Everything Was Earlier Than Expected
why we can be confident covid has been around for a long time
Humans are probability-blind.
We're like colour-blind people looking at a garden, or tone-deaf people listening to an orchestra. We have indirect evidence of the existence of probability, just as a colour-blind person has indirect evidence of the existence of hues, but we can't see it ourselves and have to figure out proxies that help us navigate a world where it's really important.
Mostly, we don't do a very good job of it, and that's not entirely our fault: it's a really hard problem.
Our probability blindness shows up in various guises. We over-estimate the significance of coincidences, for example: a one-in-a-million chance happens to someone eight thousand times a day, because there are eight billion people in the world. Every one of those people is apt to think, "The odds against that are a million-to-one! How could it possibly have happened?" when the odds of it not happening (to someone) ninety standard deviations away from the mean... when five standard deviations is considered close enough to certainty that we can safely ignore the difference in most cases.
Another more insidious kind of probability blindness is our tendency to date stuff from the earliest known instance of a thing, which when you put on ProbabilityVision goggles (if there were such things) looks... sort of dumb, even thought it’s mostly done by really smart people, with PhDs and everything. But drawing conclusions from data without explicitly considering the probability distributions that lie between the data and the conclusion is going to mostly result in "your conclusions are bad and you should feel bad", as Zoidberg might say:
This is not an insight that is original with me. Some poor guy wrote about it in Nature in the '80s with regard to fossils, but the problem with probability blindness is that it's very easy to forget what we can't see. It's work to keep it in mind all the time, particularly when it isn't widely understood.
But it's a consequence of this that almost everything we know about timelines is based on incorrect inferences, so almost every investigation into anything in the past is going to result in the "surprising" discovery that whatever phenomenon we are looking at has its origin "far earlier than we thought".
I'm writing a long piece on this phenomenon for a speculative project on things that people always find surprising no matter how often they are "discovered" and reported on, like "engineers use nature for inspiration" (regularly surprising people since the 1800s) and "spiritual beliefs and fascism are natural fits for each other" (regularly surprising people since the 1920s) and "Canada's Indian Residential Schools killed a lot of children, many of whom are buried in unmarked graves" (regularly surprising settlers since the 1930s). All these things and more have been reported again and again and again in the popular press, and each time they are reported as if no one has ever heard of them previously and they have just now been brought to light.
This is a curious phenomenon to me, as it speaks to a systemic, society-wide, refusal to update certain beliefs in the face of new evidence, while other stories seem to sweep across the globe with viral vitality, causing almost everyone they touch to update what the believe. One of those viral ideas is Richard Dawkins’ notion of "memes", which despite being almost incoherent does seem to capture something real about ideas that spread and stick in people's minds as opposed to ideas that bounce off and die.
Understanding things is very often the study of differences, and I can't help but think that the study of "anti-viral" ideas should be given as much attention as "viral" ideas if we want to understand what's going on. In the world we live in, "going viral" has the potential to make people money, so it gets studied, while anti-virality, which may hold the key to understanding, does not, except by people like me, who--unsurprisingly--can't get much traction for our ideas.
My own view is that the role played by narratives is key to the difference between ideas that are accepted and rejected: acceptable ideas tell good, positive, coherent, confirmation-bias-feeding, stories. Unacceptable ideas tell ugly, dissonant, uncomfortable stories. Which is a story for another time.
In the process of researching for the aforementioned project, I did some simple searches for things that were found to have happened "far earlier than thought" and easily dug up almost two dozen examples, from when humans populated the Americas to when dinosaurs got legs (I might be misremembering that one.)
Recent work on the Black Death is another lovely example of this phenomenon: it turns out that the earliest instance of a major epidemic caused by the bacterium Y. pestis was not in Europe in the 1340s but in Baghdad in the 1250s, almost a century earlier.
Using a combination of genetic analysis and very modern historical research, depending on scholars in multiple languages and disciplines across a huge span of time and space, historian Monica Green has been able to trace the origin of the plague bacterium to marmots in the Tian Shan mountains in southern Mongolia, and textual evidence strongly suggests it was transported to the Middle East via grain caravans, although there are no doubt other possibilities as well. How it got to Europe a century later--and why it took a century--is still an open question.
It might also be interesting to see if there were major outbreaks in the Far East... much earlier than previously thought. The historical record is full of gaps, and so the information that survives today makes for a very sparse sampling of what actually happened. That implies probability distributions with long tails that vanish into the darkness, and if we ignore them we end up being surprised, again and again and again.
Evolution is an elaborative process: each generation rings changes on what is present in the previous generation. This rarely produces dramatic positive changes between generations: significant mutations are almost always deleterious because they happen in the context of the whole organism, which is the unit of selection. Despite enough purely Mendelian cases for them to provide insight into the process of inheritance, most capabilities an organism has are the result of many, many genes working together, but even if there was a single change that could make a person twice as strong, say, it would still be bad for them unless their bones were twice as strong as well. Being able to jump high enough to break your own legs is not a feature.
This means that when a biological phenomenon first makes itself manifest it is almost always long after it has first existed in some recognizable form once we go looking for it, and most of the changes that were needed for something to become a huge big deal comfortably predate its discovery. No one was looking at sick marmots in southern Mongolia in the 1100s, but I bet if we happen to find one buried and preserved in a mudslide, or the bones of one in a cave seep, we'll find earlier versions of Y. pestis. Maybe even fairly pathogenic ones, which just hadn't yet stumbled upon the right circumstances to shine.
The same is probably true of our current little plague. Work in the past year has shown that covid is capable of infecting mink, cats, dogs, zoo animals, otters, and now deer in Quebec.
The odds that a virus that can infect a wide range of mammals just happened to start doing so immediately after it started infecting humans are staggeringly low.
We don’t do a lot of work on zoonic diseases, and viruses are particularly tricky to discover if they aren’t especially symptomatic. Bacteria in a dead animal can probably be cultured, but finding a virus is much harder work if you don’t already know what you’re looking for.
Because of this, and limited funding for research, sampling for coronaviruses in wild animal populations is low. There was work being done on bats in China, where the virus was known, but in deer? Otters? Mink? Tigers? Not so much. And we know that sparsely sampled data is almost certain to be hiding instances of things “far earlier than previously thought”. The laws of probability tell us this.
So we can predict a spate of stories in the next few years detailing how researchers have made the "surprising" discovery that covid existed—probably in animals other than bats—long before 2019. There are bound to be tissue samples from mammals in southern China taken years ago that are still preserved somewhere that could be analyzed for this.
I sometimes think I should start an academic journal that traffics entirely in papers on the discovey of things that existed “much earlier than previously thought” across all fields.
I could call it, "The Journal of Predictable Surprise".
If you like this kind of thing, please subscribe!
If you know other people who’d enjoy this article, please share!
This is brilliant. I wish I had students good enough to understand it. Not that I don't have some really good students, but the meaning is so subtle and based in a whole school of learning about the world. I'll say it again...this is brilliant.