How Effective Are “False” Labels on Facebook Posts? Evidence from India

• In Event: Combating Misinformation and Fake News

Abstract

With more people consuming information online, being exposed to false and low-quality information is inevitable. Fact-checking is one of the methods employed by social media platforms to inform their audiences about the accuracy of the information to which they are exposed. While the volume of misinformation can overwhelm the fact-checkers, the fact-checking labels are intended to reduce the amount misinformation on social media, both the spread of misinformation and belief in it. There has not been much research that examines the effectiveness of these labels using actual misinformation posts on social media, especially political misinformation, and the experimental evidence on the effectiveness is mixed. However, most of these studies are from western contexts, and there is a lack of research from the global south, where the use of social media and the visibility of fake news are both on the rise. In this study, I analyze misinformation in Facebook posts during the 2019 Indian national election campaign to answer the question: Do posts labelled as “False,” by fact-checkers reduce engagement with the misinformation post on Facebook?

I obtained partisan misinformation-labelled Facebook posts (N = 4,478) from CrowdTangle pertaining to 294 unique misinformation claims (labelled False by fact-checkers) during the election campaign (1st March 2019 – 23rd May 2019) that were identified using fact-checked stories. I noted the date of the label as False. Posts were highly engaged with before the “False” label, and the label did reduce engagement a little. Results show that users continued to share posts labelled as False (read misinformation). Negative misinformation posts were more often shared compared to positive misinformation posts. I define negative misinformation as posts that attack a party, and positive as posts that self-promote.

Even 60 days after the fact-checking labels were added, the negative misinformation posts continued to circulate, depending on the topic of the misinformation claim. I discuss these topics in the paper. However, a comparison of overall engagement with posts before and after posts were labelled as False (read misinformation) shows that there was some decrease in overall engagement (Shares and Likes). By contrast, positive misinformation posts were less often shared or liked once they were labelled False. This study has implications for social media platforms to strengthen and speed up the process of labelling posts as misinformation in order to reduce the spread.

Author

• Kiran Arabaghatta Basavaraj, University College London