BackHome
Refresh: Off
Individual Submission Summary
Share...

Direct link:

Visual Misinformation in Global Perspective: Platforms, Devices, and Users

Fri, September 16, 4:00 to 5:30pm, TBA

Abstract

The mass spread of misinformation and disinformation on social media is a key challenge for contemporary public discourse. Visual or multimodal misinformation may be particularly harmful, but we know comparatively little about its effects because most research on misinformation to date has analyzed textual content (Dan et al., 2021). This study aims to fill this gap based on a large-scale, preregistered experiment in six countries across the world.
The power of visual communication has been a classic object of inquiry in political communication research (Graber, 1990; Zillmann et al., 1999). Visuals enhance the transmission of information by helping citizens establish and retrieve memories (Stenberg, 2006; Witten & Knudsen, 2005). Images prompt audiences’ attention and emotional reactions to news (Powell et al., 2015), which leads citizens to question visuals less than text (Geise & Baden, 2015). Pictures can increase belief in false claims by enhancing familiarity with the information (Newman et al., 2012; 2015). The recent growth of platforms such as Instagram, Snapchat, and TikTok, which primarily emphasize still and moving images, highlights the relevance of visual content in contemporary social media.
In light of the potential impact of visual content in how citizens receive and interpret messages, we investigate whether social media users are more likely to believe, share, and search online to verify false information seen on social media when text is combined with images as opposed to text only, and what role the emotions elicited by images play in these processes. To this end, we test preregistered hypotheses with a large-scale experiment embedded in surveys (N per country = 3,000, N all countries = 18,000) in six countries that comprise more than 700 million internet users across four continents: Brazil, Hong Kong, Indonesia, Israel, South Africa, and the United States. Our comparative research design will enable us to assess the robustness of our findings and identify systematic sources of variation in the data.
Participants were asked a series of questions measuring demographics, political attitudes and behaviors, and social media use. Then, they saw mocked-up social media posts containing misinformation on five topics, in random order and randomly selected among eight topics we identified in each country based on misinformation that had been debunked by a reputable fact-checker. For each topic, we randomized whether participants saw a post in a text-only format or in a format that combined text and one of six different images we pre-tested to elicit different levels of negative emotions. To disentangle the role of different digital platforms, we also randomized whether participants saw mock-ups that resembled either a) Facebook, b) the main mobile messaging app in their country (Facebook Messenger or WhatsApp), or c) no platform interface. After seeing all the five posts they were randomly assigned to, participants answered questions that for all the eight issues measured: (1) whether they believe the content of the misinforming post, (2) how likely they are to share such content, and (3) how likely they are to go online to gather additional information on it. Given that each participant saw a randomly varied subset of five of the eight topics, the remaining unseen three topics serve as a within-subjects control group. As such, each participant is part of a treated group for five topics and of the control group for three topics.
The experiments have already been completed in three countries and fieldwork will be finalized in all remaining countries by April 2022. As we have already written most of the code required to analyze the data, we are very confident that we will be able to submit a complete manuscript by the deadline for the special issue.
Preliminary analyses on US data suggest that participants’ likelihood to believe and share misinformation increased when they saw misinformation presented with a combination of text and images compared with text only. The results also highlight that different images elicit different levels of negative emotions, and there is evidence that these emotions are associated with our key outcomes. By shedding light on the distinctive function of images in facilitating the spread of misinformation on these platforms, this study expands our knowledge of one of social media’s most problematic contributions to contemporary politics.

Authors