Donald J. Trump’s supporters were probably heartened in September, when, according to an article shared nearly a million times on Facebook, the candidate received an endorsement from Pope Francis. Their opinions on Hillary Clinton may have soured even further after reading a Denver Guardian article that also spread widely on Facebook, which reported days before the election that an F.B.I. agent suspected of involvement in leaking Mrs. Clinton’s emails was found dead in an apparent murder-suicide.
There is just one problem with these articles: They were completely fake.
The pope, a vociferous advocate for refugees, never endorsed anyone. The Denver Guardian doesn’t exist. Yet thanks to Facebook, both of these articles were seen by potentially millions of people. Although corrections also circulated on the social network, they barely registered compared with the reach of the original fabrications.
This is not an anomaly: I encountered thousands of such fake stories last year on social media — and so did American voters, 44 percent of whom use Facebook to get news.
Mark Zuckerberg, Facebook’s chief, believes that it is “a pretty crazy idea” that “fake news on Facebook, which is a very small amount of content, influenced the election in any way.” In holding fast to the claim that his company has little effect on how people make up their minds, Mr. Zuckerberg is doing real damage to American democracy — and to the world.
He is also contradicting Facebook’s own research.
In 2010, researchers working with Facebook conducted an experiment on 61 million users in the United States right before the midterm elections. One group was shown a “go vote” message as a plain box, while another group saw the same message with a tiny addition: thumbnail pictures of their Facebook friends who had clicked on “I voted.” Using public voter rolls to compare the groups after the election, the researchers concluded that the second post had turned out hundreds of thousands of voters.
In 2012, Facebook researchers again secretly tweaked the newsfeed for an experiment: Some people were shown slightly more positive posts, while others were shown slightly more negative posts. Those shown more upbeat posts in turn posted significantly more of their own upbeat posts; those shown more downbeat posts responded in kind. Decades of other research concurs that people are influenced by their peers and social networks.
All of this renders preposterous Mr. Zuckerberg’s claim that Facebook, a major conduit for information in our society, has “no influence.”
The problem with Facebook’s influence on political discourse is not limited to the dissemination of fake news. It’s also about echo chambers. The company’s algorithm chooses which updates appear higher up in users’ newsfeeds and which are buried. Humans already tend to cluster among like-minded people and seek news that confirms their biases. Facebook’s research shows that the company’s algorithm encourages this by somewhat prioritizing updates that users find comforting.
I’ve seen this firsthand. While many of my Facebook friends in the United States lean Democratic, I do have friends who voted for Mr. Trump. But I had to go hunting for their posts because Facebook’s algorithm almost never showed them to me; for whatever reason the algorithm wrongly assumed that I wasn’t interested in their views.
Content geared toward these algorithmically fueled bubbles is financially rewarding. That’s why YouTube has a similar feature in which it recommends videos based on what a visitor has already watched.
It’s also why, according to a report in BuzzFeed News, a bunch of young people in a town in Macedonia ran more than a hundred pro-Trump websites full of fake news. Their fabricated article citing anonymous F.B.I. sources claiming Hillary Clinton would be indicted, for example, got more than 140,000 shares on Facebook and may well have been viewed by millions of people since each share is potentially seen by hundreds of users. Even if each view generates only a fraction of a penny, that adds up to serious money.
Of course, fake news alone doesn’t explain the outcome of this election. People vote the way they do for a variety of reasons, but their information diet is a crucial part of the picture.
After the election, Mr. Zuckerberg claimed that the fake news was a problem on “both sides” of the race. There are, of course, viral fake anti-Trump memes, but reporters have found that the spread of false news is far more common on the right than it is on the left.
The Macedonian teenagers found this, too. They had experimented with left-leaning or pro-Bernie Sanders content, but gave up when they found it wasn’t as reliable a source of income as pro-Trump content. But even if Mr. Zuckerberg were right and fake news were equally popular on both sides, it would still be a profound problem.
Only Facebook has the data that can exactly reveal how fake news, hoaxes and misinformation spread, how much there is of it, who creates and who reads it, and how much influence it may have. Unfortunately, Facebook exercises complete control over access to this data by independent researchers. It’s as if tobacco companies controlled access to all medical and hospital records.
These are not easy problems to solve, but there is a lot Facebook could do. When the company decided it wanted to reduce spam, it established a policy that limited its spread. If Facebook had the same kind of zeal about fake news, it could minimize its spread, too.
If anything, Facebook has been moving in the wrong direction. It recently fired its (already too few) editors responsible for weeding out fake news from its trending topics section. Unsurprisingly, the section was then flooded with even more spurious articles.
This June, just as the election season was gearing up, Facebook tweaked its algorithm to play down posts from news outlets and to increase updates shared by friends and family. The reasonable explanation is that that’s what people want to see. Did this mean less reputable stories spread quickly through social networks while real journalism got depressed? Only Facebook knows. Worse, Facebook doesn’t flag or mark credible news websites: The article from The Denver Guardian, a paper that doesn’t even exist, has the same format on the platform as an article from The Denver Post, a real newspaper.
In addition to doing more to weed out lies and false propaganda, Facebook could tweak its algorithm so that it does less to reinforce users’ existing beliefs, and more to present factual information. This may seem difficult, but perhaps the Silicon Valley billionaires who helped create this problem should take it on before setting out to colonize Mars.
Facebook should also allow truly independent researchers to collaborate with its data team to understand and mitigate these problems. A more balanced newsfeed might lead to less “engagement,” but Facebook, with a market capitalization of more than $300 billion and no competitor in sight, can afford this.
This should not be seen as a partisan issue. The spread of false information online is corrosive for society at large. In a 2012 opinion essay in The Times, I cited the Obama campaign’s successful social media and data strategy to warn about the potential dangers of polarization and distasteful political methods, like misinformation on social media.
And the dangers of Facebook’s current setup are not limited to the United States. The effects can be even more calamitous in countries with fewer checks and balances, and weaker institutions and independent media. In Myanmar, for example, misinformation on Facebook has reportedly helped fuel ethnic cleansing, creating an enormous refugee crisis.
Facebook may want to claim that it is remaining neutral, but that is a false and dangerous stance. The company’s business model, algorithms and policies entrench echo chambers and fuel the spread of misinformation.
Letting this stand is not neutrality; it amplifies the dangerous currents roiling the world. When Facebook is discussed in tomorrow’s history books, it will probably not be about its quarterly earnings reports and stock options.
Zeynep Tufekci is an associate professor at the University of North Carolina School of Information and Library Science and a contributing opinion writer.