"We find clear evidence that the single accuracy message made users more discerning in their subsequent sharing decisions." write the authors of the study. "Relative to baseline, the accuracy message increased the average quality of the news sources shared, t(5378)=2.90, p=.004, and the total quality of shared sources summed over all posts, t(5378)=3.12, p=.002."
"This translates into increases of 4.8% and 9.0% respectively when estimating the treatment effect for user-days on which tweets would occur in treatment (that is, excluding user-days in the “never-taker” principal stratum[...] because the treatment cannot have an effect when no tweets would occur in either treatment or control); including user-days with no tweets yields an increase of 2.1% and 4.0%in average and total quality, respectively.
"Furthermore, the treatment more than tripled the level of sharing discernment (i.e., difference in number of mainstream versus fake/hyper-partisan links shared per user-day; interaction between post-treatment dummy and link type, t(5378)=3.27, p=.001)."