Visual Misinformation in Brazil: Platforms, Partisanship, and Literacy

• In Event: Misinformation in the Global South

Abstract

The spread of online misinformation has been a topic of intense scholarly concern, but most research to date has been focused on examining textual misinformation. Research that has considered the effects of visuals has been primarily concerned with deepfakes, i.e., manipulated videos created using artificial intelligence (e.g., Vaccari & Chadwick, 2020), with less attention to images routinely shared on social media and across messaging applications (for an exception, see Hameleers et al., 2020). We are particularly interested in whether visual misinformation is more credible and shareable in the context of messaging applications, as prior research has focused on more public social media platforms. Messaging apps are often neglected by scholarship focused on the Global North but are a primary concern for those studying mis- and disinformation in the Global South due to their high popularity and reach. Messaging applications also represent a significant shift in how we understand information flows and credibility, as there is no central, algorithmically curated news feed and information primarily circulates among peers through one-to-one and group encrypted chats (Rossini et al., 2021). Thus, platform dynamics are particularly relevant in the context of Brazil, where WhatsApp is considered a source of news almost as central as social media by a majority of internet users (Newman et al., 2020). Hence, we ask an exploratory question to examine whether users exposed to treatments based on WhatsApp mock-ups responded differently than those exposed to Facebook treatments (RQ1).

The role of partisanship has been one of the primary focuses of misinformation scholarship, with research in different countries and contexts suggesting that ideologically congruent misinformation is 1) more credible and 2) potentially harder to debunk (Badrinathan, 2021; Munger et al., 2020; Thorson, 2016). In Brazil, scholars have highlighted the role of partisanship and ideology to understand beliefs in, and spread of, false information online (Recuero et al., 2021; Rossini et al., 2021). Considering the role of motivated reasoning in information processing (Pennycook & Rand, 2019), we hypothesize that partisans will be more likely to believe, and share, ideologically congruent visual misinformation than incongruent misinformation (H1). We further ask whether these effects are consistent across platforms (RQ2).

We also investigate the extent to which credibility and shareability are vary depending on participants’ digital literacy. Scholarship focused on mitigating the detrimental effects of misinformation has investigated whether people’s digital skills and abilities can help them identify false information, with mixed results (Badrinathan, 2021; Roozenbeek et al., 2020). In our study, we consider whether higher digital literacy skills help users assess the credibility of the treatments, as well as their willingness to share the false information they include. We hypothesize that users with lower levels of literacy skills will be more likely to believe in, and share, misinformation (H2). We also assess whether digital literacy plays a different role depending on the platform where misinformation is shown.

To shed light on these issues, we conducted a vignette survey experiment to examine the effects of visual misinformation in different platforms (Facebook and WhatsApp), considering the role of partisanship and digital literacy skills in explaining the extent to which people believe in, and share, visual misinformation. We surveyed 3,000 respondents, recruited to resemble the Brazilian population with internet access in terms of gender, age, and education, and asked them to complete an online survey administered by Qualtrics. Before exposing participants to the treatments, we asked them a series of questions measuring key political attitudes and behaviors, digital media consumption, and digital literacy. Participants were then randomly allocated to three groups (Facebook, WhatsApp, no platform) and exposed to a series of mock-ups containing a misinforming text and an image, drawing a total of eight different topics, six political (evenly split by ideological leaning) and two non-political. Treatments were created based on misinformation that had been authoritatively debunked by a reputable fact-checking organization in Brazil and pre-tested using Amazon Mechanical Turk. After seeing each treatment, participants answered a series of questions about the credibility and shareability of the misinformation they had been exposed to. Fieldwork is currently in progress and will be completed by March 2022.

Authors

• Patricia Rossini, University of Glasgow

• Raquel C Recuero, Universidade Federal de Pelotas

• Cristian Vaccari, Loughborough University

• Andrew Ross