Who Would Volunteer to Fact-Check Twitter?

The site is hoping that its users will help stem the spread of lies—but first it has to inspire them.

The twitter logo on a black background with red scribbles
The Atlantic

I learned about the pilot test of Twitter’s new crowdsourced misinformation-labeling program the same way I learn about most news events that are relevant to my life: A bunch of Harry Styles fans were talking about it on my timeline.

Or rather, they were reacting to it, in quote-tweets, one after another, all saying essentially the same thing: “larries better hide,” “larries are over,” “it’s over for larries,” and, more explicitly, “i’m gonna use this against larries.” Maybe you don’t feel as though you need to know what any of this means before you read my take on whether Twitter’s Birdwatch program is a good idea, but I think it will help. Almost everything I know about life on the internet can be explained through “Larries,” a community of (mostly) former One Direction fans who believe that two of the boy band’s members, Louis Tomlinson and Harry Styles, are in love and secretly married, and that they have been for years, and that Tomlinson’s toddler son is either a child actor or the offspring of an ex-girlfriend’s stepdad. Larries thrived on Tumblr for years and are now experiencing an inexplicable resurgence on Twitter and TikTok, though everywhere they go, they find themselves locked in battle with a group of “Antis,” who are what they sound like. The Antis hate the Larries, often ridiculing them as delusional, which is sort of rude, and bringing up their history of viciously harassing Tomlinson’s family and friends, which is pretty fair.

Naturally, the Antis on Twitter took notice when the company announced, at the end of January, that it would launch an experiment in letting users decide which posts are true and which are false. This is not a fact-checking program, exactly, because it will not involve trained fact-checkers. Instead, participants in the Birdwatch pilot can identify tweets they find to be “misleading,” then submit notes explaining their stance—ideally these should link to reliable outside sources, and provide helpful context—as well as a judgment of how much harm the misinformation is likely to cause. In their reports, of course, the Antis say the Larries are causing “considerable harm.”

I’m telling this story because I think it’s funny, but also because it illustrates a core problem with Birdwatch at this stage: There is no good reason for most people to volunteer to fact-check Twitter. “Stans” are perhaps the category of Twitter user most willing to try it out, in a limited way, because they have a highly personal stake in the spread of an extremely specific form of, yes, misinformation. They have a “fave” to protect. They’re the most fun (though still volatile) result of Twitter’s competitive, gamified environment, and they clearly see something in this new tool that might benefit them. A few of them have already been more than willing to give of their time. But is anyone else going to feel the same way?

Twitter’s first big public push against misinformation came only last year. Throughout 2020, the company experimented with adding warning labels to tweets containing misleading or false statements—mostly about COVID-19 or the presidential election—but didn’t offer consistent or clear explanations of how the whole thing worked.

Meanwhile, the vice president of product, Keith Coleman, was working on Birdwatch, a different sort of system, and one that could end up being far more efficient and transparent than anything the company has tried before. His idea was to distribute the responsibility for catching and cornering misinformation across the entire Twitter community. Regular users might be able to move faster than machines when misinformation popped up in their social circles. The project also had the potential benefit of building trust with users. Coleman knew that some people didn’t like to see a large tech company exercise its power over national discourse in such explicit ways. “People might in fact really value something that comes from the bottom up, versus something that comes from any singular institution,” Coleman told me.

Birdwatch was made unusually transparent for the same reason. Anyone can inspect its ranking algorithm, which designates notes as “helpful” or “unhelpful,” or easily download all the notes that have been submitted through the program. Coleman acknowledged that the notes contain “quite a variety of quality,” so far, and said that his team expected this. If the ranking algorithm works properly, it will elevate the ones that elicit a “diverse consensus”—agreement among people who don’t tend to agree. Birdwatchers who consistently produce good notes will earn a positive reputation score, such that their subsequent notes will be assigned more weight—though it’s also possible to lose your reputation by writing notes that nobody rates as helpful.

As part of an ongoing experiment in locating the edges of my own sanity, I read through the notes one afternoon, about 3,000 of them altogether. (As of February 19, these were all the notes that had been submitted.) There were the Antis’ reports, as promised: “perpetuates a conspiracy theory”; “claims a relationship that is false”; “louis tomlinson is literally straight.” There were also notes defending Beyoncé from rumors that she had murdered her cousin, and defending members of the K-pop group BTS against unfair criticism of their “amazing vocals.” Dozens of notes had been added to tweets about the ongoing feud between Alexandria Ocasio-Cortez and Ted Cruz and their respective fan bases. (Some iteration of “Ted Cruz never once tried to have AOC killed” was in there at least 20 times.) Also, Nicki Minaj “did not sacrifice her dad. Plain and simple.” Anthony Fauci is “a normal doctor trying his best to keep Americans safe and healthy.” Donald Trump did not incite the riot at the Capitol, and Marjorie Taylor Greene isn’t a conspiracy theorist. According to someone.

The tech journalist Casey Newton recently described the most common question asked by Birdwatch users, up to this point, as “Is this insult fair?” That seems accurate to me. I did find some more consequential questions raised, such as “Who is actually part of Antifa?” and “Should we take the COVID-19 vaccine?” and “Does pineapple belong on pizza?” (My favorite note, if you were to make me choose, was added to a tweet from The Economist about the growing trend of insect-based food products. “We’re not going to eat insects,” a Birdwatcher countered.) But overwhelmingly, as I scrolled through the reports, the objections I found were personal.

In theory, the Antis—or other, more nefarious groups with single-minded concerns—will not be able to wage sustained war on their nemeses through the Birdwatch system. Their notes aren’t likely to produce the “diverse consensus” necessary for upranking, because other people are not likely to care about them. But that’s not going to stop them from trying. When I messaged a 20-year-old One Direction fan who had tweeted, “larries we coming for you” in response to the Birdwatch announcement, she told me that she wasn’t actually part of the pilot program, but in the future, she would definitely add notes to tweets from Larries. (She asked to be anonymous, because she doesn’t want her real name associated with her stan account.)

“Larries are spreading misinformation,” she said. She was happy to hear that other Antis had already started using Birdwatch to prosecute their case, and said that if I was able to get in touch with any of them, I should “send them a big hug.”

When Birdwatch was first announced, the company was met with an understandable knee-jerk reaction. Oh, good idea! Let’s put Twitter users in charge of deciding what is “reality.” Even seen in the best light, it represents an untested iteration of the existing system for labeling misinformation, which hasn’t yet been tested very well itself. Savvas Zannettou, a researcher at the Max Planck Institute for Informatics who has studied the impact of Twitter’s warning labels, cautions that Birdwatch could easily go wrong. “I will have to see how it works in practice,” he told me, but “I’m pretty confident that people will abuse and troll the system.” He mentioned the possibility of 4chan brigades; I saw several suggestions that warring stan armies will be the biggest abusers.

Still, Zannettou was optimistic about the idea and considers it a good one, as long as it can be executed in a way that minimizes manipulation. Other online spaces exist where crowdsourced definitions of fact and reality have produced reliable results. Wikipedia has gotten better and better over time. WikiHow, its founder told me in 2019, is a project that “doesn’t work in theory; it only works in practice.” These sites function because the people who contribute to them and edit them identify themselves as members of a community working toward a common goal. They also function because of hierarchies: Some experienced, committed editors have powers that others don’t. On Wikipedia these are called admins, a function that was added in the site’s first year to address vandalism by trolls and “editing wars” between egotistical contributors. In a 2010 study, researchers found that the majority of admins considered editing Wikipedia “rewarding” or “very rewarding,” and that 73 percent had been doing so for more than three years.

Coleman said he believes that participants in Birdwatch will be motivated by the desire to “elevate quality information in the world” and affect the national discourse. But it’s difficult to imagine Twitter users—who joke constantly about the ways in which the site has ruined their lives—coming together to dream of the platform as a valuable shared resource, let alone a “rewarding” or “very rewarding” place to spend their time. Andrea Forte, an associate professor at Drexel University who has looked into what drives Wikipedians to contribute, is skeptical. “The motivation for the kind of sustained, committed volunteer work around improving access to information that has built Wikipedia does not seem to me to be part of what’s happening on Birdwatch,” she told me. “Wikipedians are building something together; they aren’t just trying to correct someone who’s got wrong information on the internet.” (If there are any groups of Twitter users who are committed to building something together, they probably are the stan armies, but what they’re building is not really a public resource.)

Another study of what makes people want to write for Wikipedia suggested that online communities “must meaningfully structure participants’ contributions in a way that sustains involvement.” This means setting up chat rooms and mailing lists and discussion spaces, as well as providing a way for valuable users to be recognized by their peers. Wikipedia editors, for example, have user pages where they can display lists of articles they have meaningfully contributed to, as a sort of “elaborate ‘resume.’” Coleman said that Twitter might experiment with a tool that allows Birdwatch contributors to communicate with one another, and will definitely come up with some way of providing recognition to consistent contributors—either by making their reputation points visible to others or assigning some kind of digital badge.

In the Birdwatch pilot’s first three weeks, during which Twitter approved its first 1,000 participants, only 3,300 notes were submitted, many of them tests. The company will have to inspire a lot more people to be significantly more invested if this is ever to become a useful backstop against viral misinformation. Doing that will require a major shift on the platform. Wikipedia was a collectivist project from the very beginning, but Twitter is a for-profit company that operates what is often called a “data-mining operation” or a “hell site” by its own most dedicated users. The people who spend the most time there are practiced at hoarding and funneling attention—be it for themselves, a divisive political issue, or a pop star. They’re not accustomed to thinking of tweeting as a communal act toward a shared goal of filling the world with reliable information.

The internet is more personal than we tend to acknowledge, and the least terrifying but most significant problem with Birdwatch is that a lot of people who are going to use it will come in caring only about the stories—and perceived lies—that matter to them personally. Twitter has yet to provide a good reason for them to focus on anything else. In the meantime, Antis and Larries will continue their feud, which has employed many tools over many years, and they’ll experiment with Birdwatch until they figure out whether it can be bent to their will.