Misinformation & Educative Interventions: Experimental Evidence from India

• In Event: Misinformation and Fake News in Developing Countries

Abstract

Since the 2016 US election, the effects of “fake news” have received considerable attention in American scholarship, with academics pointing out its prevalence, effects on opinion, and electoral outcomes (Alcott & Gentzkow, 2017; Silverman, 2016). In response, several studies tested interventions to address these concerns, from identifying demographic subgroups vulnerable to fake news (Lazer et al., 2018) to providing warning labels on articles (Nyhan et al., 2017). Of note, however, is that all these studies focus on misinformation in the American context. But fake news is far from being an American problem: According to data journalism outlet IndiaSpend, 33 were killed in mob violence incidents in India in 2018 due to the spread of misinformation.

In this paper, I test the efficacy of a persuasive fact-checking intervention on respondents’ ability to identify fake news, in the context of the ongoing 2019 general elections in India.
Using a field experiment with 1200 respondents in Bihar, India, I seek to answer the following research questions: 1) Do strong fact-checking interventions increase respondents’ ability to identify fake news? 2) How does motivated reasoning affect partisans’ response to treatment? 3) Who is most vulnerable to misinformation? Do prior digital literacy, age, political knowledge, and media trust influence fact-checking?

The experiment involves an education, literacy training and fact-checking phase, followed by a measurement phase. In an hour-long intervention, treatment group respondents were shown demonstrations of two tools to verify information: reverse-google searching images, and navigating a fact-checking website. They also received several tips to prevent falling for fake news. Treatment group respondents were also given a flyer with these tips, along with examples of four political fake stories with their corrections. After a two week period during which respondents voted in the 2019 general election in India, I revisited respondent households to measure their ability to identify fake news. Respondents were given a series of news stories that varied in salience, political slant, and veracity. For each story, I measured respondent ability to identify the story as fake as well as confidence in that ability.

My experiment shows that an hour-long, strong persuasive and educative treatment is not sufficient to help respondents combat fake news. Importantly the average treatment effect was not significantly distinguishable from zero even though the experiment occurred during the election (when the most fake news is pedaled). Finding that a bundled, hour-long and costly intervention does not move people’s prior attitudes on fake news is testimony to the permeating, indestructible and detrimental effects of misinformation in this context. My findings challenge conventional wisdom in American politics that subtle priming treatments such as disputed tags can reduce fake news consumption. These findings also confirm qualitative evidence about the unique nature of social media in developing states that are new to the internet lending them particularly rife and vulnerable to misinformation.

While there is no evidence of a non-zero average treatment effect, there are significant treatment effects among some subgroups. BJP (right-wing) partisans in my sample possess a highly consolidated partisan identity: they are less likely to be able to identify news as fake when they receive the intervention, but only when the false news paints their party in a good light. These partisans do not demonstrate similar misinformation identification capacities for news that was not pro-attitudinal. Theoretically, this illustration is akin to a partisan backfiring effect: the treatment appears to make BJP partisans less likely to identify pro-partisan stories as fake; this effect does not obtain for non-BJP partisans at all, or for BJP partisans and pro-BJP fake stories. This is consistent with findings in the American politics literature demonstrating that partisans cheerlead for their party and are likely to respond expressively to emotionally loaded questions (Gerber and Huber 2009). It also challenges the contention that Indians lack consolidated, strong partisan identities (Chibber and Verma, 2018). I demonstrate that party identity in India is more polarized than previously thought, particularly with BJP partisans.

This is the first analysis of the effect of educative and persuasive interventions on misinformation in India. To the extent that consuming misinformation imposes costs and given that a strong, in-person treatment does not bring desired results, this study hopes to spark a research agenda on the ways to create an informed Indian citizenry armed with tools to fight misinformation as the commercial and partisan incentives to spread fake news increase in the near future.

Author

• Sumitra Badrinathan, University of Oxford