Abstract
In the midst of the COVID-19 pandemic and increasingly polarized politics, misinformation---including deliberately deceptive disinformation---has become widespread. This has been especially true of the Global South, where journalistic standards and fact-checking institutions are often comparatively less developed. The rise of misinformation has largely been fueled by social media, which has become a central source of information for many citizens in such contexts. Bolivia---the context of this study---has been particularly subjected to increasing misinformation, fueled by its political turmoil and the COVID-19 pandemic. In principle, journalists---the key producers of credible content---could play a central role in countering the dissemination of misinformation. However, journalists in low- and middle-income countries often lack training to detect and fact-check likely misinformation, the time to extensively corroborate source content, or the capacity or incentive to produce appealing articles to debunk it or compete against it. Given shrinking newsrooms, journalists have often amplified the dissemination of misinformation around COVID-19 and politics by reproducing viral news without fact-checking it, whether wittingly or unwittingly. To evaluate the extent to which providing training on misinformation and resources to journalists can overcome these challenges, Internews conducted a randomized intervention among journalists in Bolivia. Out of a pool of approximately 350 applicants, 145 journalists were screened as eligible to participate in the program. Out to those, 73 ``treated'' journalists were randomly selected to receive an invitation to participate in the program, while 72 ``control'' were not invited. The intervention provided treated local journalists with: (i) training to identify misinformation and engage in fact checking, so that they could produce stories that can combat and outcompete misinformation; (ii) seed funding to produce an original investigative journalistic content relating to misinformation; (iii) information about trending likely misinformation, as well as fact checks recently conducted by local fact checkers, to help journalists identify relevant topics for their regular work; and (iv) online materials advising on how to communicate fact checks. Ultimately, the goal was to assess whether this bundle of interventions could reduce the production and social media sharing of misinformation and increase the production and sharing of content that corrects misinformation. To evaluate the effectiveness of Internews' intervention, we will consider a series of outcomes. First, we will examine differences in knowledge and behavior between treated and control journalists about how to identify misinformation, engage in fact checking, and produce and share content to outcompete misinformation. Moreover, we will assess differences using outcomes that measure whether journalists report producing and sharing journalistic content to outcompete misinformation. To that end, we conducted a survey between March and April 2021---roughly a year after the training concluded. Second, we will look at differences in the characteristics and popularity of the content actually produced and shared by treated and control journalists, which complement the self-reported outcomes from the survey. To that end, we have scraped all publicly available content produced and shared online and on social media by treated and control journalists. We will use machine learning techniques to assess the extent to which the content produced and shared by treated journalists is of high quality in terms of informational content and style, whether it resembles misinformation, and whether it seeks to combat misinformation. Moreover, we will analyze differences in the popularity of content produced and shared by treated and journalists on social media, based on content shares and reactions on Facebook and Twitter. Measuring such popularity is important because the content that tries to debunk misinformation may not be as popular as the information that journalists are trying to debunk. Third, we will also assess citizen reactions to the content produced by the journalists that were and were not part of the program. We conducted a survey of Bolivian citizens between December 2020 and March 2021, which included a second randomized evaluation that showed respondents titles and articles produced by treated and control journalists on the topics of health and politics and elicited their perceptions about those titles and articles.