COVID-19 and Epistemic Hygiene
What the pandemic can tell us about how we ought to believe
Note: This piece discusses distressing events that have transpired during the COVID-19 outbreak.
As of this writing, we are heading into Spring Break amid a global pandemic.
South Florida beaches are packed with young revelers. Bars in New York and Nashville are full of patrons celebrating St. Patrick’s day, talking and drinking in close quarters. If your Facebook feed is like mine you are seeing a distressing amount of sentiments like “It’s just the flu!” and “It’s cooked-up media hysteria!” circulating. Arizona, Florida, and Illinois are planning to proceed with their primaries which will bring out tens of thousands of people — many of them members of high-risk populations — into high-school gyms, public libraries, and auditoriums.
Across the Atlantic but a mere tweet away, Italy has been under full lockdown for the better part of a week. All non-essential services like bars, restaurants, and salons have been ordered to close. Only grocery stores and pharmacies remain open. The streets are empty. In a video posted to Twitter, an Italian man compares the obituary sections in two editions of his local newspaper. The February 9 edition contains one and a half pages of obituaries. The March 13 edition contains ten full pages.
Europe no longer has the luxury of disbelief. In North America, we seem determined to hold on to it until the last possible moment — which is perhaps mere weeks or even days away. As Dr. Fabiano Di Marco, head of the respiratory unit of the Hospital Papa Giovanni XXIII in Bergamo, said in a recent episode of The Daily as he fought back tears: “I have some colleagues in other parts of Italy who are not prepared. They are doing the same thing we did three weeks ago. This is incredible. This new reality we are living in started on the 23rd of February. Fuck three years ago. After three weeks we are living in another dimension. For me, it is difficult to think about my life before this. No one can be prepared for this. Impossible.”
Normalcy bias is a significant driving force of this pandemic. Most people don’t want to believe that everything is going to change because it is too painful and scary to contemplate. Discovering the truth is also difficult and time-consuming, especially in a highly polarized and degraded media environment like ours. The average person’s beliefs about global issues typically don’t matter much. Except right now, there are few things more important to our collective future.
Today, we are rightly focused on improving people’s awareness of personal hygiene practices. We should bring the same focus to improving our collective epistemic hygiene — the processes we use to develop accurate beliefs and discard inaccurate beliefs, both personally and more broadly in society.
As we grapple with the consequences of widespread disbelief that continue even in a declared pandemic, I want to take a bit of time to examine how one person saw the possibility of COVID-19 going global early on.
Balaji S. Srinivasan was one of the first, clearest, and loudest voices on the Internet in North America to sound the alarm on COVID-19. Balaji is an angel investor, former CTO of Coinbase, and former General Partner at Andreesen Horowitz. He also founded a molecular diagnostics company and taught bioinformatics at Stanford where he holds a Ph.D. in Electrical Engineering.
Balaji started writing and tweeting about the virus back in late January when it was still mostly confined to Wuhan. He seems to have written and tweeted about little else since then. At that time, he kicked off a lengthy tweet thread this way:
It’s hard to imagine from the vantage point of mid-March, but Balaji’s speculations were initially received in some corners of digital media as irresponsible alarmism shaded with anti-Asian xenophobia. Many users on social media ridiculed what they saw as the overreach of cocky “tech bros” and VCs who thought they knew it all because they’d made some money in software. Now those speculations have become our shared unfolding reality.
In a recent conversation with Reason’s Nick Gillespie, Balaji briefly outlined the process by which he came to believe that the coronavirus would become a history-altering pandemic. I want to share Balaji’s process here with a bit of commentary because I think it is a clear account of how to identify weak signals of change in the informational environment and how to begin substantiating and extrapolating trends.
Gillespie’s first question to Srinivasan during their conversation — which you can watch in its entirety here — was about how he came to learn about the coronavirus and why he was early in sounding the alarm.
Here is his full transcribed response, lightly edited for clarity:
“I had been following the coronavirus out of my peripheral vision as one of many stories until I saw the lockdown on January 23rd. That was unprecedented. It was an unthinkable signal. The Chinese government is completely focused on economic development, so to quarantine millions of people — which would soon become hundreds of millions of people — that made me pay attention to what was going on.
And so I began talking to many friends of mine — Chinese engineers, entrepreneurs, angels, and the like — and the qualitative picture that I got from this was hair-raising. We’re talking videos and photos of people being blockaded in their apartments and dying on the street. Even one of a man throwing himself off a bridge because he couldn’t get to the hospital and he didn’t want to bring the virus home to his family.
I didn’t know what to believe from this. Some of it could be fake. But on the other hand, those images seemed fairly difficult to fake even with today’s technology and they seemed to match the general circumstances.
Even if some of them were true, these were sci-fi movie-level events happening in China. If I was too skeptical, I would remind myself that there was an officially acknowledged scale of the lockdown which was unprecedented and that was an independent reference point of ground truth that one could anchor on. I also have a background in bioinformatics, molecular diagnostics, genomics, and so on.
The third thing I did is I dug into the literature and read all these dispassionate medical reports. These kinds of articles in the New England Journal of Medicine and the Lancet are written in a style which is almost like the opposite of clickbait.
The first U.S. case study written up in NEJM said something like ‘thirty-five-year-old patient with no pre-existing conditions contracted bilateral pneumonia and by day ten we pursued compassionate use’. In English, that means ‘a healthy 35-year-old man was brought to death’s door and we needed an experimental drug to save his life’. I mean zero insult whatsoever toward folks writing these articles. That’s an accepted style. But when you translate them it’s eyebrow-raising. Here’s a healthy 35-year-old about to die and it took ten days and a team and an experimental drug to save him. And that was the first case in the U.S. That’s a contagious disease, so you start mapping that to not just one person but to ten people or a hundred people or a thousand people and that starts to get very expensive in terms of suffering and in terms of medical resources and so on.
So the combination of (A) China’s official state action, (B) unofficial social media reports from people in China and (C)articles in biomedical journals — that’s what made me triangulate on this early as something of potentially critical importance.”
Let’s examine how Balaji thought about (A), (B), and (C) so that we can extract some practices and principles that everyone can adopt to improve their epistemic hygiene and, in turn, their ability to develop more accurate forecasts.
(A) China’s official state action
Balaji uses a ‘vision’ metaphor to explain the field of awareness in which he seeks signals of change. Practically speaking, this means consuming information on a wide range of topics across media spectra and not allowing ourselves to become too blinkered by our current area of focus. Simple enough. But consuming a broad diet of information is not enough. We also need to learn how to notice things and what’s worth noticing.
When news broke that China was suspending all highway traffic and public transport in three large cities in Hubei province it struck Balaji as immensely odd because the action ran against his strong prior conviction that the Chinese government was laser-focused on economic growth. Wuhan is the largest city in central China and the commercial, cultural, and transportation hub for the region. Shutting the city down could not be done without serious economic consequences for China.
A prior is a concept from probability theory that describes the beliefs we hold before receiving new information. Priors can be strong, weak, or something in between. Typically, the more evidence we have for our priors, the stronger they are.
If we are striving for accurate beliefs — which is the point of epistemic hygiene — the quality of the evidence we gather matters a great deal. It is possible to have strong priors based on very poor quality evidence. Someone who has earnestly watched thousands of hours of Flat Earth videos on YouTube has likely developed very strong priors that the Earth is flat; so strong that satellite images from space showing the Earth’s curvature do not put an appreciable dent in that prior.
Our priors are critically important when it comes to noticing change because to be surprised by new information we first need a good sense of what we believe and why we believe it. Priors help us to smartly process new evidence, so we ought to consume information not just to stay informed but to form strong beliefs and opinions that calibrate our ability to sense when things are changing.
(B) Unofficial social media reports from people in China
The next thing that Balaji does is explore his real-world and virtual social networks to get a sense of the reality unfolding on the ground in Hubei province.
I would venture that — from an epistemological perspective — Balaji is seeking what we might call informational contrast. Moving from state bulletins to social media posts allows him to examine the situation from two vantage points that are radically distinct from one another in tone and content.
The sources for the initial reportage on the shutdowns appear to be Chinese state media outlets and statements made by Chinese government officials. While in this case — given how unfavorable this news would be for China’s perception abroad — there may have been little reason to doubt the veracity of these sources, they still likely presented a sanitized view of the situation that leaned heavily on bare statement of fact and the language of objectivity.
Going to his social media feeds and whisper networks for information derived from experience, testimony, and good old-fashioned gossip would give him a better sense of the lives of the Wuhanese dealing with the reality or threat of COVID-19 infection during a city-wide quarantine.
These media sources and testimonies also furnished Balaji with a feel for the affective features (“the qualitative picture that I got from this was hair-raising”) of the facts he was learning. In strategic foresight, which I practice professionally, we employ storytelling and design techniques to create an imaginary encounter with the future which engages the affective components of our belief formation processes. Sometimes — as in the case of an emerging global pandemic — press reports, charts, figures, and graphs do not drive home the potential real-world impact of change. For that, we need stories, like the story of the man on the bridge.
We also should be mindful of how stories can lead us astray. Balaji understands that these types of networks often contain high-value information hidden among baseless speculations, insinuations, and outright falsehoods, and so he proceeds with a measure of skepticism. At the same time, his skepticism is tempered by the mind-bending nature of what he has already learned.
This tells us something important about the interplay of our beliefs, especially as it occurs across informational types. When immediate comprehension fails, we should weigh the pieces of evidence we’ve gathered against one another. We do our best to see the whole in light of its parts and vice versa, cycling through “big picture” and “street-level” information in an interpretive cycle.
(C) Articles in biomedical journals
Once Balaji felt he had a grasp of the official line and the reality on the ground in Wuhan, he turned to peer-reviewed medical journals to better understand the nature of the disease itself.
Generally, peer-reviewed papers published in prestigious journals are an oasis of high-quality information. Reading journal articles and other technical sources of information like patent filings and policy papers allow us to step out of the world of clickbait journalism and knee-jerk social media posts to further round out our knowledge base with dispassionate research which has been pre-vetted by experts in the author’s field.
Peer-reviewed journals are by far the most overlooked source of information in our epistemic practices. There are at least two reasons for this.
First, many of them are sitting behind paywalls that are prohibitively high for most people. But there’s no such thing as a wall that can’t be scaled.
Second, the language of peer-reviewed journals is often prohibitively technical. Abstracts can help to demystify the details but sometimes even these summaries elude our understanding. Talking to experts in the field in question can help. When all else fails, try to render it in plain language for yourself as Balaji does in his response. In the retelling of his epistemic practice, it’s this plain language rendering that gives him the final pieces of the puzzle which lead him to the belief that a pandemic was likely to follow.
Because journal articles and other pieces of technical writing are written by professionals for professionals, the insights they contain often do not circulate outside of a niche profession or academic sub-discipline. Informed generalists with some ability to translate technical language into digestible takeaways for the laity provide a valuable public service in connecting the dots between mainstream media ecology and walled-gardens of institutional research.
Note that even with peer-reviewed journals, it is important to read widely and critically. Since so few people read studies in peer-reviewed journals, the mere existence of such a study can be used as an intellectual cudgel to silence dissenters and to create a veneer of respectability. So we should keep our critical hackles up even when we’re consuming peer-reviewed information. If we do, there is enormous potential for the information hidden behind paywalled peer-reviewed journals and in other sources of technical writing to drastically improve our beliefs about the world.
If you’re like me, you might be wondering what it is that people who are going about their lives as if nothing is happening are seeing and hearing when they watch the news or browse social media. The only thing that could make the reality starker is for it to arrive — as it inevitably will and as it already has — on our doorstep.
Many of us are failing to take the necessary measures to contain COVID-19 on a personal level because the situation as it exists elsewhere simply defies belief. It exists as an abstraction mediated through phones, computers, and TVs. We don’t feel it as an imminent danger to ourselves and our loved ones.
The inertial tendencies of our minds are safeguarded by a phalanx of defenses: denial, deflection, ridicule, vitriol, self-deception, conspiratorial thinking, and so on. We fight against changing our minds because it often comes with real-world costs. Publicly admitting wrongs can be status-lowering; this is why politicians so rarely do it. It can damage the trust others invest in us. It can mean admitting to ourselves that we’ve made bad investments — with our time, money, passion, and so on — or that we might not be as smart as we think.
Having accurate beliefs about the future is no easy feat. The human faculty of foresight is not evenly distributed, but it seems at least somewhat trainable. Practicing a bit of epistemic hygiene can help.