The New Psychology of Fake News
New research helps explain why fake news can be so believable.
Posted Jun 29, 2017
Last year, an armed man entered a neighborhood restaurant called Comet Ping Pong in Washington, DC. He fired his assault rifle multiple times inside the family-friendly establishment. Thankfully, police apprehended the gunman and no one was hurt. This well-known eatery is not far from where I both live and practice, and I’ve been there on a number of occasions myself.
Needless to say it was far too close for comfort. And the man’s explanation of his actions was particularly disturbing: He was investigating a conspiracy theory about Hillary Clinton operating a child sex ring out of the restaurant. Since fake news has been essentially been waging war on truth and democracy, perhaps it’s unsurprising that it would eventually lead to an event with real guns.
But why is fake news so believable—despite that it’s so often preposterous? New research led by Gordon Pennycook of Yale University helps explain why it’s so easy to judge fake news as the real deal.
The research extends work on the illusory truth effect, which has demonstrated that familiarity increases the perceived accuracy of plausible statements, even if they are untrue. This is because we cognitively process material that is familiar to us with much greater ease, and this familiarity subsequently serves as a heuristic (a cognitive method of problem solving). For example, one study found that when participants repeated the false statement,“Chemosynthesis is the name of the process by which plants make their food,” they evaluated it as more accurate even though they reported that they knew the right answer.
Building on this body of work, Pennycook and his team conducted a series of three studies based on the 2016 election that investigated whether the illusory truth effect can help explain the phenomenon of fake news.
Study 1: Familiarity makes fake news seem accurate.
In the first study, Pennycook and his collaborators examined the relationship between familiarity and the perceived accuracy of news stories. To that end, they had participants view actual news headlines. Five stories were real news, and five were fake news (i.e., five items were factually accurate, and five items were complete falsehoods). The fake news headlines were sensational and highly partisan. For example, “Donald Trump Sent His Own Plane To Transport 200 Stranded Marines,” and “FBI Agent Suspected in Hillary Email Leaks Found Dead In Apparent Murder-Suicide.” Quite cleverly, the team used the same format as postings on Facebook—that is, a headline with a corresponding photograph and byline.
Participants then rated each headline for both familiarity and accuracy, and the researchers crunched the numbers. The results were striking. In line with expectations, familiar news headlines were much more likely to be perceived as accurate by comparison to unfamiliar news headlines. What’s more, even though the real news headlines were generally viewed as more accurate than the fake ones, familiarity was an overriding factor in that participants rated familiar fake news to be more accurate than unfamiliar real news.
Study 2: We use low-level thinking when we encounter fake news.
In their second study, Pennycook and his collaborators explored the cognitive mechanisms that underlie the relationship between familiarity and the perception of accuracy of news headlines. More specifically, they wanted to see if participants engaged low-level rather than high-level thinking processes. In other words, are people engaging reasoned and deliberative thought processes when reading headlines?
The investigators designed a three-stage experiment. In stage one, the familiarization stage, participants viewed equal numbers of real and fake news headlines and were asked whether they would share each headline on social media (of note, some were collected from Snopes.com, a website that fact-checks news stories). In stage two, the distractor stage, participants completed a questionnaire that asked about their demographic background. This was done to temporarily refocus their attention. In stage three, the assessment stage, participants were presented with the 12 news headlines they viewed in the familiarization stage and 12 new headlines (six fake news, six real news). Participants rated each headline for familiarity and accuracy.
The authors tested lower-level vs. higher level thinking processes in two ways. First, they analyzed the effect of providing participants with explicit information about accuracy as they would experience it in real life. Half of the participants were assigned to a “warning condition,” in which the fake news stories were flagged with a caution emoji and the text “Disputed by 3rd Party Fact-Checkers.” Facebook specifically developed this warning in order to stanch the spread of fake news.
The authors reasoned that if the familiarity effect is driven by high-level thinking processes, a warning should weaken its influence. According to this logic, a participant’s thinking would change from “I saw this before so it’s probably true” to “I saw this before and it’s probably not true.” Put another way, a direct warning should make participants question the accuracy of a fake news headline when seen in the familiarization stage.
However, if low-level thinking processes are being used, then an explicit warning would have no influence—and familiar headlines would still lead participants to perceive them as accurate, regardless of whether they were real or fake news. What were the results? The familiarity effect was found for fake news headlines in both the warning and no-warning conditions. That is, participants rated familiar fake news headlines that they were explicitly warned about as more accurate than unfamiliar fake news headlines that they were not warned about.
Pennycook and his team also examined the influence of “political concordance,” or agreement, on the familiarity effect. A sizable body of work has demonstrated that politically charged material can give rise to what’s known as motivated reasoning. This is reasoning that is biased in favor of conclusions that match one’s already established view. To test for the possible impact of motivated reasoning, the researchers presented participants with equal numbers of pro-Republican and pro-Democrat news headlines. In particular, they wanted to see if participants would be more likely to assess news headlines as being accurate if they were concordant with their views.
The authors contend that politically discordant headlines should activate high-level reasoning. What did they find? Low-level thinking processes were again at play, as the familiarity effect was found for both politically concordant and discordant headlines. In other words, participants judged familiar headlines to be more accurate, whether or not the stories matched their political views.
Study 3: Familiarity makes fake news seem more accurate, even after a week.
In the third study, Pennycook and his team followed the same three-stage procedure they used in Study 2. Yet the researchers made key changes. First, they increased the amount of time between the familiarization stage and the accuracy stage to assess the persistence of the familiarity effect. They also increased the length of the distractor stage by adding 20 more unrelated items to the demographics questionnaire. And the participants were invited to return for a follow-up session one week later, at which they were presented with the same headlines they had seen in the assessment stage as well as a set of new headlines not included in the first session. By designing the study this way, the researchers could look at how time influenced the stability of the familiarity effect over both the course of a testing session and the span of a week.
Again, headlines presented in the familiarization stage were rated as more accurate than novel headlines. And one week later, the investigators found a clear effect of familiarity on accuracy ratings. Moreover, the perceived accuracy of a story rose linearly with the number of times the participants had been exposed to that particular story, suggesting a sort of compounding effect. The relationship between number of exposures and accuracy remained when only fake news stories were under consideration. And it was found yet again for fake news in both the warning condition and the no-warning condition.
This series of studies helps to explain the pernicious reach of fake news, and how misinformation can fold into our memories with such disturbing ease. As the findings reveal, familiarity can breed trust. Though an explicit warning that a news item may be fake did help curb the problem to some degree, we clearly need to do more to stop the spread of fake news and falsehoods.
Pennycook, Gordon and Cannon, Tyrone D and Rand, David G., Prior Exposure Increases Perceived Accuracy of Fake News (April 30, 2017). Available at SSRN: https://ssrn.com/abstract=2958246