On March 20th, an article about the novel coronavirus started to spread. “Evidence over hysteria — COVID-19” crystallized a viewpoint conservatives had been hawking for some time: the hysteria over the virus was overhyped. The United States needed to reopen the economy.
To a non-expert, the argument sounded compelling. The author, Silicon Valley growth hacker Aaron Ginn, used Johns Hopkins data to make his point. On Medium, the piece blended in with fact-checked articles from doctors and epidemiologists, until it was debunked in an eviscerating tweet thread by an actual infectious disease expert, Carl T. Bergstrom, a biology professor at the University of Washington. Shortly after, it was taken down.
In many ways, this was Medium working as intended. The company operates both as a blogging platform and as a media outlet. Some articles, written by professional journalists who work at one of Medium’s publications, are fact-checked. Others are written by conspiracy theorists and contain dangerous misinformation. After all, the company’s mission, according to CEO Ev Williams, is to level the playing field and encourage “ideas that come from anywhere.”
As the pandemic disrupts life in the US, Medium has made strides to stop the spread of misleading health news. Its own publications, like OneZero and Elemental, have covered COVID-19 with the journalistic ethics, editing, and fact-checking you’d expect from a traditional outlet. Medium also started an official COVID-19 blog to promote articles from verified experts. It rolled out a coronavirus content policy and hired a team of science editors.
But the decision to curate some content — to hire professional journalists and promote verified articles — has made it harder to tell fact from fiction on the platform. While user-generated pieces now have a warning at the top telling users the content isn’t fact-checked, they look otherwise identical to those written by medical experts or reporters. In some ways, this is the promise of Medium: to make the work of amateurs look professional.
Reading a 2,000-word article that contains misinformation about COVID-19 also seems notably different than reading a few of the same ideas in a tweet. It might not have mattered when Medium was a home for productivity hacks. But coronavirus misinformation could put people’s lives at risk.
The situation has forced Medium to wade deeper into the waters of content moderation, where big tech firms have been floundering for years. Now, the platform that was built as a home for the world’s “unique perspectives” is in the position of deciding which perspectives actually matter.
One of the most prominent voices on Medium about COVID-19 belongs to Tomás Pueyo, another Silicon Valley growth hacker. Pueyo wrote “Coronavirus: Why You Must Act Now,” which has over 40 million views, making it one of the most widely read articles about the virus.
This is the best-case scenario for Medium. Pueyo is a credible non-expert who has a lot to say but no clear place to say it. Also, he can make sense of data. He learned how to read complex academic papers while building social apps in Silicon Valley. When the virus started to spread in the US, he immersed himself in Johns Hopkins’ repository on GitHub to figure out what was going on.
Pueyo’s insight, which now seems obvious, was that novel coronavirus cases were growing exponentially. It was a pattern he recognized from when one of his apps had gone viral, adding 7 million users in three months. This comparison — viral apps to viral disease — would normally sound too convenient. This time, it actually worked.
Pueyo argued on Medium that the coronavirus problem was worse and more immediate than most people knew or expected. Overnight, his article exploded. Andrew Yang shared it, as did the psychologist Steven Pinker. It was quoted in The New York Times. BuzzFeed called it the “defining piece on the outbreak of COVID-19.”
The attention brought with it criticism, mostly focused on Pueyo’s background. He doesn’t have experience in epidemiology, and, to date, he has mostly built apps. One of these, Zoo World, helps people build virtual zoos.
Asked about this feedback, however, Pueyo seems delighted. “It’s great!” he says. “You need the checks and balances, especially in my case, because I’m nobody. I have no training in epidemiology.”
When A.J. Kay, an essayist in Tempe, Arizona, began learning about COVID-19 cases in the US, she didn’t think the information added up. It seemed off that the disease was being called “highly transmissible,” but only one person in Arizona person had it. (Testing for COVID-19 has notoriously lagged in the US, meaning that many people who may be sick with the disease aren’t counted in the official numbers.)
Kay isn’t an epidemiologist. Until last month, her most popular article on Medium was about having her breast implants removed. But she grew more suspicious when her ex-husband lost his job at a local hospital. If there was such a shortage of health care workers, why on earth was one being fired? Also, her child support payments were now at risk.
Kay compared the number of known COVID-19 cases with flu data from the Centers for Disease Control and Prevention (CDC) and came to a surprising conclusion: the virus had come to the United States far earlier than was previously reported. What people were predicting would be the peak — the menacing time when COVID-19 cases would overwhelm hospitals in cities across the US — was likely already happening. If she was right, the draconian measures state governments were enacting, like shutting down nonessential businesses, seemed woefully out of proportion with the problem.
In the article, Kay admitted the situation was personal. Last year, she’d discovered she had a tumor in her liver but was told it was too low-risk to remove. Recently, she’d begun developing symptoms that seemed to indicate the disease was spreading. She needed to get a scan but couldn’t because the procedure was considered elective. If Kay could convince people that the virus wasn’t as bad as they thought, she might be able to finally address the problem.
She also knew that her entire family had been extremely ill in early January. The first case of COVID-19 in Arizona was reported at the end of January. “I started putting things together in my head and I was like wait a minute…” she says.
Her article, “The Curve Is Already Flat,” got 275,000 views in the first 48 hours it was published. Like Ginn’s piece, it fed into the conservative narrative that the mainstream media had gotten this wrong. Kay says she didn’t intend to take a political stand, but she was happy to see the article go viral.
Two days after publishing, Kay received an email from Medium. “Due to elevated risk of potential harm to persons or public health, Medium’s Trust and Safety team has removed the following content,” it said. Medium had concluded that her article contained “health claims or advice which, if acted on, are likely to have detrimental health effects on persons or public safety.”
Kay had watched a similar situation play out with Ginn. She’d liked his article, calling it the “first coronavirus piece that made sense” and was surprised when Medium took it down. Now, she felt like she was also being censored for refusing to play into the popular narrative. “I was naive to think that couldn’t happen to me because I’m telling the truth,” she says.
Kay’s piece was quickly republished on a conservative blog, after receiving support from libertarian thinkers who derided Medium for censorship. An editor’s note at the top now reads: “This piece originally appeared on Medium.com and was removed without explanation or warning.” (Ginn’s piece was republished on a different conservative blog.)
Medium’s communications team wouldn’t say how, exactly, the company decides to take an article down. They also declined to comment on how they discover misinformation on the platform and wouldn’t say how many people are tasked with finding and regulating content. When asked why, specifically, Kay’s piece was taken down, a company spokesperson sent over the COVID-19 content policy, declining to comment on specifics. Medium did say that it takes into account the newsworthiness of the piece as well as the “context and nature of the posted information, the likelihood and severity of actual or potential harms, and applicable laws” when deciding whether to take action.
When asked about Ginn’s piece, Sandee Roston, the company’s head of communications, sent the following statement:
We’re giving careful scrutiny to coronavirus-related content on Medium to help stem misinformation that could be detrimental to public safety. The Ginn post was removed based on its violation of our Rules, specifically the risk analysis framework we use for ‘Controversial, Suspect and Extreme content.’ We’ve clarified these rules with our Covid-19 Content Policy to address more specific concerns around the evolving public health crisis, and we are actively taking down Covid-19 stories that violate this, such as the Kay piece.
Part of the problem Medium is facing is that the situation is changing so fast. Information that seems true one day can quickly look wrong or out of date by the next. In February, the surgeon general told everyone to stop buying masks. Now, the CDC is urging everyone to wear cloth masks all the time.
Journalism, by its nature, is rarely predictive. It tells people what’s true at the moment, often with a bit of analysis. “What seems to be true today may be wrong tomorrow,” wrote Charlie Warzel in The New York Times. BuzzFeed’s late-January article urging people to worry about the flu rather than the novel coronavirus looks laughable today — as does Vox’s January 31st tweet saying the situation would not become a deadly pandemic. At the time, however, these ideas were the prevailing wisdom.
In this period of uncertainty, Medium is giving people a place to speculate and offer concrete solutions — even if those solutions are unverified. For the writers, the stakes are low. If they’re wrong, most of the time, no one will notice. If they’re right, as Pueyo was, they could suddenly become an “expert” overnight.
From that lens, Medium’s mission seems like a relic from another era, one where people believed giving everyone a voice on the internet would make the world a better place. Williams admitted as such when he spoke to The New York Times in 2017 about why he thought the internet was broken. “I thought once everybody could speak freely and exchange information and ideas, the world is automatically going to be a better place,” he said. “I was wrong about that.”
It’s unclear how Medium solves that, and its opaque policies surrounding COVID-19 allow the company to enable the spread of misinformation while absolving itself of responsibility. Even Pueyo is self-aware of the platform’s dangers. When asked about people criticizing his article, he was okay with it: “I have no training in epidemiology,” he said. “You should definitely not trust me.”