There’s been increasing talk the last few days about “black swan” events. Like many folks, I was at least semi-familiar with the term when I first heard it.
I went looking for its origin and found Nassim Taleb’s The Black Swan, a fascinating book that placed the phrase in modern parlance. It offers important lessons for understanding our current situation, how we got here, and what we can do next.
Taleb defines a “black swan” as a massively impactful outlier event beyond normal expectations because nothing in the past could have reliably pointed to its likelihood. You heard an echo of this sentiment when President Trump said that no one could have predicted COVID-19.
Yet despite this, Taleb points out that human beings only have a tendency to create explanations for black swan events after they occur, though there may have been prior warning signs. By doing so they fail to learn from them or expect different black swan events in the future.
This failure involves three delusions called the Triplet of Opacity. They are:
- The illusion of understanding. Everyone thinks they know what is going on, even though the world is far more complex and random than they realize. (I am often quite susceptible to this fatal flaw.)
- The retrospective distortion. Events often appear more clear and obvious after the fact, leading to recriminations about what everyone should have done differently – basically, “hindsight is 20/20.”
- The over-valuation of expertise. Experts often rely on categorization and selected facts to react to or diagnose black swans during and after their occurrence. But the smoothing and flattening of inherently complex data removes needed nuance.
The last delusion, which speaks of the dangers of what Taleb calls “Platonization” – the desire to fit everything into tidy boxes and narratives – is especially significant. It speaks to the paradox of increased information we often experience in the modern world.
As we gather more data, it’s more likely that we’ll simplify and create predictions. And these models will inevitably end up being some degree of wrong, because we tend to interpret random noise as valuable insight. Dr. Anthony Fauci pointed to this issue yesterday when he said that, frankly, our best predictions about COVID-19 right now turn on models based on assumptions that may be incorrect.
So where is the Triplet of Opacity most at work right now? A few key points.
The current coronavirus outbreak is a black swan. This means it is an out-of-nowhere, unprecedented event. But the way the pandemic progresses is rather ordered. It will follow the much-publicized curve.
This is actually a good thing. Because the outbreak will unfold in a predictable way, it is possible to model it. The steepness or shallowness of the curve in a given locality will determine the stress on the local healthcare system. But that steepness or shallowness is itself another black swan. Depending on the steps we take individually and socially, the curve will steepen or flatten. But determining the degree of steepening or flattening is hugely complicated.
Another lesson that I learned: do not assume that any data is perfectly accurate. Instead assume some data is more reliable than others due to fewer factors influencing its gathering and consistency.
For instance, COVID-19 death counts are off by some factor because attributing cause of death is often difficult. But death counts are inherently more reliable to calculate viral severity and spread right now than active case numbers, which depend on testing (which still isn’t widespread) and don’t account for an unknown number of asymptomatic COVID-19 cases. And per-capita numbers are yet more helpful for determining severity, because they are tied to a relatively known number: population. My current project charting the pandemic’s growth on a per-state basis uses per-capita numbers.
We also cannot assume any model is perfectly accurate. Again, though, we can say some models are better than other for estimating the breadth of the COVID-19 outbreak. Models using more reliable data, like death rates or per-capita numbers, will likely be better founded.
We must beware theorizing or forward-facing generalization based on past data points. When you hear someone ask whether the United States will become the next Italy or the next South Korea, you should inwardly cringe. This is a false dilemma. It assumes the United States is bound to one path or the other, and not an entirely different one based on a multiplicity of unique factors, some known (like geography and population density) and some unknown.
Everyone can fall prey to “information smoothing.” I’m not immune. None of us are. Media of all kinds especially thrives on this habit, whether social or mainstream. It often manifests as sloganeering (“It’s just a bad flu”) or myopia (“We just need to test more people”). It takes the form of the fallacy of induction: data point, data point, data point, thus inescapable conclusion.
Enslavement to trend lines based on cherry-picked details is narrative creation. You and I should both remember that this may be tempting, but it’s ultimately unhelpful. As Taleb says, “favor experimentation over storytelling, experience over history, and clinical knowledge over theories.”
Finally, averages are dangerous in situations with extreme possibilities like a black swan event. This is why medical experts are rightly providing exceptionally wide ranges of future active case numbers or death tolls, and why the media sadly reduce these ranges to singular numbers. Averages must be viewed in a broader context that tells us what went into their calculation.
So, we can learn all that just by applying the Triplet of Opacity to the coronavirus outbreak. But how do we solve the Triplet? Can we overcome its delusions?
Yes. Taleb suggests looking for situations and favoring strategies where estimable favorable consequences are much greater than estimable negative ones.
We should seize any opportunity, or anything that looks like an opportunity. Such serendipitous events can themselves become positive black swans. One example: the somewhat miraculous possibility that a common anti-malarial drug, hydroxychloroquine, may be an effective treatment for the virus. But we cannot stop there. We must try every other possible treatment as rapidly as possible, from convalescent serum to remdesivir.
We must embrace simple solutions with large upside. Nearly every American can wash their hands and remain six feet from others. These practices dramatically cut down the likelihood that you will contract the virus, breaking the chain of transmission.
Additionally, look for blaring trends in data. What’s showing up, over and over, in every model or even in the raw numbers? For instance, COVID-19 causes severe illness in people over 70 or with underlying health conditions at a far higher rate. It also spreads like wildfire in areas with high population density – not just packed cities, but environments that look like them. Think large crowds.
We should also avoid embracing only the most extreme, potentially harmful predictions. One of the earliest models for COVID-19, emerging from Imperial College, forecasted apocalyptic numbers of death and contagion even with social distancing measures. And the new UW Health study predicting widespread ICU bed shortages and high death tolls in a few weeks follows a similar path. I say this not to imply these studies are entirely wrong. But a myopic focus on worst-case scenarios strips everyone of hope and leads to dramatic, unnecessary overreaction in the face of unpredictability.
And last, but certainly not least, if a blaring trend becomes clear or new opportunities present themselves, we must be ready and willing to change our approach. Any solution may work against an inherently chaotic and ever-shifting black swan.
I hope my point is simple and hopeful. Right now we live in strange, unsettled times. And to emerge from them we must fight the tendencies which further obscure our way. But The Black Swan demonstrates how we can embrace approaches that will slowly but surely lead us out of the fog and into a better, more prepared future.