"I Don't Think That's True, Bro!" An Experiment on Fact-checking Rumors in India

• In Event: Misinformation and Fake News in Developing Countries

Abstract

Online misinformation creates serious challenges for policy-makers. The challenge is especially acute in developing countries such as India, where misinformation tends to be disseminated on private, encrypted discussion apps such as WhatsApp. Faced with this challenge, platforms have encouraged users to correct the misinformed beliefs they encounter on the platform.

When, if at all, should we expect such a strategy to be effective? Specifically, to what extent does the source and the sophistication of the corrective messages posted by users affect one's ability to counter misinformation? To explore these questions, we experimentally evaluate the effect of different types of corrective, user-driven messages on the persistence of misinformed beliefs among a large sample of social media users in India (N=5104). This study constitutes one the first attempts to evaluate users' perceptions of misinformation and corrective messages on an encrypted discussion app, on which users by design control input. It is also one of only a handful of studies to focus on this question in India. This allows us to explore the extent to which social media misinformation and potential corrections are perceived in a context of low literacy and low digital literacy, and in which partisan identities may play a weaker role than in the US.

In keeping with existing findings on fact-checking on other social media platforms and/or in other contexts, we show that user-driven corrections can be effective relative to a no correction condition, albeit on some rumors and not others. We also show that the heterogeneous effects partisanship and motivated reasoning are negligible, pointing at important differences in the mechanisms through which misinformation persists across contexts (Nyhan and Reifler 2011). More importantly, we show that the source and the sophistication of corrective messages does not strongly condition their effect: in our experiment, brief and unsourced corrections achieve an effect comparable to that of corrections implying the existence of a fact-checking report by a variety of "credible" sources (domain experts and specialized fact-checkers alike). This suggests that corrective messages may need to be frequent rather than sourced or sophisticated, and that merely signaling a problem with the credibility of a claim (regardless of how detailed this signaling is) may go a long way in reducing overall rates of misinformation.

This has implications for both users and platforms. Rather than unrealistically expecting users to refer to fact-checking reports (a practice they are unlikely to engage in in the first place), users should be encouraged to effectively ``sound off" as easily as possible and express their doubts about on-platform claims. The group-based nature of chat applications such as WhatsApp may ensure that social pressures to not give in to misinformation decrease expressed beliefs in fake news. Our results suggest that creating a simple ``button" to express doubt in reference to on-platform claims may be a complementary, cost-effective way to limit rates of beliefs in common online rumors.

Authors

• Simon Chauchard, Leiden University

• D.J. Flynn, IE University

• Sumitra Badrinathan, University of Oxford