1Cademy - Perverse Downstream Consequences of Debunking

Concept

Perverse Downstream Consequences of Debunking - Analysis

Ran linear regressions with Huber-White heteroskedasticity and cluster-robust standard errors on users for the user-day level analyses
Included a “post-treatment” dummy variable to account for stepped-wedge design
Computed average relative scores all the user-days = individual scores – average pretreatment score
User-days without any data were given a 0 as a conservative estimate
Overall treatment effect determined by adding the estimates for different days together
Model specifications: number of mentions, political ideology, number of accounts following, whether political ideology of users corresponded with corrector, whether corrector and user had any interaction before correction message was sent

0

1

Updated 2021-08-07

Contributors are:

Who are from:

References