Zannettou, S. (2021). " I Won the Election!": An Empirical Analysis of Soft Moderation Interventions on Twitter. arXiv preprint arXiv:2101.07183. https://arxiv.org/abs/2101.07183

University of Michigan - Ann Arbor

Researchers wanted to study how users interact with soft moderation on Twitter, which they accomplished through a mixed-methods approach and qualitative analysis. They noted that previous research, although they made insightful discoveries such as the implied truth effect, backfire effect, and illusory truth effect, were done in hypothetical settings. Hence in this study, researchers targeted tweets associated with the 2020 presidential election to answer the following research questions: 
- What are the various warning labels used on Twitter during the 2020 US elections and what kind of users have their tweets flagged more frequently? Are there differences across political leanings? 
- Is the engagement of content that includes warning labels significantly different from the content without warning labels?
- How do users on Twitter interact with content that includes warning labels?

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

- Used Twitter API to get tweets and retweets from verified users on Twitter because they have a larger following, as well as information regarding whether a tweet has warning labels
- Between December 7th to 31st, 2020, researchers obtained a total of 79,361,081 tweets from 168,126 verified users
- To obtain tweets with warning labels, tweets that quote other tweets containing warning labels, and retweets of tweets with warning labels, researchers again used Twitter API 
- Final dataset consisted of 18,765 tweets containing warning labels from 8,142 users
- Divided dataset into two groups – tweets with warning labels and tweets quoting other tweets with warning labels

“I Won the Election!”: An Empirical Analysis of Soft Moderation Interventions on Twitter Data Collection

- Republicans were responsible for sharing the majority (72%) of tweets with warning labels on the topic of the 2020 US Presidential Election
- Higher engagement observed for tweets with warning labels, indicating that political bias contributed to people’s perceptions of the warning labels or that warning labels were not added until substantial discussion of the tweet was found 
- Many users take advantage of soft moderation to debunk misinformation and provide resources to accurate and reliable information 

Learn Before

Related