Science used to be done by a select few, inside high-tech laboratories and dusty university offices. Studies would go on for years, and when they were published, results would be locked behind journal paywalls, ready to be read by a handful of fellow specialists. But this was before we had a pandemic on our hands.
Now, everyone from billionaire Elon Musk to your high school friend on Facebook is an “armchair epidemiologist.” Data, analyses and opinions on Covid-19 are flooding social media feeds and news sites. And why not? Even public health experts are struggling to make sense of this extremely unusual situation. This crisis affects everyone, so why not offer your views on how to dampen the R rate?
But, according to the World Health Organisation, we are fighting not only a pandemic; we are also fighting an infodemic. And, it’s much harder to identify fake news than previously thought. Most of the Covid-19 misinformation contains some truth and authority, blurring the boundaries between fact and fiction.
Rise of the data bros
Certain posts exacerbate the problem. In March, Aaron Ginn, a Silicon Valley product manager, published a Medium article titled “Evidence Over hysteria—Covid-19.” He argued that the world was in panic mode, and that this hysteria “is pushing aside our protections as individual citizens and permanently harming our free, tolerant, open civil society.” Facing backlash, the article was soon taken down from Medium and Ginn was chastised by public health experts on Twitter. Renowned epidemiologist Carl Bergstrom tore apart Ginn’s analysis in a thread of 30 tweets. Bad modelling can lead to unintentional misinformation, which is especially dangerous when the novice modeller has a big online platform.
This is far from a unique case. Mehmet Alpaslan, an astrophysicist and science communicator in New York, tweeted that “a significant majority of online bros” believe that they are as skilled as epidemiologists because they have taken statistic courses and learned how to code. These so-called “data bros” speak over experts and offer mathematically complicated, misinformed analysis.
“I haven’t really seen a single actual statistician come out with these absurd Medium essays about how epidemiologists are all wrong” Alpaslan told me through Twitter. Data science can be used as a tool, often by men who like to explain things to people, and talk over them, using “cold hard facts and logic,” he continues.
These “data bros” throw around weighty terms like “sampling size” and “machine learning” without really having much substance in their work, he says, and “they definitely do have some false sense of confidence when it comes to opining about things they don’t know about.” “I think data science is almost like a tool in some ways that allows men to extend their desire to explain things to people,” he continues.
Alpaslan does believe that good science can come out of outsiders coming into new disciplines, including his own. “But when it comes to a public health crisis this deadly, I think the risks of spreading misinformation are not worth it.”
There are clear examples of data bros getting things very wrong. But what about when an outside perspective offers invaluable help?
A Twitter DM gone right
David Yu is a sports data analyst in Canada, where he predicts trends in different ice hockey leagues. Yu confesses that at the beginning, he “actively resisted attempts at armchair epidemiology during this crisis.” But, in early April, Yu came across something which didn’t make any sense. He was looking at the model used by the White House to understand how Covid-19 spread. On closer inspection, Yu noticed that the number of Covid-19 deaths predicted by this model had suddenly dropped overnight. The swift change from 80-95k projected deaths to just over 60k alarmed him, especially because this influential model was being used by senior policymakers. Why had the numbers plunged downward so dramatically?
At first, Yu thought it could be due to individual states updating their figures. However, when he examined the data, it became clear that the drop was consistent throughout the US. Without warning, the scientists behind the model had changed the inner parameters for their data and now the curve of the model predicted that the Covid-19 death rate was falling quicker than previously assumed.
Yu felt compelled to investigate. “I was very concerned that these trendlines would give ammunition to people who wanted to relax social distancing,” he says. However, he was also cautious about his own lack of expertise. So, he decided to do what any millennial in his situation would do—he found a renowned infectious disease epidemiologist on Twitter and DMed them.
Carl Bergstrom, the professor who debunked Ginn’s analysis, replied to Yu that same night. Bergstrom was the perfect person to chat with—he had also recently co-written a book, Calling Bullshit, which argues that you don’t need specialist technical expertise to call out problems with data.
His Twitter message led to an unlikely but fruitful collaboration. Yu built a tool, Covid Projections, that compares different Covid-19 models across the US. By centralising and visualising all models in one place, his tool helps scientists understand how old and new models compare, giving people like Bergstrom “a little bit of time back in their day.” Bergstrom recently told the Atlantic that Yu’s tool offered him renewed clarity on Covid-19.
A crisis of trust
However, even if some contributions from laypeople are valuable, the sheer amount of information can make it difficult to know who to trust. Busy frontline workers such as medical doctors don’t have the time to sort through thousands of articles. Becky Jones, a medical student at McMaster University, saw this problem first hand, and decided to step in.
Alongside colleagues, she created an online filtering system which funnels in only useful scientific work. Anyone’s efforts can be included, although “we are only including about 30% of published articles,” says Jones, which gives a clue to how much unhelpful information is actually out there.
Organisations such as Crowdfight Covid-19 are also helping streamline research efforts by connecting people who want to offer their skills to Covid-19 scientists. The organisation has already accrued 40,000 volunteers, hopefully helping to put “armchair epidemiologist” accusations to rest.
Science reformulating its boundaries has positive aspects. We just need to make sure science is in the hands of people with know-how—and crucially, know the limits of their own knowledge—and not the all-explaining data bros.