groups.
  • Topic Distribution: "Section 230" (27% of tweets) and "Trump Ban" (21%) were the most dominant topics in the discourse.
  • Stance Volatility: Stance towards Section 230 was volatile, particularly among right-leaning users, showing signs of increasing polarization towards the end of the observation period.
  • Bot Activity: Bot detection estimated that approximately 9.9% of users and 12.4% of tweets were from bots, a level deemed not to significantly skew the findings.

Limitations and Future Work

  • Limited Timeframe: The study's findings are based on a specific period (January 2020 - April 2021). The authors suggest that extending the time frame, particularly to examine the polarization trends of Section 230, would be valuable.
  • Conceptualization of Frames: While topic models were helpful, future research could aim for a more precise conceptualization and measurement of the frames used to describe content moderation as a political problem.
  • Broader Arenas: The study focused on Twitter. Future work should incorporate other arenas like traditional media, political statements (e.g., Congressional debates), and tech company communications for a more comprehensive understanding.
  • Theorizing Politicization: The process by which content moderation emerged as a politically salient issue needs more thorough theoretical exploration, including the roles of different actors and interconnections between various communication arenas.

Summary updated

Abstract

Content moderation — the regulation of the material that users create and disseminate online — is an important activity for all social media platforms. While routine, this practice raises significant questions linked to democratic accountability and civil liberties. Following the decision of many platforms to ban Donald J. Trump in the aftermath of the attack on the U.S. Capitol in January 2021, content moderation has increasingly become a politically contested issue. This paper studies that process with a focus on the public discourse on Twitter. The analysis includes over 9 million tweets and retweets posted by over 3 million unique users between January 2020 and April 2021. First, the salience of content moderation was driven by left-leaning users, and "Section 230" was the most important topic across the ideological spectrum. Second, stance towards Section 230 was relatively volatile and increasingly polarized. These findings highlight relevant elements of the ongoing process of political contestation surrounding this issue, and provide a descriptive foundation to understand the politics of content moderation.
BibTeX
@article{alizadeh2022content,
  author = {Meysam Alizadeh and Fabrizio Gilardi and Emma Hoes and K. Jonathan Klüser and Maël Kubli and Nahema Marchal},
  title = {{Content Moderation As a Political Issue: The Twitter Discourse Around Trump's Ban}},
  journal = {Journal of Quantitative Description Digital Media},
  year = {2022},
  doi = {10.51685/jqd.2022.023},
}
Figure