By studying how more than 16,000 American registered voters interacted with fake news sources on Twitter during the 2016 U.S. presidential election, researchers report that engagement with fake news was extremely concentrated.
Only a small fraction of Twitter users accounted for the vast majority of fake news exposures and shares, they say, many among them older, conservative and politically engaged. This work, like past studies, shows that the vast majority of exposure to political news, across all groups, still came from popular non-fake news sources, which is “reassuring,” say the authors; it’s in contrast to claims of fake news garnering more engagement than real news during the election.
Thrust into the spotlight during the 2016 U.S. presidential election season, the concept of fake news to describe the sharing of misinformation on social media has quickly become ingrained in the modern zeitgeist. Newly developing research on the topic has attempted to explain its broader impact on society.
However, the scope and scale of the phenomenon has been difficult to fully grasp – complicated by the challenge of measuring human behavior through social media data and obfuscated further by the role of automated “bot” accounts.
To overcome these challenges and better understand how real humans interact with fake news, Nir Grinberg and colleagues leveraged Twitter data linked to public voter registration records, studying the tweets sent by 16,442 accounts from August to December 2016 (the U.S. election season).
To define fake news, the authors followed past work defining fake news outlets as those that have the trappings of legitimately produced news, but “lack the news media’s editorial norms…” (The attribution of “fakeness” is thus not at the level of the story but that of the publisher.)
According to Grinberg et al.’s findings, only about 1% of the Twitter users they studied accounted for 80% of exposures to fake news content. Furthermore, 0.1% were responsible for 81% of the fake news shared. For most people – voters spanning the political spectrum – political news exposure came from factual media outlets.
In a related Perspective, Derek Ruths discusses Grinberg et al.’s study within the context of misinformation research writ large. According to Ruths, studies on the topic that appear to report competing views are often actually complementary, each providing insight into a single cog of what could be described as a “complex misinformation machine.”
Conflicting conclusions from similar studies could be due to a general lack of consensus on how fake news is defined. Grinberg et al., for example, approach the question by looking at the sources of fake news and their publication practices, while in a 2018 study also published in Science, Vosoughi et al. defined fake news at the level of the individual story. The two studies drew seemingly conflicting conclusions.
However, Ruths emphasizes that when underlying approaches are more carefully considered, the two studies together tell a detailed, cohesive story: Grinberg et al.’s results indicate that specific and small communities engage with fake news sources, and that publishers with weak journalistic standards are far more likely to produce misinformation.
Vosoughi et al. continue this trajectory by showing that, once produced, fake news stories spread with greater effect and speed than true stories. The findings of Grinberg et al. also show conservatives’ disposition for consuming and sharing fake news.
However, Ruths suggests these conclusions may not be as cut and dry as they seem, and suggests a more likely scenario where liberals, too, embed and spread misinformation in ways that have not yet been identified – an understanding of which is central to placing fake news within the context of larger systems of misinformation.