"I Won the Election!": An Empirical Analysis of Soft Moderation Interventions on Twitter

Abstract: Over the past few years, there is a heated debate and serious public concerns regarding online content moderation, censorship, and the principle of free speech on the Web. To ease these concerns, social media platforms like Twitter and Facebook refined their content moderation systems to support soft moderation interventions. Soft moderation interventions refer to warning labels attached to potentially questionable or harmful content to inform other users about the content and its nature while the content remains accessible, hence alleviating concerns related to censorship and free speech. In this work, we perform one of the first empirical studies on soft moderation interventions on Twitter. Using a mixed-methods approach, we study the users who share tweets with warning labels on Twitter and their political leaning, the engagement that these tweets receive, and how users interact with tweets that have warning labels. Among other things, we find that 72% of the tweets with warning labels are shared by Republicans, while only 11% are shared by Democrats. By analyzing content engagement, we find that tweets with warning labels had more engagement compared to tweets without warning labels. Also, we qualitatively analyze how users interact with content that has warning labels finding that the most popular interactions are related to further debunking false claims, mocking the author or content of the disputed tweet, and further reinforcing or resharing false claims. Finally, we describe concrete examples of inconsistencies, such as warning labels that are incorrectly added or warning labels that are not added on tweets despite sharing questionable and potentially harmful information.

• The paper empirically analyzes soft moderation on Twitter, focusing on warning labels used during the politically charged 2020 US elections.
• It reveals that 72.8% of flagged tweets originate from Republican accounts, highlighting a notable political bias in content moderation.
• The findings indicate that warning labels can boost engagement, as labeled tweets receive higher likes, retweets, and comments.

Empirical Analysis of Soft Moderation Interventions on Twitter

The empirical study conducted by Zannettou provides a nuanced examination of soft moderation interventions on Twitter, particularly during the politically charged period surrounding the 2020 US elections. In this paper, the author scrutinizes warning labels—a form of soft moderation intended to curb misinformation and questionable content without infringing on free speech. The empirical analysis relies on a dataset comprised of tweets from verified users between March and December 2020, focusing especially on the effective yet controversial methods of moderating content without resorting to censorship.

Key Findings

1. Warning Labels as a Moderation Tool: The study identifies 13 distinct warning labels, predominantly linked to claims concerning election fraud during the 2020 US elections. The temporal dynamics reveal that some warning labels are applied persistently across months, while others are rapidly deployed and withdrawn as contextual needs change.
2. Political Distribution of Warning-Labelled Content: A striking observation is the political skew in content subject to soft moderation; 72.8% of such tweets are attributed to Republican accounts compared to just 11.6% from Democrat accounts. This suggests a potential political bias either in the distribution of misinformation or in the moderation process itself.
3. Impact on Engagement Metrics: Contrary to expectations that warning labels might reduce engagement, the paper reports that tweets containing such labels actually receive more likes, retweets, and comments. This observation challenges conventional understandings derived from controlled environments like surveys, and suggests that controversial topics might drive engagement despite being flagged.
4. Challenges and Inconsistencies: The qualitative analysis uncovers inconsistencies in the application of warning labels across different formats and languages. Moreover, there are instances where labels seem arbitrarily applied, underscoring the complexities faced in moderating vast platforms like Twitter.

Implications

The research has several implications for both practical application and theoretical development in the field of social media moderation:

• Political Bias and Transparency: The skew in political affiliation of flagged content raises questions about the transparency and impartiality of moderation systems. It necessitates enhanced mechanisms to audit moderation practices for fairness and consistency.
• Effectiveness of Soft Moderation: Despite theoretical claims that such moderation should curb misinformation spread, empirical data suggest otherwise. Users might perceive warning labels as challenges or invitations for debate rather than deterrents, which could imply a need for redesigning such interventions or complementing them with more comprehensive strategies.
• Challenges in Automated Moderation: Misapplications of labels highlight the difficulties in automation, especially in diverse linguistic contexts and varying content formats. This study reinforces the necessity for human oversight and continual refinement of moderation algorithms.

Future Perspectives

Looking ahead, the results of this study open several avenues for research, particularly in the continued evolution of AI-powered moderation systems. Enhancements might include more robust multilingual support and tools that can seamlessly integrate human judgment with algorithmic efficiency to minimize erroneous flagging. Furthermore, understanding user interactions in dynamic political environments can inform better-targeted and adaptive moderation strategies, improving both user experience and content quality.

Overall, Zannettou’s exploration into soft moderation interventions on Twitter underscores the complexity behind simple acts of tagging content and reveals the layered interactions influencing platform dynamics. Future research should build upon these insights to optimize the balance between free expression and misinformation control in digital ecosystems.