- Dependent variable of user engagement with corrections 
- Used linear regression Fisherian randomization for predicting user engagement 
- User engagement: no engagement (0) or replies/retweets/likes (1)
- User political view: Democrat (-0.5) or Republican (0.5)
- Correction political view: Democrat (-0.5) or Republican (0.5)
- Social interaction before correction sent: no interaction (-0.5), bot followed user and liked 3 posts under the social condition (0.5)

University of Michigan - Ann Arbor

- Procedure
- Experimental Design
- Analysis

Social correction across party lines in a Twitter field experiment - Methodology

Social correction across party lines in a Twitter field experiment. https://www.cmu.edu/ideas-social-cybersecurity/events/conference-papers-2021/conference-2021-paper-4-socialcorrection.pdf 

Social correction across party lines in a Twitter field experiment

- Obtained 10 articles that were deemed false by Snopes.com, 5 pro-Democrat and the other 5 pro-Republican 
- Searched for users who shared these 10 articles on Twitter 
- Created 12 identical bots (6 Democrats and 6 Republicans) to send correction messages 
- Sent 1,585 correction messages but only 1,454 were actually delivered 
- Corrections were direct replies to the original tweet with misinformation and included a short explanation and link to the fact-checking article 

Social correction across party lines in a Twitter field experiment - Procedure

- Blocked users based on follower count (status), tweets in the last two weeks (activity), mentions in the last two weeks, replies in the last two weeks, and political stance 
- Users in each block randomly assigned to one of four conditions (social, asocial, partisan agreement, partisan discordant)
- Used a 2x2 between-subject design 
- Social condition: bots followed users and liked their three most recent tweets one day before corrections were sent 
- Partisan agreement condition: bots with the same political view sent correction message to users
- Partisan discordant condition: bots with opposite political view sent correction message to users 

Learn Before

Related