Debunking Disinformation Myths, Part 3: The Prevalence and Impact of Fake News

Contrary to conventional wisdom, online fake news is rare and mainly consumed by a minority of social media users with pre-existing fringe views.

Aug 24, 2024

This is the third essay in my “Debunking Disinformation Myths” series.

This series aims to draw on a wide range of scientific evidence and arguments to challenge popular myths about disinformation and related topics such as social media, misinformation, and deepfakes. Although these essays are “contrarian” in that they contest popular ideas about disinformation among many politicians, pundits, and ordinary people, they stick close to mainstream scientific findings.

Moreover, I try to be thoughtful and balanced. Many critiques of misinformation research—“Big Disinfo”, the “Censorship Industrial Complex”, and so on—are hysterical, conspiratorial, uninformed, and biased. This series aims to be different. Lies, propaganda, and misleading content are problems, and social media generates novel informational challenges. Nevertheless, public discourse and conventional wisdom about these topics are often misguided and alarmist, resulting in bad analysis and decision-making.

The first essay challenged the popular idea that we are living through an unprecedented “epistemic crisis”, “disinformation age”, “misinformation age, or “post-truth era.” The second challenged the popular idea that modern disinformation research is completely objective and politically neutral. In this essay, I will turn to a third popular myth: that fake news is rampant online and highly consequential.

Myth: Fake news is widespread and harmful

“Disinformation” is typically defined as intentional misinformation. In the previous essay in this series and elsewhere, I pointed out that classifying content as disinformation is vulnerable to various biases and errors.

Perhaps to avoid such risks, much of the early post-2016 establishment panic about dis/misinformation focused on online fake news, technically defined as fabricated news stories designed to mimic real news. For example, if a website publishes a phoney news story alleging that the Pope endorsed Donald Trump for president, that is fake news. Because such stories are fabricated and easily debunked, classifying them as dis- or misinformation invites less controversy.

Survey evidence consistently shows that ordinary citizens think fake news is widespread and highly impactful in shaping political outcomes. For example, a 2019 poll found that 51% of Americans think fake news “is a very big problem in the country today.” To put this in perspective, this percentage is higher than those who believe the same about violent crime (49%), climate change (46%), racism (40%), illegal immigration (38%), terrorism (34%), and sexism (26%). A 2018 survey shows that Americans also think 65% of the news they see on social media is misinformation. The pundit class often expresses similar views about the prevalence and dangers of fake news.

Fake news is not widespread

As multiple scientific review articles document, this alarmist view of online fake news is unsupported by scientific evidence, at least when it comes to well-studied Western democracies (“the Global North”). Consider just a few examples:

Allcott and Gentzkow (2017) find that the average US citizen saw and remembered just 1.14 fake news stories during the 2016 election.
Guess, Nyhan and Reifler (2018) find that most citizens (75%) didn’t visit a fake news website during the 2016 election.
Grinberg and colleagues (2020) find that on Twitter during the 2016 election, fake news only accounted for 6% of all news consumption.
Allen and colleagues (2020) find that on average fake news makes up only 0.15% of people’s overall media diet (i.e., once you include offline content like television news).

These and numerous other studies show that online fake news makes up a small percentage of people’s overall media diet. Moreover, such analyses tend to overestimate the prevalence of online fake news because they typically measure its prevalence at the source level, treating 100% of the content from untrustworthy websites that sometimes publish fake news as fake news. However, even extremely low-quality, fringe news websites often refrain from publishing outright fake news.

The fake news minority

In addition, existing estimates of the prevalence of online fake news are also misleading because they focus on average exposure. Another robust finding from scientific research into online fake news is that exposure to such content is overwhelmingly concentrated among a small percentage of the population.

In other words, whereas most people encounter very little fake news, a small minority encounters a lot, which inflates the average. Moreover, this “misinformation minority” is not a random sample of the population who fell down online “rabbit holes”. As Joe Uscinski puts it, someone who engages with QAnon content online

“wasn’t looking at recipes on YouTube then slipped on a banana peel and got inadvertently pulled down the QAnon rabbit hole. Maybe they were on YouTube looking for fringe conspiracy theories or extremist religious stuff; maybe they were already into all sorts of Bible conspiracy nonsense. The internet didn’t persuade them of some foreign idea. It gave them exactly what they already believed.”

More generally, those who engage with large amounts of online fake news have particular pre-existing traits such as extremist attitudes, highly conspiratorial worldviews, and intense distrust of (and often hostility towards) establishment institutions. These traits lead them to seek information that aligns with their worldviews and supports their intentions. Because their worldviews are typically wrong and their intentions are often antisocial, this means seeking out a lot of extremist, low-quality information, including fake news. Nevertheless, such content primarily “preaches to the choir.” Given this, it is unlikely to have dramatic consequences in shaping the audience’s beliefs and behaviours, except for reinforcing pre-existing ideas.

Summary

These findings are entirely mainstream discoveries of modern social science. Nevertheless, they contradict much popular discourse and conventional wisdom about fake news. As a recent review article in Nature puts it,

“The conclusions of academic research are clear – exposure to misinformation is low as a percentage of people’s information diets and concentrated among a small minority. However, this reality is not reflected in public discourse about social media.”

What is going on?

Many things might explain this disconnect. For example, mainstream media outlets have an obvious incentive to spread alarmism about—and hence discredit—their social media competitors. "Fake news” also offers an appealing, simple explanation of complex problems, which the expert class can point to without addressing deeper social, political, economic, and cultural factors in society. Finally, people tend to exaggerate the prevalence and dangers of threats across the board, especially when associated with fancy new technologies.

Objections

1. Big Numbers

Much of the alarmism surrounding online fake news is fuelled by reports of the large numbers of people exposed to specific fake news stories. For example, social scientists or journalists will report that some huge number of people—hundreds of thousands, millions, or even tens of millions of people—were exposed to fake news.

However, such isolated statistics are highly misleading without providing additional context about the vast scale of content exposure on the internet. To illustrate, on Facebook in the first quarter of 2023, the 20 most widely viewed posts had 777.6 million views in the USA alone, yet this represented just 0.04% of total views during that period.

Consider the influential story that content generated by Russian trolls potentially reached as many as 126 million Americans via Facebook before the 2016 US presidential election. That sounds like a lot. Moreover, the big number fits the influential narrative among liberals that Trump’s 2016 presidential victory was due to Russian interference. However, such content represented only 0.004% of the content citizens saw in the Facebook news feed. Combined with the fact that people are difficult to influence and Russian disinformation preached to the choir, this is why research provides little evidence that Russian disinformation campaigns influenced the 2016 presidential election.

2. Broadening the meaning of disinformation

Another objection is that online fake news is only one form of online disinformation (which is only one form of online misinformation). There are many other forms of problematic and misleading content online, including false, misleading, and inflammatory commentary from politicians, pundits, and influencers, as well as accurate content (e.g., true claims, real videos, etc.) that is nevertheless misleading because it is cherry-picked or published without appropriate context.

Broadening the definition of online disinformation to include a broader range of content will obviously increase estimates of its prevalence. However, as I have argued elsewhere, the problem with more expansive definitions of disinformation is that they will never be applied reliably and impartially. For example, many claims about social media-based disinformation by journalists and social scientists are extremely misleading. Would those worrying about “disinformation” also characterise such content as “online disinformation” on these expansive definitions? Of course not. This is a microcosm of a much bigger issue. For example, if one includes true but unrepresentative information in one’s definition of “disinformation”, that includes pretty much all mainstream news media.

3. Fake news doesn’t need to reach a large audience to be dangerous

Another objection is that online fake news does not need to reach large audiences to have dramatic effects. For example, elections are often decided on small margins, which means fake news only needs to persuade a small number of people to have dramatic political effects. Moreover, online fake news could also play a role in enabling fringe or extremist groups to coordinate, even if it does not persuade anyone to adopt radically different beliefs. People will often point to things like 2021’s January 6 Capitol attack in the United States or the United Kingdom’s 2024 race riots to defend this view.

This is a reasonable point. Even if online fake news is relatively rare and preaches to the choir, it can still have harmful consequences. Whether or not it has such consequences must be decided on a case-by-case basis.

However, this argument is also misleading in many contexts.

First, it typically ignores one of the main lessons of research: engagement with fake news is not just very low on average but overwhelmingly preaches to the choir. Who engages with lots of pro-Trump fake news? Trump supporters! Given this, it seems unlikely such content is shaping enough people’s political attitudes or voting behaviours to swing elections or cause major changes in political outcomes.

Second, when one looks into specific cases, such as the UK riots, people frequently exaggerate the impact of fake news. Moreover, even when online fake news plays a role in such events, it does not follow that similar events would be less frequent or less severe without social media. As a team of researchers put it,

“In a very narrow sense,… these outcomes could not have occurred exactly as they did without the platforms. But the causal question of interest requires consideration of a more difficult counterfactual: whether social media sites like Facebook cause violent protests or anti-vaccine movements to be more severe or prevalent than they would have been in their absence. The answer to this question is unclear.”

Finally, a motte-and-bailey fallacy occurs in this context.

To begin with, people make extreme claims about the threat of online fake news. This is the “bailey”. For example, earlier this year, the World Economic Forum surveyed roughly 1,500 experts to put together a global risk report, which listed disinformation and misinformation as the top global risk over the next two years, ahead of things like military conflict, nuclear war, economic catastrophe, and so on. The justification of this ranking appealed primarily to fears about AI-based disinformation, most of which are founded on the misguided assumptions I have identified in this essay.

When people point out that these alarmist narratives are unsupported, the “motte” comes: “Are you denying that online fake news sometimes has harmful consequences?!”.

Online fake news can be dangerous in some circumstances, and it is reasonable to worry about that. However, this makes it all the more important to have an accurate—evidence-based, non-alarmist—understanding of its prevalence and characteristics.

4. The Global South

Whenever you point out that online fake news is relatively rare and primarily preaches to the choir, people respond that this analysis is based on research on “the Global North” and hence ignores many countries throughout Africa, Asia, and Latin America.

This is true. The claims about the prevalence and impact of fake news in this series are explicitly focused on affluent Western democracies. (These are the societies where much of the alarm about social media and fake news has arisen.) The situation is likely very different in underdeveloped, poor, non-democratic societies with significant corruption and low levels of education and literacy. However, we currently need more data on this.

5. Elon Musk

Estimates of the prevalence of online fake news were mostly conducted prior to Elon Musk’s takeover of Twitter. Although there does not seem to be much rigorous scientific research on this, I suspect fake news is more prevalent on the platform now than it used to be, not least because Musk himself sometimes posts it.

Nevertheless, it is important not to exaggerate the impact of this change. For example, most people (≈75%) in the UK and US do not use Twitter, and roughly half of these users do not engage with news on the platform. Moreover, those users who engage with lots of political content online are, like most politically engaged citizens, likely to have entrenched political identities and worldviews that will not be easily influenced by exposure to a slightly larger amount of online fake news.

It is also important not to exaggerate the magnitude of the change. Although fake news likely has become more prevalent on the platform, most of the troubling content on the platform is not fake news. Elon Musk tweets lots of hateful, uninformed, and conspiratorial content, as do numerous accounts that achieved a bigger profile since he took over (not least because many were previously banned from the platform). Nevertheless, there is a difference between online fake news and stupid, biased, low-quality contributions to public debate, which are widespread, ancient, and not specific to social media.

Moreover, Musk also brought in some good innovations, such as Community Notes. This has many virtues, one of which is that certain forms of blatant left-wing misinformation (e.g. here) now get flagged on the platform, which probably would not have happened before Musk’s takeover. However, these are just my impressions. Given the absence of rigorous data and analysis, it is difficult to say anything definitive.

Conspicuous Cognition

Discussion about this post