### The Volokh Conspiracy

Mostly law professors | Sometimes contrarian | Often libertarian | Always independent

# Was the New Coronavirus Accidentally Released from a Wuhan Lab?

## It's probable.

|The Volokh Conspiracy |

The Washington Post Fact Checker provides a thorough discussion of the relevant facts concerning whether the COVID-19 pandemic is a result of a lab accident in Wuhan. By the time I reached the conclusion, I was convinced that it almost certainly is. And then I reached the article's conclusion: "The balance of the scientific evidence strongly supports the conclusion that the new coronavirus emerged from nature — be it the Wuhan market or somewhere else. Too many unexpected coincidences would have had to take place for it to have escaped from a lab."

This conclusion is puzzling. The video accompanying the article acknowledges that it is a coincidence that the disease emerged in the same city where coronavirus bats are studied. But the video and article point to a much bigger coincidence: The Wuhan Center for Disease Control & Prevention, where researchers engaged in bat research without using appropriate protective equipment, was "right around the corner" from the Wuhan Seafood Market, the location tied to multiple initial carriers of the virus–apparently, fewer than 300 meters.

Let's do some simple math. First, let's try to make concrete the point of Michel Trottier-McDonald that we've long expected the possibility of a pandemic, and so we should have a strong prior that a pandemic that actually occurs is a result of something other than an accident. Let's assume that labs studying bats are actually very safe, despite the safety violations documented by the Fact Checker, more specifically that the probability that a lab somewhere in the world causes a bat coronavirus pandemic in a given year is only 1 in 100,000. And let's assume that non-laboratory zoonosis was ex ante much more likely, with perhaps a 1/1,000 chance of producing a pandemic in a given year. So, with no other evidence, when a pandemic occurs, we can calculate P(Accident | Pandemic) = P(Accident) / (P(Accident)  + P(Zoonosis)) = 0.0099. Moreover, let's suppose that there are 20 labs worldwide (just a wild guess, my guess is it's on the high side) that study coronaviruses and bats. So, the probability that any one lab is responsible, which we can denote P(WCDCP) when speaking of the Wuhan Center for Disease Control & Prevention, is approximately 0.0005. From this vantage point, the idea that the virus escaped from that lab does look like a conspiracy theory.

But we then need to account for the coincidence that the first location associated with the outbreak was in a 300 meter radius. As Some Guy comments in response to Trottier-McDonald, surely this should figure in the calculus. We can calculate the probability that this neighborhood would be the first location associated with the outbreak as follows: P(Neighborhood) = P(Neighborhood | WCDCP) * P(WCDCP) + P(Neighborhood | ~WCDCP) * P(~WCDCP). That is, the unconditional probability that a bat coronavirus pandemic would first be associated with this particular neighborhood is equal to the probability that this would occur assuming that the coronavirus originated in that lab, multiplied by the probability that the coronavirus originated in that lab, plus the probability that the pandemic would first be associated with this particular neighborhood assuming that the coronavirus did not originate in that lab, multiplied by the unconditional probability that the coronavirus did not originate in that lab. Let's suppose that there is 50% chance that an accident at the WCDCP would lead to the pandemic first being observed in the neighborhood. So, P(Neighborhood) = 0.5 * 0.0005 + P(Neighborhood | ~WCDCP) * 0.9995.

What about P(Neighborhood | ~WCDCP)? As a first approximation, let's assume that the probability of a pandemic starting in any location is roughly proportional to population density. Wikipedia reports population density of as much as 20,445 per square kilometer in the approximate area of WCDCP, so the circle with radius 0.3 km would have a population of about pi * 0.3 * 0.3 * 20,445 = 5,781. Because the world population is 7.8 billion, P(Neighborhood | ~WCDCP) = 5,781/7,800,000,000 = 7.41 X 10-7. But let's suppose, to err against the conspiracy theory side, that, lab aside, this neighborhood was much more likely to be the source of an outbreak than other random neighborhoods on earth with equal population density (even though the relevant bat species' habitat is far from Wuhan)—say, 100 times, because it had a wet market that sold wild animals (though apparently not bats). Then P(Neighborhood | ~WCDCP) = 7.41 X 10-5. So, we can calculate P(Neighborhood) = 0.5 * 0.0005 + 7.41 X 10-5 * 0.9995 = 0.00032.

Now, we can apply Bayes' Theorem. We want to calculate P(WCDCP | Neighborhood)—the probability that an accident at the WCDCP was the source of the pandemic given the evidence that the first cases were detected there. This will equal P(Neighborhood | WCDCP) * P(WCDCP) / P(Neighborhood), so 0.5 * 0.0005 / 0.00032 = 0.781. That is, with these conservative assumptions and this simple model, there is a 78.1% chance that the virus originated in the Wuhan lab.

This is a back-of-the-envelope calculation and hardly definitive. Ideally, one would like to take into account other evidence. For example, Chinese scientists report that they had no record of this coronavirus and that the most similar coronavirus they studied was only 96% similar. That doesn't seem all that relevant–couldn't an infection have occurred without the virus being detected and its genome decoded?–but it might move the needle a little bit. Other evidence, such as intelligence reports, might change our conclusions as well. But at the very least, it suggests that this is no mere conspiracy theory and that the Fact Checker ought to document its numerical assumptions. The Fact Checker understandably prefers hard facts to ruminations about coincidences. But if the Fact Checker is going to make a probabilistic assessment, it needs to think about coincidences, and it should focus on the most glaring coincidence (that the WCDCP was in the same neighborhood) rather than on the less remarkable one (that the Wuhan Institute of Virology, which apparently had much safer practices, was in the same city).

Of course, conspiracy theorists often thrive on irrelevant coincidences. What conspiracy theorists often forget is that there are so many possible coincidences that the occurrence of a few coincidences is not a coincidence. If you told me that the director of the lab and one of the victims had the same last name or enjoyed eating in the same restaurant, I wouldn't put much weight on that. But it at least seems to me that the location where the disease originated is not just one possible coincidence of many possible coincidences, but rather one of the first facts one would want to know in assessing the pandemic's origin.

Does all this matter? People are not very good Bayesians, but we should strive to think rigorously about probability in matters of public policy. I don't think that the likelihood of origination in a lab means that China needs to pay reparations. But if this pandemic did originate in a lab, that should lead the international community to pay close attention to how such labs operate and how they could be monitored. And while it seems to me highly unlikely that the virus release was intentional, origination in a lab might nonetheless reasonably focus attention on bioterrorism. Modern genetic technology certainly makes it more likely that scientists, intentionally or by accident, could create stronger strains of viruses. Now that we have seen what a naturally occurring virus can do, the possibility of a man-made virus worse than the novel coronavirus should terrify us, perhaps even posing an existential risk on the same order of magnitude as climate change.

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

1. Thank you. I said the same thing here, much more dumbed down, over a month ago.

1. You were wrong a month ago. You are wrong now.

1. [Citation needed]

2. This article shows its work. You merely assert a negative, with no alternative, no reasoning, and no work.

1. No he doesn’t.

1. Well,that’s empirical.

1. “Having an argument isn’t just saying ‘no it isn’t.'”
“Yes it is.”
“No it isn’t!”

2. Throw in a country so corrupt that lab animals are sold for food — and the government having convicted a couple of people for having done exactly that…..

3. However, Baysian Theory does allow for guess and gut instinct, it can also be notorious sensitive to those initial inputs. Garbage in, you get garbage out. So the first couple of numbers, the odds of producing a pandemic, 1/100000 and 1/1000 pretty much guesses.
So you mention the because the virus outbreak started near the lab, it has to up the chances that it came from the lab. However, you also mention that an open market, that sells bat products, is also in the same area. So the odds that it came from the market is also markedly higher. All you’ve really said was that outbreaks are more likely near a market or lab.

Bottom line, supposition on top of supposition gives you nothing. And worse, just because the most likely outcome of a pair of dice is 7, it has no bearing on what you actually rolled.

So Yeah, you did the whole conspiracy theory thing.

1. “…because it had a wet market that sold wild animals (though apparently not bats).”

So. No bats.

2. Yeah, I don’t need to buy into this 78.1% chance computation to note that this is a striking coincidence, enough so that it merits consideration of whether it is likely to truly be a coincidence. That was my comment, and I appreciate the Prof’s taking a stab at the question. The critics in this thread will just poke holes in his model, which is always easy to do, without ever offering their own answer.

Since I made that comment the US intelligence community came out and said that the virus probably came from the Wuhan lab.

If you want to discount “the US intelligence community says so” from holding any evidentiary value, I’m OK with that. But not if you’re one of those people who have been taking it as gospel that Russia “interfered in our elections.”

1. the US intelligence community came out and said that the virus probably came from the Wuhan lab.

Cite? And not Pompeo talking out his ass.

2. I would only be terrified if they were doing human testing in the lab. Otherwise, it more likely nature was doing its own human testing.

3. The question that arises is why the U.S. media and some others appear so eager to carry water for the Chinese Communist Party.

1. Maybe to prevent irresponsible saber rattling that would get even more people killed down the line in a war?

1. Nope. We can rule that out conclusively by looking at how they covered the Russia conspiracy theory hoax.

1. Well, it was neither of those, and I don’t feel like litigating that whole thing, here. But that’s a fair point. There is a lot of overly heated anti-Russian rhetoric.

2. Please. No one is going to go to war with China.

Cutting off trade…that’s a different matter.

1. Cutting off trade…that’s a different matter.

Newsflash: Then there will surely be war with China.

Regardless of the origin of the virus, China lied and obfuscated relative to the nature of the virus. China closed off internal travel for Wuhan residents, but allowed thousands of Wuhan residents to travel all around the world. Those are established facts.

These purposeful acts by China have lead to the deaths of many thousands of Americans. That is nothing that can just be dismissed. There will be a response. I am quite certain it will be comprehensive: political, economic, diplomatic, and military. And long lasting.

1. There will be a well deserved response. Not a shooting war.

1. Japan considered such to be grounds for war back in 1941….

1. China’s not that dumb to get into a shooting war with the US at sea.

2. Japan needed embargoed resources. The US oil embargo was the match tossed into the powder keg. Japan used war to seize new resources.

China needs markets for its products. War for China will not open new markets, but close existing ones.

2. Cutting off trade…that’s a different matter.

We’re not going to look very good without a nose.

Trump’s obsession with China trade is destructive, and is yet one more example of what an ignorant incompetent jackass he is.

1. China…. Let’s review…

1. They release a very dangerous virus, through negligence, either from a lab or through unsafe food practices.
2. They then deliberately lie about the person-person transmission for weeks, ensuring it spreads further.
3. Then, critical medical supplies produced and owned by American companies in China are banned from leaving the country.

No, our trade with China needs to be revisited for our own public safety. Having critical medical supplies and drugs produced in China is UNACCEPTABLE in the long run, given their current practices. Unless you want millions of Americans to die during the next epidemic China releases, lies about, then bans the needed supplies from leaving China.

Trump realized having so much manufacturing in China was extremely dangerous for our country as a whole. It NEEDS to change, and critical manufacturing NEEDS to come back to the US. China has a history of using its dominance in certain manufacturing markets as a political tool to be cut off when convenient. It can’t be trusted.

1. “No, our trade with China needs to be revisited for our own public safety. Having critical medical supplies and drugs produced in China is UNACCEPTABLE in the long run, given their current practices. Unless you want millions of Americans to die during the next epidemic China releases, lies about, then bans the needed supplies from leaving China.”

You don’t have to revisit trade with China to ensure that critical medical supplies owned by US governments is not produced in China. You just… buy them from other suppliers. Or make them domestically. What “revisit[ing]” “trade with China” are you even proposing to save us from China hoarding things manufactured in China? How would tariffs get critical medical supplies to the US more quickly?

I think you’re confusing government guaranteeing domestic supply (say with oil reserves) with trade policy prohibiting or regulating private people from trading with other private people, in other countries.

1. “You just… buy them from other suppliers. Or make them domestically”

Unfortunately, it’s not quite that simple. The major issue here is the overwhelming majority of worldwide production for several critical medical supplies are made in China. For the current crisis, that’s what we have, and we still need to import. We can’t build up new factories that fast. For future crises however…

The major issue long term issue here is Chinese manufacturers producing an item and undercutting US (and other manufacturers) on price. This drives the US manufacturers out of business, which increases the Chinese market share. Until they have dominance in a market, then once a crisis hits, suddenly the production…doesn’t get to the US, as the Chinese government uses their physical control of the means production as a proverbial hammer.

The way around this is guaranteeing US domestic supply (as you mention). But in a free market, that requires support from the government, either with internal pricing support, or tariffs on Chinese goods, or a combination of the two.

This argument extends beyond N95 masks. The Chinese, for instance, have a history of using their current dominance over rare earth production as a political lever. Irritate them on a separate issue, and suddenly, the needed supply of neodymium for your manufacturing is held up in “delays” for months.

It, frankly speaking, can’t keep going on like this. Chinese production cannot be trusted because the Chinese government can, will, and has cut it off when politically advantageous. US corporations want into the large Chinese market, but it’s not worth the long term risk. And the markets are simply too connected. I’d advocate for across the board, long lasting, 50% to 100% tariffs on all Chinese produced goods. That will ensure protection for US and allied country’s industry to compete, and will ensure that if the Chinese cut us off again, the US remains safe.

1. “Unfortunately, it’s not quite that simple.”

Don’t get this mixed up. It’s incredibly complicated. That’s why I reject simple solutions like generic “revisit[ing]” trade with China.

“We can’t build up new factories that fast. For future crises however…”

So what do you propose be done to force China to hand over its valuable medical supplies quickly?

“The major issue long term issue here is Chinese manufacturers producing an item and undercutting US (and other manufacturers) on price.”

This is a wonderful result. It means the Chinese government is producing goods cheaper than market prices and distributing them to American consumers. We should hope for more of this. It effects a tax on Chinese citizens and manufacturers for the benefit of American consumers. If the concern is that we need to ensure domestic production, the answer isn’t a trade war. The more optimal solution is for the government to buy up the goods or produce them, and wait to sell them until the Chinese jack up the price after the price-fixing has served its ostensible purpose (to kill American industry). Alternatively, we could just use the existing trade agreement to deal with the issue, or prop up the domestic industry with direct investment. The dumbest thing we could do is tax American consumers and had the money to domestic industry. If the hypothetical industry is critical to American infrastructure, all Americans should help pay for it, not just the people who consume or produce those goods.

“But in a free market, that requires support from the government, either with internal pricing support…”

There are about a million other ways to deal with it, too.

“Irritate them on a separate issue, and suddenly, the needed supply of neodymium for your manufacturing is held up in “delays” for months.”

Just so I understand, the Chinese will refuse to sell neodymium to American phone battery manufacturers because the US government insists on Taiwan being recognized? Why is that the problem of the American public? Those American phone battery manufacturers intentionally chose to secure supply chains through a mercurial autocracy. Heaven forbid that people who make dumb economic decisions go out of business. We used to call that capitalism, but whatever.

“That will ensure … the US remains safe.”

WTF are you talking about? The US is safe right now. A 100% tariff on Chinese goods is not going to cause N95 masks to fall from the heavens. It will have the exact opposite effect. The best we can hope for is a long, brutal, economically devastating trade war, resulting in a trade agreement that looks virtually identical to the one we have now. You’re a keyboard trade warrior who thinks if we just burn our own ports enough, the Chinese will eventually cave and give us whatever we want. It’s a fucking pipe dream. They aren’t going to bend to our will for the precise reasons you want to get in the trade war in the first place. And it is a true fact that China is able to manufacture a lot of things more cheaply than us, and that will be the case forever. Any attempt to bend China will fail, and will cost us immeasurably.

2. Why do commentors here appear so eager to carry water for racists?

1. Criticizing a government isn’t racist.

1. That’s true. But people have a hard time separating that.

1. The Chinese Communist Party and Chinese government isn’t all Chinese people, anymore than the evil Nazi Germany is all Germans, or a corrupt Nigerian government is all Nigerians.

Not sure why you think people have a hard time separating this.

China (In the form of the current Chinese Government) has a number of major issues that make our continued relationship problematic, and it may require severe recalculation. They may be responsible for far more American deaths than Al Queda. Perhaps not deliberately, but likely due to negligent actions. Severe recalculation of our current trade and industrial practices in regards to China may be required.

1. Because historically people absolutely do have a hard time separating it? Particularly when politicians and pundits aren’t careful about the distinction.

1. You overstate things. Even historically, people were able to separate out the actions of Nazi Germany and German Americans. Chinese Americans today are easily as well integrated as German Americans in 1940.

1. Umm, no. German Americans were treated quite badly during both wars.

1. If by treated badly, you mean drafted and sent to fight the Nazis, then sure.

2. You know who might disagree? All the Japanese people we put in camps during World War II.

And how about Muslims in the post 9/11 world? For instance , a sitting congressman named Tom Tancredo was suggested we nuke Mecca in response to a future attack by an Islamic militant group.

1. The 1940’s were a different time, and the Japanese-Americans were less well integrated. Believe it or not, things have changed since 1942.

As for Muslims in the post 9/11 world? A good way to look at this is the FBI’s hate crime list. A fairly small number of crimes, considering it’s nationwide. But Muslims only make up ~15% of the religious hate crimes. Again, a fairly small number. To put this in context, there were more than three times as many hate crimes against white people (because they were white) than Muslims (because they were Muslims). (And yes, arabs is a separate category that is much, much lower their either).

https://ucr.fbi.gov/hate-crime/2018/topic-pages/victims

The argument you seem to be making is that governments and countries can’t be criticized at all for their actions, because of the potential for potentially discriminatory/racist actions by people here, whether or not it actually exists here.

1. A fairly small number of crimes, considering it’s nationwide. But Muslims only make up ~15% of the religious hate crimes. Again, a fairly small number. To put this in context, there were more than three times as many hate crimes against white people (because they were white) than Muslims (because they were Muslims).

So 15% against Muslims, and 45% against white people.

Well, Muslims are about 1% of the population, white people about 75%.

And here you are, suggesting that whites are more victimized by hate crimes than Muslims. That’s idiotic.

2. “And here you are, suggesting that whites are more victimized by hate crimes than Muslims. That’s idiotic.”

Sigh…. No, it was to put the numbers in context. And unless you think there are a raging number of anti-white hate crimes, that should indicate there aren’t that many anti-Muslim hate crimes.

Also, look at the link and the actual statistics. The way you report the numbers is incorrect.

3. No. I’m saying that you need to be careful to draw the distinction.

Also, you implicitly admitted I was right about that historically being the case.

I’m not sure “hate crimes” is the necessary measure of anti-Muslim animus in America. That’s just reflects the people willing to commit crimes that are violent enough to get reported to the FBI. It doesn’t actually address the fact that there is widespread anti-Muslim feeling in the country post 9/11 world, and that people tend to conflate Islam with terror.

2. Are there people suggesting we attack Taiwan that I’m not aware of?

3. Why do emotive ‘thinkers’ believe accusations of isms/phobia w/o proof is a victory?

3. The question that arises is why Trumpists are so eager to find a reason, any reason, to deflect blame from Trump.

Why are the media “so eager to carry water for the Chinese Communist Party?”

Maybe because they don’t think the virus originated in the lab. IOW, unlike Fox, they care about facts.

1. “The question that arises is why Trumpists are so eager to find a reason, any reason, to deflect blame from Trump.”

There has to BE blame before you need to deflect blame. In what way do you suppose Trump is to blame for Covid-19? It didn’t originate in the US, he shut down travel from China early enough that Democrats criticized him for doing it, and death rates are running twice as high in Democrat run states as Republican run states.

Yes, it would have been nice if the US had shut down travel even sooner, and shut down travel from nations besides China at the same time. If he’d done that you’d have called for him to be impeached over it.

1. Brett,

This has been argued endlessly. Your position, as always, is “Trump Can Do No Wrong,” despite the mountains of evidence that he ignored the threat, ignored recommendations from both the Bush and Obama administrations, etc., etc. I know. You have excuses ready. Don’t bother. I’ve heard them.

You maintain that about both his business career and his term in office. Consistently.

Haven’t you been gloating about 2016 long enough?

2. Believe it or not, it’s possible for two things to share blame. If my kids play with matches and start my house on fire, then the fire department is slow to respond or doesn’t make sure the fire is completely or out before leaving (allowing it to flats back up), I can be mad at my kids and the fire department at the same time.

The kids were the initial cause, made worse by the next person’s incompetence. Neither one relieves the other of culpability.

4. See? M L is just asking serious questions. He’s definitely interested in the truth.

4. Oof. Here’s a story from yesterday that, while probably a coincidence, sure as heck won’t put a damper on conspiracy theorizing.

Researcher ‘on verge of making very significant’ coronavirus findings shot to death in apparent murder-suicide

1. If the researcher in the story was “on the verge of making very significant’ coronavirus findings” then unless his work was also stolen, other researcher can learn from his documented findings. Simply killing him without destroying evidence of his work wouldn’t make as much of a difference and I assume much of his work was documented at the University of Pittsburg lab.

1. Eh, maybe. Good researchers are supposed to thoroughly document their work along the way but are human like the rest of us – we keep minimal documentation along the way and get everything cleaned up for publication. There will inevitably be undocumented assumptions and steps that seemed too routine to write down at the time but that will not necessarily be obvious to whoever picks up the research.

And that’s assuming that anyone wants to drop their own research to take a chance on someone else’s claims of “very significant” findings.

1. Yeah. The University’s statement says “We will make an effort to complete what he started in an effort to pay homage to his scientific excellence.”

Sure doesn’t sound like a foregone conclusion that whatever he was working on can be picked up and completed.

2. “If the researcher in the story was “on the verge of making very significant’ coronavirus findings” then unless his work was also stolen, other researcher can learn from his documented findings.”

People don’t do this kind of research on their own, it’s done in teams. The whole team would have to be killed before “other researcher can learn from his documented findings” would become necessary.

Assuming for the sake of argument that “Researcher ‘on verge of making very significant’ coronavirus findings shot to death in apparent murder-suicide” is true, my first thought as to motive would be a dispute over credit for the discovery.

3. Very very few inventions and scientific discoveries occur in a vacuum. They occur because other inventions and discoveries have made later ones possible. No one can definitively say who invented the car because there was no single inventor. Watt gets the credit for steam engines, but he built on Newcomen’s work, and he held back high pressure steam engines needed for locomotives because he did not have the skill to work on higher pressures and used his patents to block others with more skill. The Wright brothers get all the credit for the first powered controlled flight, but the power was not enough for takeoff on its own and needed a catapult to get off the ground, and the control was so lousy that they spent several more years working on it.

The headline for that story was right out of a comic book.

5. Why do Law Professors and commentors here feel the need to “analysis” (using that phrase loosely here) area’s outside their expertise? I see only lawyers and engineers doing this. Is it because both see the world in black and white?

You are not helping. Leave the analysis to the experts.

1. Leave the analysis to the low IQ propagandists that call themselves journalists at the Washington Post.

2. Lawyers don’t see things in black and white , they have a different problem. The law touches many areas of life, therefore lawyers have to learn a little bit about a lot of things. They also have to be able to communicate that little bit to someone who knows even less than they do about the topic: a judge or jury. As a result they tend to have an overinflated sense of their competence in a lot of different areas and feel free to opine with authority on anything. This phenomenon is even worse in legal academia.

Fully aware that I might be an example of the the way of thinking I think is flawed by offering an unfounded opinion, to the extent that engineers think they’re experts on everything it’s probably because they have to learn complicated things to solve complicated problems. Even though they learn that for a particular field, they might assume that means they can solve anything else too.

3. Statistics and basic bayesian analysis is outside the expertise of engineers? It shouldn’t even be outside the expertise of good lawyers. A failure to understand Bayes is at the root of the Prosecutor’s Fallacy.

Contrary to your other claim, I know of almost no lawyers who see the world in black and white. The most common complaint against them is that they see everything in shades of gray.

Engineers, on the other hand, do see the world starkly in terms of those-who-understand-math and those-who-do-not. The Washington Post author apparently does not. And you pretty clearly do not.

1. It shouldn’t even be outside the expertise of good lawyers.

If that’s true there are damn few good lawyers.

1. And damn few good engineers.

I don’t claim to be an engineer, but when the first ten ignore me, only to have the 11th say “oh, shyte, he’s right”, I tend to question the qualifications of the first ten.

2. I would classify anyone who endorses this article as someone who-does-not-understand-math.

Almost every input is a fairly arbitrary choice that could easily change by an order of magnitude either way.

And since the whole analysis is just multiplying and dividing these inputs you can easily change the answer by orders of magnitude in either direction with only mild adjustments.

1. I would classify anyone who makes a post like this as someone who doesn’t understand math.

He makes all his assumptions a forte, so they are as against his conclusion as he thinks could possibly be reasonable.

And you clearly don’t understand how Bayes works.

But please, try different numbers. See what kind of assumptions are needed to move the needle an order of magnitude. I’ll wait.

1. 99.7%

That’s what you get if you remove his “oooh, I’m being skeptical!” factor of 100.

Does that sound reasonable?

Not to mention a bunch variables that he treats as independent are anything but. Which means that even if his individual probability estimates were somehow correct his calculations would still be completely invalid.

I think this is literally the kind of analysis that would cause you to fail a 1st year stats assignment.

1. It’s not stats. It’s Bayes Theorem. (Probability is not statistics).

If you have specific criticisms of his priors or evidence, make them, and do your own calculation.

Bayes is a method of measuring independence, ultimately. (Is the location of the lab independent of the outbreak, for a pertinent example). And the terms in Bayes are based on assessments of evidence and priors, which can (and should) explicitly take into account independence issues.

It’s also the case that a formal Bayes analysis would look to establish conditional independence of variables. But for a quick ‘back-of-napkin’ calculation, Bayes is reasonably robust to failures in indepedence assumptions.

(It’s also not clear which variables you are claiming lack independence here.)

1. I learned Bayes Theorem in a stats course which I think is fairly typical. Stats is largely about understanding probability, and Bayes is one of the basic tools for that.

And you keep ignoring the fact I did do my own calculation. Removing his arbitrary division of 100 yields a nonsensical probability of 99.7.

And Bayes is only “robust” to failures in independence assumptions is the covariance between the variables is small. Once you start chaining calculations and introducing highly correlated variables the robustness goes away.

1. aluchko:

I hope the squirrel replies again. I’m enjoying watching you pimp-slap him around the comments.

4. Why do commentors here feel the need to rebuke others for expressing an opinion on a subject where they have some knowledge? Is it because they lack the knowledge and think it elitist for amateurs to show off where they cannot?

1. I think they’re skeptical of the presence of knowledge given the field and the training of the writer.

2. There is nothing wrong with expressing an opinion.

But opinions of the form, “I have discovered something obvious, wbich has been overlooked by thoise who actually know something,” are silly.

Let’s call them Epsteins, in honor of Richard Epstein.

6. We should have wiped China off the map long ago.

1. We should have killed one 1/6 of the world population?

1. Even if we killed it’s entire population (I’m skeptical this would even be possible), the geographic region known as China would still exist, and would therefore still properly appear on maps.

1. We don’t have to do anything to the actual country or its people, just wipe it off the map. Like Camp David.

7. “The possibility of a man-made virus worse than the novel coronavirus should terrify us, perhaps even posing an existential risk on the same order of magnitude as climate change.”

Ooh, I’m scared now.

8. I gave up on The Washington Post Fact Checker a long time ago. I saw too many cases where the conclusion appeared to me counter factual and politically motivated. The final straw was when they actually said that a politician’s statement was factually correct, but they awarded 2 Pinocchios anyhow.

1. The probability of a ‘fact checker’ confirming a fact is not necessarily correlated with the factuality of the fact. It is not merely a random coincidence if the fact-checker’s conclusions conform with the publisher’s prejudices.

2. Ad hominem is pretty boring. Don’t take them as authority, but read their analysis at least.

1. I read their analysis from time to time. They’re not bad on topics that have no political implications at all. Then again, they don’t tend to fact check such topics, so that doesn’t get you very far.

Once a topic has political implications, the thumb starts coming down on the scale, and they’re perfectly capable of demonstrating that a claim is factual, and still rating it only half true, just because they don’t like the implications.

Their “fact checker” maintains their database of Trump “lies”, too, and that’s a real joke as well.

9. As far as I can tell, your Bayesian analysis ignores the purported evidence that this coronavirus is a distant ancestor with research coronaviruses, not that I have the qualification to evaluate that evidence.

The problem with a Bayesian analysis like this is it’s easy to include only the evidence that supports your opinion. I’d need a further smoking gun to actually support this position. Not that I put any stock in WaPo’s objectivity in weighing this sort of evidence.

One advantage of this theory is it helps explain why the government was so interested in keeping this under wraps, but I’m guessing that was more incompetence and China being China.

10. “emerged from nature” and “escaped from a lab” are not mutually exclusive

Virus sample taken from bat for study. Sloppy handling results in infected researcher who then infects others outside the lab.

1. And: Oh, look there’s a wet market conveniently near the lab where the infected researcher probably did at least some of his grocery shopping.

1. Or had close contact with someone who did.

It seems that the simple analysis is to ask which is more likely, a virus that originated in bats 50 miles away magically appears in this particular market where no bats are sold, or a lab that had been collecting samples of coronaviruses from said bats accidentally allowed one virus to escape and spread in a wet market adjacent to the lab.

2. “Virus sample taken from bat for study. Sloppy handling results in infected researcher who then infects others outside the lab.”

We need to regulate the bats before they get into labs!

11. What kind of molestation of statistics is this?

This is not how you decide on how likely something is. This is just confirmation bias with some numbers painted on.

Coincidences happen. Using math to see how unlikely a given one is is a great way to talk yourself into a conspiracy.

1. Once is happenstance. Twice is coincidence. Three times is enemy action.

1. Depends if you count the following lies and cover ups…

1. Don’t change the goal posts. That’s not part of the numbers game being pushed in the OP.

I learned about how probability evidence isn’t really proof of anything in freaking law school, what’s this law and economics professor’s excuse?

1. Law and economics is about 95% amateur hour, sarcastro.

But, probability evidence can actually be pretty strong, no matter what your innumerate law professors taught you.

Certianly better than “the demeanor of the witness,” or similar BS.

1. There are criminal law cases on why it’s disallowed. I could look them up but I’m tired and lazy.

And why is fallacious stuff like this – assuming an outcome around the defendant being innocent and then calculating that it’s super unlikely.
‘What are the odds there was another green car on that street?’ is just a bad question to ask.

2. Probability evidence can actually be pretty strong when it’s used correctly and understood correctly.

The problem is that lawyers and jury layman don’t understand how to correctly understand and use statistics. Even “experts” often misuse statistics. A great example is Sally Clark.

3. Fair, Max. I shouldn’t say it’s never useful, just that it’s on a whole more prejudicial than probative to laypeople.

2. NO. Third — that we know of….

1. Yes, let us not forget all of the secret things we don’t know about in calculating the perfidy of our nemesis.

3. Did you forget 2003, Sarcastr0? I think you did. This is not the first time China lied and obfuscated regarding a contagious virus.

You think whatever China did in 2003 makes it more likely that the OP’s nonsense about a lab is correct?

1. Your statement I responded to was: This is once.

My point to you was this is repeated behavior on China’s part. The origin of the virus to me is irrelevant. China’s purposeful actions post-discovery of the virus is very relevant to me. And I think, to every American.

The fact that it is repeated behavior makes it all the more worse.

1. The thread is about whether this is coincidence. AL says this happened 3 times. That’s not what happened 3 times, and it’s not part of the post I was criticizing.

We’ll see what people say about China once this shakes out. I think by overplaying it’s hand this hard the Trump administration is giving up on more legitimate issues it could take with China. And making it partisan.
We shall see what happens, but if Trump goes full China truther, I would not be surprised if China didn’t get away with only some international grumbling and treating this like a purely natural disaster.

I’m also really not sure about the case that this could have been contained if China had acted better, given it’s extraordinary virulence. I don’t think the facts are in for either side of that scenario.

2. Like, this is offensively ignorant.

3. All sound argumentation can be reduced to Bayes Theorem. You can literally prove that. (At least one person already did, actually, but I’m not looking it up right now). This isn’t statistics, this is reasoning via Bayes Theorem, and it’s on firm mathematical footing.

If you have a problem with his priors or his evidence, state your case and do your own Bayes calculation. Put up or shut up.

1. All sound argumentation can be reduced to Bayes Theorem.

Why would anyone suppose either politics or policy is, or should be, dependent on, “sound argumentation,” whatever that is? Extending Bayes to analyze real-life human experience is rationalist fallacy run wild.

Politics, policy, and human experience generally present neither measurable initial probabilities, nor independent variables. Experience is full of feedback loops, irrational responses, perverse resistance, hormones, and god knows what all. That, at least in part, is why William James was correct when he said that human affairs will be very little governed by reason, but instead by, “prejudices, partialities, cupidities, and excitements.” Do you suppose there are any measurable points on those axes?

Of course, economics is not yet even aspirationally a science, and almost never solves any non-trivial problem relating to future outcomes. Statistics applied to politics, policy, and human affairs do no better.

But just for the fun of confounding the argument in the OP, let’s play along, and add a few initial variables. First, the number of coronavirus importations per unit of time at the lab. Second, the number of coronavirus importations per unit of time at the market. Third, the number of coronavirus manipulations per unit of time at the lab. Fourth, the number of human interactions with coronavirus carriers (animals or humans) per unit of time at the market. Fifth, the probability of a coronavirus security breach at the lab. Sixth, the probability that a lab coronavirus security breach, if it occurs, will infect a human at the lab. Seventh, after reckoning zero coronavirus security at the market, the probability of a coronavirus transmission to humans at the market, if coronavirus is present at the market.

Finally, what is the probability that a coronavirus transmission to humans at the market ends up infecting humans at the lab—by whatever undetected chain of transmission—compared to the probability that an infection found at the lab originated from coronavirus imported to the lab, but not from the market?

So here I am Squirrelloid, ready to put up or shut up, and insist that without command of the probabilities I mentioned—and doubtless other kinds of convolutions without practical limit—nobody can do the kinds of analysis the OP pretends. It is all bogus rationalism. Don’t fall for it.

More generally, avoid rationalism in politics. It is not useful to suppose anyone can reason from axioms to discover the facts of human affairs.

1. “”sound argumentation”, whatever that is”

You’re not off to a good start. ‘Sound’ has specific meaning in logic. (Specifically “logically valid and having true premises”.) If an argument is unsound, it is necessarily false. So when I say all sound argumentation is reduceable to Bayes Theorem, that means the logic of all true arguments is equivalent to Bayes Theorem.

Second, Bayes is specifically intended for human experience. Look up the epistemic interpretation of Bayes. It’s certainly not a rationalist fallacy, because it’s not rationalism at all, since it doesn’t depend on a priori reasoning. (Maybe also look up philosophical rationalism). Bayes is primarily a materialist perspective, because it evaluates evidence to see if enough is known to overcome prior beliefs. I might suggest you have no clue what you’re talking about.

And why those variables? Why is that the most relevant information? Why not use more specific data, like ‘How many bats are traded at the wet market?’ (none) Or what is the source location of the virus (Yunnan, quite far from Wuhan)? Had the lab collected samples from there? (Yes) Did they have samples of a similar coronavirus (Yes, the most similar one we know of)? What’s the likelihood of non-scientific transport from Yunnan to Wuhan without intermediary detection? And that it still appeared in the immediate vicinity of the lab as the first detection?

And you haven’t put up at all. If you think those are the relevant variables, do the math. Assess how that affects the actual probabilities involved. Use your best guesses. The epistemic version of Bayes is about updating your beliefs as you get new evidence, not being an absolute judge of the truth. It’s always subject to change as more evidence is received.

I’m not sure what politics has to do with this. Whether or not the virus escaped from the lab is not primarily a political question. (The kinds of answers governments will tolerate is, but at its base this is a question of fact).

1. Squirrelloid, what you argue amounts to this: that you know how to predict how humans would behave if humans were numbers. Then you take an extra, utterly fanciful leap, and insist that even if humans are not numbers, they will behave just as if they were numbers. That is the fallacy.

By the way, it seems like you do both those things without noticing you are doing them. If so, that puts you in plentiful company. Mathematically inclined rationalists habitually think that way. Maybe you draw encouragement from that. But it will never work.

Bayes gave you a repeatable method for thinking about systems simple enough to permit measurements of their inputs—or, in the absence of measurements, to bound speculative inputs within some still-simple limits—limits situated among tidily imaginable interactions, and not-too-numerous feedback loops. All that has its uses.

However, it does not, and will never, describe human experience. Or natural experience either, except under constraints so severe they limit analysis to simpler cases. The relevance here is that modeling the origin of a pandemic is not a simple case.

As for filling in numbers useful for the extra inputs I suggested? Realistic assumptions for those inputs would rule out any inference that the laboratory is the most likely unique source of the pandemic. If you cannot see that at a glance, then I suppose I am wasting my time trying to exchange views with you.

Or, correct me. But don’t change the subject, as you did above. Use your imagination, and pick inputs for the values I asked for, values you think look reasonable, which will also falsify what I just said. I will take a look, and if your values are not obvious nonsense, we can talk about them.

One final question. Do you like to play poker? In college, I had liberal-arts-type friends who enjoyed good times playing poker against mathematically-oriented rationalists—opponents my friends learned to prefer above others.

My friends discovered they could bait a reliable supply of math types. They did it by posting on the bulletin board at the Hall of Graduate Studies an inconspicuous invitation to a little game among undergraduates. For some reason, liberal arts scholars shunned that invitation, but the mathematically proficient saw it as an opportunity.

Thus, the liberal arts undergraduates took on the math pros, and one-by-one cleaned out the engineers, mathematicians, and physicists who showed up—all of them starting out just as confident as you are about universal advantages to be had from mathematics (and, I suppose, from sound argumentation as well).

That confidence never lasted. The results were not subtle. The key realization came quickly, or less quickly, but always, and promptly enough. That was the realization that the longer they played, the more they would lose.

Neither luck nor superior mathematical insight would usefully affect the outcome. Only poker acumen really mattered, and these mathematical types did not know what that was. After that dawned, as the grad students went out the door, you could see on their faces signs of real emotional turmoil. These were smart people—smart enough to figure out that what they had experienced was systematic failure. They had learned abruptly that principles they expected to rely on professionally had limits they had not supposed. It was disturbing.

Squirelloid, when the pandemic passes, take your mathematical rationalism to Vegas, and put it to the poker test. You might learn something too.

1. What does any of this have to do with how humans react to things? You’re reading tea leaves.

You’re still badly mischaracterizing Bayes (and, for that matter, rationalism, which Bayes is totally unrelated to. Bayes is an empirical method).

Bayes is a formalism of all proper reasoning. When you reason properly, you’re reasoning is described by Bayes Theorem. Any reasoning which cannot be represented by Bayes Theorem is necessarily wrong.

I can’t do the analysis with your suggested elements. That’s not evidence we have. Bayes is about dealing with the evidence we do have. Nor is it particularly relevant if the market is swimming in coronavirus if it isn’t the right kinds of coronavirus. Because covid-19 didn’t just come from any possible coronavirus, it came from a chimaera that almost certainly arose in a bat.

So let’s take a stab at some considerations *adjacent* to your highly specific requested data which no one has.

Note: Background information is estimating the relative likelihood (odds) of each hypothesis given known background information. Evidence is giving the relative likelihood (odds) of having a piece of evidence given each hypothesis.

Bat CoV: I won’t say the flux of bat coronavirus in the Wuhan market is zero, but it’s pretty small, because *bats aren’t traded at the Wuhan market*. Ie, odds of having a bat origin CoV as evidence is unlikely if it originated in the market. By contrast, we know the Wuhan lab had bat CoV samples, because they’ve published papers characterizing them and manipulating them. Some other origin is possible, but R. affinis, the bat species with the most similar CoV strain, isn’t known to be native to Hubei, so the evidence is unlikely if it was some other bat encounter, but not as unlikely as the market (ie, i have less certainty the natural populations of R. affinis have been accurately mapped, so it’s more likely it’s present in Hubei generally than at the Wuhan market specifically).

Pangolin CoV: Pangolins, which are a possible (but less likely) intermediary, also weren’t listed as being traded in the market, but unlike bats, pangolins are highly illegal, so possibly someone illegally smuggled them in. On the other hand, we know the lab also had Pangolin CoV, because it had gotten samples from some illegally trafficked sick Pangolins that had been recovered by chinese authorities. Pangolins are native to Hubei (where Wuhan is located), so some other method of transmission is possible.

There’s no evidence in favor of other intermediaries, but we know tigers and dogs have been infected with Covid-19, so some other intermediary is possible.

To put some odds to this:
Bat vs. Pangolin vs. something else as immediate pre-human host: 5:3:1. That’s a loose guess. Those aren’t terrible odds for any particular immediate origin.

If it is a bat, the relative odds of the lab:market:other,
arguing a forte, 10:1:2. It can’t reasonably be any less likely than that for the bat CoV origin to be the lab. The last number is least certain, but without evidence of R. affinis native to the area, i find it hard to credit higher odds.

If it is a pangolin, I’m going to start with (lab:market:other) 5:1:2. We know the lab had relevant Pangolin CoV. I rate the market at half some other pangolin encounter, because the market requires deliberate smuggling of a live animal, which is less likely than simply encountering a pangolin because it requires assuming additional unknowns (we don’t know pangolins were smuggled into the market).

If it is some other animal, we don’t have any specific information. Lacking evidence, we assign equal probabilities, which means 1:1:1 odds.

Accidental release, background knowledge
Accidental lab releases are, as far as we know, relatively rare but not unheard of. The 1977 detection of 1950 flu in Russia and China, for example. On the other hand, natural zoonotic pandemics are not uncommon (at least 5 flu strains since 1889, for example). And if we should be counting all zoonotic transmissions which attain human-human transmission, even more common than pandemics. Reasonable ‘worst case’ guess on background here is 100:1 (natural:lab accident), but we want to argue a forte, so let’s go with 1000:1, which is almost certainly over-stating the case, but it’s surely no more disparate than that.

We’ll split the ‘natural’ between ‘other’ and ‘market’ two different ways, and run two calculations. Most favorable to the market hypothesis: 1:500:500 (I can’t possibly imagine it being higher). And favoring ‘other’, 1:100:900. (I’m also going to assume these odds are constant regardless of immediate pre-human host organism identity).

Geography
The most similar CoV we know of is from Yunnan province, near Pu’er. It’s pretty similar to covid-19, aside from the insertion from Pangolin CoV, and certainly much more similar than other known CoV strains. Pangolins also occur in Yunnan, so natural recombination is possible.

But Pu’er is quite far from Wuhan. A travel website says air travel between the two takes over 7h, although the same site says you can drive it in a little under 20h. (Must not be a direct flight). 20h is the same driving time as Chicago to Miami, so that’s a pretty far distance.

We know the lab in Wuhan collected CoV samples in Yunnan near Pu’er. They’ve published several such sequences, including RaTG13, which is the most similar strain to covid-19.

For a live animal to get to the Wuhan market, it would need to be smuggled that whole way (meaning air travel is out). That requires an unevidenced assertion, not just of smuggling, but of a particular smuggling route. That seems relatively unlikely, bar evidence of such a route. And if it was a pangolin intermediary, it makes little sense to smuggle pangolin from Yunnan to Hubei when they also live in Hubei. I’m not sure I can justify being any more generous to the market theory than 10:1 in favor of the lab on this point. (I don’t have any specific evidence of live transport of animals in china over long distances, so i’ll concede it’s possible, though reducing weighting due to unevidenced assumptions, and make that 10:1:4 to include some other origin. I’m also including in the ‘other origin’ the virus spreading via natural populations that crossed that significant distance, which also involves a number of unevidenced assumptions).

Let’s work with just this evidence.

The odds form of Bayes states that the ratio of odds for each possible hypothesis is equal to the ratio of the product of odds for each hypothesis given the background and the odds of each piece of evidence given that hypothesis. ie, O(h|b)*O(e1|h)*O(e2|h)*… for all evidence, where O is odds and h is the current hypothesis.

Immediate intermediary given as Bat:pangolin:other
5:3:1 (used as weights on that bat/pangolin/other odds)

All others given as lab:market:other
Background, natural zoonosis vs. lab accident – A. 1:500:500 or B. 1:100:900
Bat – 10:1:2
Pangolin – 5:1:2
Other – 1:1:1
Evidence – Geography: 10:1:4

Resulting final odds (please check my math) lab:market:other:
A. 75:75:1200 (6%, 6%, 89%.)
B. 75:15:2160 (3%, 0.7%, 96%)
Failure to sum to 100% due to rounding. (The 6%s in A, for example, are closer to 5.55%).

Considering this was estimated to argue a forte against a lab origin (the real odds should likely be more generous towards a lab origin), it’s not surprising that ‘other’ came out so strongly, especially with how much I freely admit I don’t know regarding other vector possibilities. But the big story here is that the Wuhan market, even with generous estimates in its favor, doesn’t come out particularly well. All it has going for it is a large background weighting, because none of the evidence favors the market over anything else. (Including some other evidence, like early cases, doesn’t help here either – only 1 of the first 4 cases had visited the market).

I would characterize this result as a lab origin is possible (despite arguing a forte against it) and cannot be ruled out. Given generous estimates in its favor, i’d lean towards rejecting the market as the initial source (at best it performs as well as a lab origin, even though i argued a forte in its favor and against a lab origin). But the best chance is some other origin. I welcome critiques on my estimates of how likely the evidence is given each hypothesis, or on what the evidence actually is.

This also points to specific evidence that could be crucial to shifting the estimation, including better information about wild animal populations in Hubei, a source population for the chimaeric precursor of covid-19 itself, or evidence regarding live animal transportation over long distances in china.

1. Squirrelloid, I think that is the biggest subject change I ever saw on the VC, which is saying something.

Want to play poker?

2. SL makes some good point.
Plenty of others on this thread have a problem with his ‘evidence’ which is speculative claptrap that works hard using made up numbers to stare at a coincidence until it’s evidence. That’s not even how Bayesian stats operate.

As for your putting Bayes on a pedestal, more often than not that is just endorsing confirmation bias as a fine thing to have. Plenty lf other ways to look a facts, but I guess you don’t much care about that.

1. You Nick from HMC?

2. He makes no good points. He misuses basic terms and doesn’t understand Bayes Theorem at all. His list of evidence he wants probably isn’t the most relevant, and he never does the work to show how that evidence should change our conclusions.

Bayes updates as more evidence is compiled. It doesn’t confirm bias unless you intentionally exclude relevant known evidence in a biased way (at which point you’re doing it wrong).

I’m not putting Bayes on a pedestal. It’s a mathematically proven fact that all true reasoning is identical with Bayes.

1. You remind me of when someone first learns about Nash Equilibrium and thinks they’ve solved every policy problem ever.

The thing is, there is currently very little evidence. I count exactly 1 actual data point in the OP’s analysis. That’s what’s so ridiculous in this abuse of statistics.

As to your appeal to logical reasoning, that’s like saying all good decisions are identical with mathematics. Technically true, but not in any useful way.

1. I’ve never maintained Abramowicz is right in his conclusion, just that the method is sound.

Obviously we need better evidence. (Actually, we know more than is in this post in several substantial ways, and the way he goes about thinking of ‘natural’ wild zoonotic transmission vs. lab released disease is probably not the best way to do it).

But impugning valid methodology instead of challenging his priors and evidence is precisely the wrong way to go about criticizing the post.

1. I should probably additionally note that i’d want at least a 99% result to have any faith in a conclusion on this either way. Being in-between means it can’t be definitively answered one way or the other. Which, as you note, given the analysis is light on evidence, doesn’t seem like an unreasonable conclusion.

2. He makes all his assumptions a forte, so they are as against his conclusion as he thinks could possibly be reasonable.

And you clearly don’t understand how Bayes works.

This sounds quite supportive to me, including the totemic invocation of Bayes.

12. What bugs me about this analysis is where the probability of a bat corona virus moving to humans being so unlikely comes from …. in face of the fact that both SARS and MERS were corona viruses moving from bats to humans (via civit and camel respectively) in just the last 18 years.

1. When having fun with numbers, and every number in a long series of products has great possibility for error. When doing order of magnitude estimates of a probability, the error bars widen by an order of magnitude for every multiplication within that calculation potentially. Two wild-ass guesses might be wrong in opposite directions, and the error partly cancel out, but if wrong in the same direction, then the result will be not just wrong, but fantastically wrong.
See this xkcd for illustration of garbage maths

13. 538 has an excellent article on this.

https://fivethirtyeight.com/features/why-scientists-think-the-novel-coronavirus-developed-naturally-not-in-a-chinese-lab/

I think the conclusion was: You know the old story about a million monkeys with a million typewriters–one will finally end up typing out “Hamlet.”? Well, the odds of a scientist/lab creating this virus is about the same as you–standing above a huge room with those million monkeys–happening to point out, at random, that particular monkey. In other words, astronomical odds.

I found the 538 article much more compelling that referenced by the OP here. But different strokes . . . .

1. Haven’t read the 538 article, but based on your summary, that only argues developing the virus de novo. That’s a strawman which no one is arguing. What people are arguing is that they were investigating the corona virus itself, as taken from bats, probably collected from the wet market, one of them had this particular strain, and their sloppy work let it escape.

1. Understood. The article also addresses the 3 main conspiracy theories: (a) Intentional lab-created virus, (b) Accidental creation in the lab, and accidentally spread outside of the lab to the wet market, or (c) truly came from bat (ie, nothing to do with the Wuhan lab. All 3 are covered, and I found it compelling.

For instance, the Chinese govt (all would agree, no shrinking violet) has acted inconsistent with the first two theories, by allowed workers there to speak to media, etc.

The article is not conclusive…nor could it be, being impossible to disprove a negative. But it is not very long, is not super-technical, and is worth the read…even if only to see the other’s side’s arguments.

1. “Allowing workers to speak with the media” Yes, I’m sure those workers aren’t at ALL afraid that they or their family members might be “detained at an unknown location” like the people who posted info on Covid to github. Not at fucking all.

1. Of course your point is well-taken. But it’s worth noting that the Chinese govt always blocks people from talking to the media, foreigners, etc, when it has something to hide…in spite of the overarching fear ordinary Chinese have of speaking against those in power. It’s worth noting when there is an exception to this. Sort of like when Trump doesn’t try to block people from testifying…it’s rare enough that it’s worth mentioning when it does happen.

2. Not accidential creation in the lab, but isolation and being worked with in the lab and accidental release.

It’s kinda like if, oh, there were suddenly a smallpox epidemic. And it suddenly appeared in the same city with a major viral lab that was working on smallpox. But, also, there were some African rodents nearby, that had been there for a while….

1. Armchair,
That analogy does work, if those African rodents nearby had smallpox. Otherwise, the analogy does not make sense.

1. Have we found the specific strain covid-19 in a bat population yet? If we haven’t (and i presume it exists in one somewhere – that the article is arguing for ‘accidentally released a natural virus’ and not ‘the lab created it’), then the proper level of knowledge regarding said rodents would be that they *could* carry it.

Because obviously if we knew which bat population, that might affect our assessment of the likelihood it came from the lab. (Is it a nearby population or far away? Is it one the researchers had live specimens from? etc.)

1. If it were exactly the same strain, it’d still be in bats, not tearing its way through the global human population. Good Christ, man, learn some fucking biology.

More importantly, yes, there are very, very similar versions of the Spike protein (with a polybasic furin cleavage site) in both bats and pangolins. And it’s that Spike protein feature that scientists think makes Covid-19 so much more virulent than SARS/MERS.

1. “Neither the bat betacoronaviruses nor the pangolin betacoronaviruses sampled thus far have polybasic cleavage sites. Although no animal coronavirus has been identified that is sufficiently similar to have served as the direct progenitor of SARS-CoV-2, the diversity of coronaviruses in bats and other species is massively undersampled.”

https://www.nature.com/articles/s41591-020-0820-9

1. I could’ve sworn I read something like two days ago about cleavage being found in bats. (Cleavage… in bats… teehee.) But as I explained to squirrel, I don’t read gud. Maybe just too hastily read this. Not sure.

¯\_(ツ)_/¯

2. It jumped species. It didn’t magically mutate 1000x overnight. Of course it wouldn’t be exactly the same – there would be some differences – but it would be substantially similar.

Zoonotic diseases have a reservoir animal population from which the infection spreads to humans. No such reservoir population has been located yet, nor has a similar chimaeric virus been identified in an animal population.

The closest virus is known from bats (R. affinis) in Yunnan, except for the RBM segment, which is from Pangolins. (That’s why it’s a chimaera). That combination of coronaviruses didn’t happen *in humans* (because otherwise we’d have some evidence of, separately, both the bat coronavirus and the pangolin coronavirus in human populations – and both viruses separately making the jump, meeting in the same human, and then forming a chimaera, is highly unlikely), but if it happened in nature, it happened in some unknown animal intermediary, probably a bat.

1. Actually, no. Just, no.

I will, however, take my own advice and learn to read gud. I misread what you were saying and went somewhere else in my head. It’s lovely in here.

I’m rooting for the yaks.

14. Hence why it is important to call this the CHINESE VIRUS. It originated in China and proliferated throughout the world because the Chinese failed to contain it. At best, it was just straight up negligence from a public policy standpoint. At worst, they let the virus out of the lab (either knowingly or not) and got it into the population.

1. Why would they let it out knowingly without first having the vaccine?

2. Don’t you mean Chinese Government Virus? After all, I’m reliably told, no one would conflate the government with the people.

15. I think it’s fair to call it the Wuhan or China virus, to the extent that China’s disgusting dishonesty helped it spread across the world. But we can also call it the Trump virus, to the extent that Trump’s pathetic performance in Jan and Feb helped spread it across America.

1. Seriously…. January? And Trump was somehow to blame for spreading it?

1. Well, he did not spread it, in the sense of walking around the country, sneezing on people. But he deliberately ignored repeated warnings. (Yes, in January, as well as many many many more warnings in Feb) And he provided zero leadership, in terms of taking decisive action here at home. (He did block much of travel from China, which I give credit for. And would give more credit, if the policy had not been inept enough to allow in about 40,000 people coming from China.)

Thank God for the amazing leadership from the state of California (esp local officials around the Bay Area), which either shamed other states into issuing stay-at-home orders, or showed those states how important such orders were.

It is, of course, impossible to know how many lives Gov Gavin Newsom (and those Bay Area officials) saved around the country. Only 10,000? 100,000? More? It’s all speculation. But, again, thank God for decisive and courageous leadership. Trump’s cowardly refusal to institute a federal approach to making/obtaining PPEs and setting up a nationwide tracing system stands in stark contrast.

1. Mini-dictator Newsom? Who seems to like arresting protestors…

I’ll give California this, they did far, far, better than Cuomo and De Blasio in New York. But Newsom didn’t do anything for the entire month of January or February, only acting on March 4th.

Meanwhile Trump declared a public Health Emergency in late January, attempting to quarantine everyone who had the disease and came into the US.

As for the “importance” of Stay at home orders? Some states don’t actually have them. Like Arkansas. And the death count there is still pretty low. (Just 80 for the entire state). Something to think about.

1. Trump declared a public Health Emergency in late January, attempting to quarantine everyone who had the disease and came into the US.

Come on, AL, that’s complete BS.

2. Ah, yes, January, when Trump should have been doing — something? What?

On January 2nd, there were 44 total reported cases of an ‘unknown’ disease.
On January 11th, China admitted the first death to this ‘novel coronavirus’.
On January 19th, the US patient who was first to later be identified as having the COVID entered the hospital.
On January 21st, China and the WHO first admitted cases of human-to-human transmission. At that time, there were 280 worldwide cases… none known (at that time) to be in the US.
On January 23rd, the first patient (entered hospital on the 19th) was identified as having COVID.
On January 31st, when the first travel restrictions went into place, there 9800 cases world wide (9700 in China) and a total of six in the US.

Again, exactly WHAT should Trump have done, and when? And with what legal justification?

1. Yes, the WHO initially denied human to human transmission.

3. I think it’s fair to call it the Wuhan or China virus, to the extent that China’s disgusting dishonesty helped it spread across the world.

How about just because that’s where it originated, just like we often do? Why the pervasive need to overthink it?

1. Brian,
The Spanish Flu (along with the Black Death, the most widely-known pandemic in human history) was so-named not because that was where it originated, but for other reasons. (Free press, etc). So, since that super-famous pandemic was named for a country where it did not originate, why not allow that for this pandemic? Why not for the country where it has had the worst impact? (ie, us) Or where it received the most media attention (ie, similar criterion or criteria to the Spanish Flu)? Again, us.

I have no religious objection to calling this the China flu. Except that Trump did not do this for quite some time. It was only when his performance was being criticized in the United States and worldwide media that he felt the need to deflect blame, and renaming it fit nicely into his narrative that “It’s all China’s fault.” Remember, this is the same guy who insists that he and his administration have done nothing wrong in response to this pandemic. Nothing. Not one thing. No delay was in any way his fault. Nada.

So excuse me if I am a tad touchy on this topic. President Trump has done nothing to make me reluctant to ascribe sinister ulterior motives to anything he does. About the 30th time he did or said something horrific (ie, by his second month of his presidency), I gave up and said, “No more. No more benefit of the doubt to you.”

1. The Asian Flu (’57) came from Asia. The Hong Kong Flu (’68) came from Hong Kong. The Russian Flu (1889) came from Russia (Bukhara, specifically). And they were all first identified in those places.

But they aren’t named for these places just because they’re origin points, but because that’s where they were first identified. And while I can’t tell you offhand where ‘Spanish Flu’ was first scientifically identified, it is where it was first identified to the public by the press.

We traditionally identify pandemics by where they were first identified publicly. The Spanish Flu isn’t an exception, it’s just the only time (at least to my knowledge) mass censorship meaningfully affected public knowledge of where a pandemic started.

16. From the outset I’d trust the virtually unanimous opinion of researchers in the field that an accidental release is very unlikely, as opposed to some back-of-the-envelope calculations from a law prof.

But lets dive into the analysis because it’s actually kinda funny:

The video accompanying the article acknowledges that it is a coincidence that the disease emerged in the same city where coronavirus bats are studied.

This strikes me as a lab doing it’s job. Now if it was an Ebola outbreak in China and they were studying Ebola then I’d be suspicious.

a lab somewhere in the world causes a bat coronavirus pandemic in a given year is only 1 in 100,000. And let’s assume that non-laboratory zoonosis was ex ante much more likely, with perhaps a 1/1,000 chance of producing a pandemic in a given year.

Because made up numbers are always the basis for sound conclusions.

And really, a 1/1000 change of a non-laboratory zoonosis causing a pandemic in a given year? We literally just had a non-laboratory zoonosis pandemic in 2009 (swine flu).

Wikipedia reports population density of as much as 20,445 per square kilometer in the approximate area of WCDCP, so the circle with radius 0.3 km would have a population of about pi * 0.3 * 0.3 * 20,445 = 5,781.

No, the 33km^2 district has a population density of 20k, which is also irrelevant since the patrons of the seafood market weren’t necessarily residents of that kilometer.

Because the world population is 7.8 billion, P(Neighborhood | ~WCDCP) = 5,781/7,800,000,000 = 7.41 X 10-7.

Is this serious?!? Plugging in the population or Wuhan, or even the province could be vaguely defensible, but the world?!?

I guess you didn’t want a <1% probability.

But let’s suppose, to err against the conspiracy theory side, […] say, 100 times

That is, with these conservative assumptions and this simple model, there is a 78.1% chance that the virus originated in the Wuhan lab.

Garbage in…

Does all this matter? People are not very good Bayesians, but we should strive to think rigorously about probability in matters of public policy.

I’m sorry, but this is possibly one of the best examples I’ve seen of why not to use Bayesian probability in matters of public policy. It’s a bunch of guesstimated numbers multiplied together. Change any one of them and the conclusion is completely different.

1. I’m sorry, but this is possibly one of the best examples I’ve seen of why not to use Bayesian probability in matters of public policy. It’s a bunch of guesstimated numbers multiplied together. Change any one of them and the conclusion is completely different.

Completely agree — way too complicated. He should have just guesstimated an exponent and extrapolated it out to infinity. That seems to be the in thing these days.

2. Amen. Perhaps fair to criticize WaPo for any confidence in its analysis (although they seemed rather hesitant). Crazy to open with a confident statement of probability and lead to purported probabilistic precision, even when peppered with “back of the napkin” qualifiers when the only premise with an evidential base is Archimedes formula for measuring the area of a disc. I offer, with 95% confidence, that there is a 10-90% chance it came from the lab (I made up those numbers).

3. Thank you for clarifying this analysis–it is hard to find the patience to examine it all after finding the first errors. Besides H1N1 2009…There’s MERS 2012 and SARS 2003. So That’s three major zoonotic epidemics. Most people working in this field “inevitable”. Just basing it on these three epidemics we must conclude that the prior for zoonotic based epidemic is on the order of 20%. 1/1000 — where did he get that? It is an incredible claim and would mean we can write off all pandemics in the last millenium ;). Again thanks for your re-analysis.

1. MERS is not a pandemic. With only ~8000 cases worldwide, it’s arguable SARS wasn’t really a pandemic either. Pandemic doesn’t just mean ‘novel virus’. It needs to infect large numbers of people.

That said, I agree 1/1000 is too low. But you’d probably do better taking the 5 recent flu pandemics (1889, 1917, 1957, 1968, 2009), all of which were probably zoonotic. For round numbers, call the year range 1880 to 2020, or 140 years. 5/140 = 1 in 28. That’s probably all the true zoonotic pandemics in that range.

Just making that adjustment reduces the calculated result to ~0.9%. But we’ve ignored a bunch of other evidence, like the relevant bats aren’t native to the area, that bats aren’t actually traded at the wet market, and that safety at the lab was sloppy. Accounting for that knowledge will push the probability up, probably by a decent amount (depending on how you weigh those factors).

While 78% likely from the article is an overestimate, something in the neighborhood of 10-30% would be enough to make us want to find more evidence one way or another. (I don’t know if the relevant bats were present in the lab, for example).

I haven’t done anything with the population distribution assumptions, because fixes would all tend to increase the probability it came from the lab, if only because they would tend to reduce the probability outside China (and since the lab is in china, it’s relative probability would increase).

4. This is one of the most substantive critical comments, and of its 6 specific criticisms, only the one concerning zoonosis seems like it might have merit, while the ones concerning the world population and the neighborhood population tell me that you don’t understand the model.
More generally, the level of hostility to this well-reasoned analysis with clearly identified assumptions and caveats is surprising. Part of me wonders whether most of the criticisms made by folks applying the principle “that which I cannot understand, I can at least deride.”

1. “while the ones concerning the world population and the neighborhood population tell me that you don’t understand the model.”

Using the world population means that everyone on the planet would have an equal probability of being the source of the non-lab cross-over.

Do you really believe that the type of wild virus a lab studies is an independent variable of the location of that virus on the planet?

Because if they’re not independent variables he’s doing Bayesian probability wrong.

I’m hostile to the “analysis” because it’s useless. Just try playing with the numbers. You could easily get a probability 99% just by tweaking any number of his assumptions in ways that don’t look any less justifiable.

1. 1. That’s not a failure of independence that matters for the argument.

2. Yes, I’d believe lab location and virus/host location are reasonably independent. Scientists collect samples all over the world – very few actually study things in their own back yard.

3. The presumptive bat host origin isn’t located anywhere near Wuhan, but in Yunnan, which is quite a long way away even by human standards, much less bat standards. Strongly argues against any geographical dependence issues.

Regardless, any alteration of the ‘location’ calculation would increase the relative probability it came from the lab, because the lab is in china, and the probabilities for elsewhere would on average go down relative to china.

If you think you can get 99% with reasonable assumptions, prove it. Show us what assumptions would be required to reach that conclusion.

1. 1. I may have answered this elsewhere

2. A group of viruses in China (that previously crossed over into humans) being studied by a Chinese lab? I think there’s a pretty high covariance. And just how many labs or Universities in China were studying bat Coronaviruses? Remember, we’d be having a similar discussion if the origin was close to another facility that had access. Or if patient zero was in another city, but had been in close contact with someone from the lab.

3. Ok, I hadn’t checked that bit of geography. But here’s another critical factor missing from the calculation

P(Virus origin in Yuhan bats | Closest studied ancestor is in Yuhan)

I don’t know the answer to that. That closest ancestor is apparently ~20 years distant in viral evolution, does that mean the original virus is probably just in that local population of bats, or there bat populations with similar sets of Coronaviruses all over China.

That’s the problem with Bayes theorem on problems like this. It only works for giving you reasonable rough answers if you only have very few terms, their all fairly independent, and you have good estimates for all of them.

And I can’t believe I bothered to do this but here you go,
all the numbers and calculations from the article in runnable form
.

There’s actually a change of over a percentage from the article just from rounding.

In fact, try deleting all of the numbers and trying to recreate them based on your own “reasonable assumptions” for the values, see how wildly the percentage changes.

2. My point was more general. In this case no one – including the author – believes the 78%, or that the model is or ever will be complete. But it is a rough cut at a framework within which more discussion can occur, and insights result. It has already converted (for some) a vague intuition concerning the locality coincidence into a clearer understanding of the influence this may have on the overall conclusion.
Systematized clear thinking – even if it involves numbers, even if it makes one’s head hurt, even if it reduces the relative value of rhetoric – is not the spawn of the devil. It is one of the lights of the world.

1. It’s a rough cut of a dead-end framework. I’ve already posted the Google co-lab, you can try yourself and see how numerically unstable the framework is.

The problem with this kind of “analysis” is it primes you to start basing your estimates off that 78%. In reality, the author’s “gut instinct” probably put the number between 10% and 90%, and so he fiddled with the model and and probabilities until he got a “reasonable” number.

I base that conclusion on the fact that if he constructed his model and made his estimates before running anything I’m guessing he’d be choosing the probability from a fairly uniform distribution of about [0.001-99.99]%

It’s worthwhile discussing some of the factors in a systematized way, but this analysis generates nothing but authoritative looking nonsense. Sometimes all we have is “good judgement”, something the WP fact checkers exercised in talking to researchers who dismissed the possibility with good reason.

17. The post didn’t address the prior for how well SARS-CoV-2 matches the type of virus that might escape from a lab vs jump the species barrier in a wet market. The Medium link below covers this, as well as documenting the several (not exactly rare) instances of viruses escaping from labs.

18. Almost everyone seems agreed that the virus was not man-made. Almost everyone seems agreed that the first cases detected and explored were in China. Studies on similar viruses were known to be in progress at three labs (one in Texas, one in China, and one in Russia) and one of those labs is known to have had an accident.

There are two basic difficulties — timing and propagation — with the various Wuhan origin theories. If we make an assumption that most mild CoViD cases were dismissed as flu, CoViD must have been in circulation far earlier than the first known case. If we look at a graph (like https://nextstrain.org/ncov/europe?l=radial ) and search for “miracles” [strains which have no ancestor but magically appear in a location], the origin choices narrow. It seems too soon to declare anything “fact.”

1. What everyone agrees or not isn’t really evidence of anything. I mean, almost everyone agreed there was a medium light propogated through in space at one point (the ‘ether’) rather than a vacuum.

What the evidence shows is that:
1. the most similar strain we know of in nature came from a bat in Yunnan (far from Wuhan),
2. that said strain is not a chimaera,
3. that said strain was known in full only to the Wuhan lab before the outbreak,
4. that there is no known reservoir population of any species from which the chimaera covid-19 could have come from,
5. that it almost certainly did not come directly from Pangolins (the species host of the coronavirus from which other viral genome bits primarily came from),
6. and that the lab in Wuhan also had samples of that Pangolin CoV strain before the outbreak.

Whether it recombined in nature, recombined in the lab naturally, or was artificially created are not readily discriminable from those facts.

19. I winder if Abramowicz’ ass hurts after pulling so many numbers out of it. Note that his 1/100,000 WAG is noit really the probability of an accident, but the probability of an accident leading to a Covid-19 infection

Those are two different things.

1. Ok, now that was funny = I winder if Abramowicz’ ass hurts after pulling so many numbers out of it.

2. The product of two garbage numbers is worse garbage.
Every time two wild ass guesses are multiplied or divided, then the error bars for those assumptions are also multiplied or divided. And once the error bars include both 0% and 100% simultaneously, then the resultant calculation is most definitely non-conclusive.

20. That is, with these conservative assumptions and this simple model, there is a 78.1% chance that the virus originated in the Wuhan lab.

This kind of reasoning is why I always go to an epidemiologist for legal advice.

But then I also believe that there have only been 500 deaths due to COVID-19. Or is it 5,000? Orders of magnitude are so confusing.

21. This is, perhaps, the most shameful thing ever written on the VC. And I include the writings of ex-conspirator Clayton Cramer and mostly-absent Jim Lindgren.

When the so-called “Conservative Elite” tsk tsks the people running around, spouting the next moron theory espoused by either the President or Alex Jones (but I repeat myself), they should instead spend more time looking in the mirror.

+1 Useful Idiot.

1. Yeah, this is a shamefully ignorant post. The next generation of conservative luminaries, folks!

22. How enlightening. Someone tries to do numerate calculus of logic in public and the Epistemological Trespassers line up to put their toe across the line in another’s field.

Before there was a calculus of logic there was only rhetoric with which to seek truth and the caution against ERISTIC. Read and understand James Franklin’s The Science of Conjecture: Evidence and Probability Before Pascal.

Bayesian Inference is tough beyond descriptive frequentist statistics, especially for all other than specialist epidiemiologists.

23. “Now that we have seen what a naturally occurring virus can do…”

I take it you mean smallpox, measles, rabies, ebola, marburg, HIV…

We haven’t just “now” seen what a naturally occurring virus can do. The first three have been around for ages, and the last three for decades. The Wu Flu isn’t in the same league.

1. Prof. Volokh mentioned about a month ago how our willingness to accept deaths as a price to pay is lower than it was in the past.

I think that’s a good point, and one that deserves more attention, maybe from a risk analyst academic guest blogger.

24. Anyone who thinks this analysis is giving a reasonable probability estimate should try running the numbers themselves and see just how wildly arbitrary the output is.

25. I don’t think this post is the best possible Bayesian analysis of the question. I think it leaves out a lot of known evidence, and doesn’t do a good job trying to estimate the parameters it does use. But as a back-of-napkin characterization, it’s an okay *starting point* to clarify where better estimates or models may be needed.

It should also not be taken as a measure of truth, but as a measure of the author’s beliefs, before and after the analysis. Someone with better informed beliefs would have better priors. This is of course a soluble problem. As Abramowicz gets better evidence, it continues to adjust his viable beliefs about the question (which become his priors for the next iteration), and this is well modeled through a recursive application of Bayes.

I think the explanation could also be significantly improved in how it discusses the terms involved, and an increase in clarity could likely also be achieved by using the odds form of Bayes instead of the probability formulation.

But the objections should not focus on trashing Bayes (which is fallacious, since the Bayes method is mathematically proven), but rather on:

1. Better estimations of key terms, preferably with references
2. Identification of additional evidence which should be incorporated into the analysis (with estimates of actual probabilities)
3. Redoing the calculation under 1 and/or 2 to update the estimate of belief strength.

Going through Abramowicz’s calculation is a good way to figure out where you specifically agree or disagree with his argument. That has a lot of value on its own.

1. “It should also not be taken as a measure of truth, but as a measure of the author’s beliefs, before … the analysis.”

So true.

We have learned that someone believed something, then made up some number to justify what they already believed, and then published it so gullible people will now be able to spread more poison on the internet, but can say, “LOOK! A REAL PERFESSER SAID IT’S, LIKE, 80% LIKELY THAT IT’S A CHINESE BIOWEAPON! DERP DERP WUHAN BIOFLU!!!!!!”

Bayesian analysis is great; but Bayes does not prevent GIGO, and it does not stop reasonably intelligent people from abdicating fro any and all responsibility for their words and actions.

This is shameful.

1. Professor Abramowicz’s post is useful, admirable, and intellectually honest. What’s clear is that we don’t have a lot of evidence to answer the question with any degree of certainty, and yet this is a reasonable and appropriate question to look at.

What’s shameful is the Washington Post, and others, who to varying degrees pretend that they do know the answer with relative certainty. They use unwarranted conclusions and misleading headlines for propaganda purposes, to the effect of shouting down the question.

1. No, that’s not it at all.

Here, let me examine the difference between the two.

A right-wing conspiracy theory start circulating among the usual suspects. “Hey, wouldn’t it be SO AWESOME if this was all China’s fault? Maybe … a bioweapon? Or an accidental leak?”

Then they just keep pushing the theory.

Responsible journalists try to check the details. They inquire with responsible sources. You know, scientists (“No, it’s not manmade.”). With intelligence operatives (“No, we have no evidence that it came from a lab.”). With experts in epidemiology (“No, we can’t even say for certain that it originated in Wuhan.”).

In other words, there is absolutely no more evidence for this, right now, than me saying, “There is a giant hamster in Donald Trump’s head that is making all of his decisions.”

So what is shameful is someone making a terrible argument, with numbers that he has completely made up, and propagating to support people, well, like you, who are ignorant, stupid, and will latch on to anything to support you small and pathetic little partisan wishes.

Of course, the danger is that useful idiots like the poster give credence to the administration’s harebrained ideas that might hurt us; and this wouldn’t be small stuff like hydroxychloriquine, either.

No, this would be monumentally stupid things like what the morons are thinking about now. Like a trade embargo (quick, who makes the PPEs?). Or refusing to pay out on the debt China holds (if you think the current economic meltdown is bad …)

Look, when people said in 2016 that by 2020, we’d have an impeachment, Americans dying while Trump ignores it, and the economy in a freefall … I’d take it! This is the best possible outcome, all things considered.

But let’s not go to the full-on meltdown, just because y’all are pissy that you were wrong. Or at least don’t take us with you.

1. The published comments from scientists are best understood as political. The nature paper, which is the best statement of the evidence, literally argues ‘covid-19 isn’t ideally targeted at humans, therefore, it must be natural’. From the nature article:

“While the analyses above suggest that SARS-CoV-2 may bind human ACE2 with high affinity, computational analyses predict that the interaction is not ideal and that the RBD sequence is different from those shown in SARS-CoV to be optimal for receptor binding. Thus, the high-affinity binding of the SARS-CoV-2 spike protein to human ACE2 is most likely the result of natural selection on a human or human-like ACE2 that permits another optimal binding solution to arise. This is strong evidence that SARS-CoV-2 is not the product of purposeful manipulation.”

It’s certainly not strong evidence, despite the Nature article’s claim. It’s probably not evidence either way – it’s assuming an unreasonable background assumption (that all purposefully manipulated viruses are ‘ideal’). Needless to say, that assumption is false, purposefully manipulated viruses aren’t designed peptide by peptide to be ‘optimal’ by some metric, but are frequently formed by making chimaeric combinations with other natural viruses. It’s a surprisingly badly argued claim for a publication of Nature’s stature. (Chimaeras also do happen in nature, so covid-19 being a chimaera isn’t indicative of manipulation, but it isn’t an argument against either.) And that’s not even the only flaw with the Nature paper’s argument.

Now, I’ll certainly agree we don’t know for sure where the virus crossed-over into humans. We only know where it was first identified. But we have to reason under uncertainty all the time. We’ll probably never know who patient-0 was. (And if it didn’t originate in Wuhan, that argues against the Wuhan wet market origin just as much as the lab origin). That said, Wuhan seems like the likeliest origin point – the earliest cases were there, and the highest concentration of cases in China are there. Given how infectious covid-19 is, you’d expect relatively early evidence from its initial area, or at least near its initial area. (It’s an ~7h flight from Pu’er, where RaTG13 was found in bats, to Wuhan, that’s more than twice as long as New York to LA, so even if the covid-19 precursor arose naturally in a bat population near Pu’er, it’s quite unlikely it made the first hop to humans there).

The thing is, an accidental lab leak is not an inherently unreasonable theory. Dismissing it as a ‘conspiracy theory’ without properly vetting the evidence is a type of poisoning the well fallacy, that it’s too implausible to consider. But we know safety at the Wuhan lab wasn’t good. We know accidental lab releases have happened in the past and gone live in human populations. (Most famously, an escaped H1N1 strain from ~1950 showed up in 1977 in China and Russia). And we know the Wuhan lab had both strains from which Covid-19 seems to be derived (the R. affinis-derived CoV that forms most of covid-19, and the Pangolin CoV from which the RBM module comes from).

And none of this has anything to do with Trump. (Nor is China uniquely culpable for poor lab safety – US labs have had some glaring errors in the past too).

Why origin matters is in determining proper responses going forward. If it arose in the lab naturally, that’s probably an animal protocol problem that needs addressing. If it was created in the lab, that’s an argument against doing Gain of Function research. (Both of those potentialities are in addition to the need to improve safety protocols at the lab itself – an error of this magnitude isn’t caused by just one thing, but by a series of mistakes that started well before the critical event). If it’s origin was outside the lab, then different types of precautions involving different parties might be necessary, depending on the circumstances. Anyone concerned about reducing the incidence of future pandemics should be keenly interested in a fact-driven hunt for the origin of the virus.

1. The published comments from scientists are best understood as political.

What. The Crap. Is this.

Screw. You. And your ‘I know better than the scientists how evolution works versus design, so I’ll just discard what they say as flawed due to bias’ Dunning-Kruger nonsense.

It’s one thing to say it’s incomplete science or a flawed study, but you’re impugning the integrity of the authors.

You are one of those who lives in a world of one expert. Thus far you seem mostly an expert of vocabulary.
You and the OP seem peas in a pod.

1. Your May.7.2020 at 3:47 pm post is better on this. It’s still invoking an appeal to expertise to yourself which you don’t have, but at least you don’t attack the scientists you’re trying to contradict via ipse dixit about how genetic design works.

1. https://www.nature.com/articles/nm.3985

How actual coronavirus viral genome manipulation research works. The synthetic virus in this article would, by the standards of the nature correspondence, be rejected as even possibly coming from a lab. Which is obviously the wrong conclusion.

2. Oh look another unhinged partisan rage-rant with no substantive responsiveness.

tl;dr

1. You still down with the OP’s reasoning after these comments?

2. You’ve misinterpreted. His stated priors were strongly against it being released by a lab. The evidence he considered shifted the probabilities to be in favor of it.

Whether or not his evidence was well estimated or most relevant are valid questions. But it isn’t his prior beliefs which drove the conclusion here.

2. It should also not be taken as a measure of truth, but as a measure of the author’s beliefs, before and after the analysis. Someone with better informed beliefs would have better priors. This is of course a soluble problem. As Abramowicz gets better evidence, it continues to adjust his viable beliefs about the question (which become his priors for the next iteration), and this is well modeled through a recursive application of Bayes.

Ah, so you like a foolish surmise with crap numbers versus a foolish surmise without them.

That just makes you easily fooled by numbers.

No one is trashing Bayes, but rather it’s application here. Which I think you know, since it rather looks like your argumentation has changed from your more supportive stance last night to ‘this is still a good exercise!’

3. But the objections should not focus on trashing Bayes (which is fallacious, since the Bayes method is mathematically proven), but rather on:

No one is trashing Bayes, they’re trashing an attempt to apply Bayes in a completely inappropriate manner.

1. Better estimations of key terms, preferably with references
2. Identification of additional evidence which should be incorporated into the analysis (with estimates of actual probabilities)
3. Redoing the calculation under 1 and/or 2 to update the estimate of belief strength.
Going through Abramowicz’s calculation is a good way to figure out where you specifically agree or disagree with his argument. That has a lot of value on its own.

Abramowicz’s calculation has zero value except as a cautionary tale of trying to use methods you don’t understand to prove the experts wrong.

The problem isn’t that some specific number is wrong, or he needs to consider slightly different conditional probabilities. It’s that the entire exercise is a fool’s errand.

Bayes is great for a simple question when you understand the priors well, ie “I work at a biolab and Bob came in with a Flu today, what’s the probability that Bob caught it from work?”

But the reason Bayes isn’t used for public policy questions like this is it’s completely unsuited to questions where the prior probabilities are unclear, covary, and you start chaining things.

But if someone really wants to try refining the estimates then go ahead, you have my full support!!

Just remember to include error bars.

26. “there is a 78.1% chance that the virus originated in the Wuhan lab.”

Damn, three places of precision presented. Every number in the calculation is subject to order-of-magnitude error. There should be some acknowledgement that the precision of the calculation is less precise than even the best single estimate of the numbers input to the equation.

27. I assume that Michael Abramowicz must also be helping Kevin Hassett model Coronovirus predictions.

It is amazing what you can do with math when you both have no knowledge of the area you are modeling and also absolutely no shame or, really, any empathy for the people that will be hurt by your total irresponsibility.

28. The author’s beliefs is precisely the proper expression of Bayesian probability, would you bet a dollar?

Remember that Professor Jaynes taught us to maximize the entropy of the subjective naive prior (don’t lie and guess) and to eschew ad-hockery (make your argument and walk away, don’t try to shore it up ad hoc).

29. For what it’s worth, there is a decent and accessible article in Nature.

https: // http://www.nature.com / articles / d41586-020-01315-7

(just take out the spaces I put in before and after each backslash)

The opening explains a bit about current thinking into the origins. And, best of all, it doesn’t use total crap math and cater to the tinfoil crowd.

1. That paper ignores a known similar strain (RaTG13), published in January by the Wuhan lab. It ignores the methods and outcomes of a lot of actual virus manipulation work, including the published methods of the Wuhan lab and other coronavirus labs. In short, it’s a poorly argued paper that seemingly ignores the most relevant evidence.

This is a much better researched and presented exploration of the question:

The author concludes that it’s not determinable whether it was naturally arising or artificially created. He also takes apart the Nature article’s conclusions pretty convincingly.

1. I strongly support people who learn outside of formal educational settings, this includes biotech.

But there’s always a risk that someone falls into quackery for lack of experienced guidance. Given his field of “expertise” (life extension) there’s are massive warning signs around Yuri Deigin.

And, as a rule of thumb, when a potential quack writes an Internet articles that claims to disprove peer-reviewed research I heavily discount it.

I have no doubt Mr. Deigin knows a lot of stuff about genetics and biotech, but I have severe reservations about his ability to properly apply and contextualize that information. And none of us here are qualified to properly vet his article.

I see no reason to believe that Mr. Deigin is better and applying genomics and gain-of-function research than Abramowicz is at applying Bayes theorem.

I’m definitely not taking his word over a Nature article that represents what seems to be a consensus among qualified researchers.

Here’s a Bayesian problem to solve. Which article is more likely to be correct? The peer-reviewed Nature article. Or the Medium post that implicates China that you found by looking through information sources that were looking for ways to implicate China?

1. Just FYI, you guys aren’t discussing what Loki linked, which was actually just a May 4 “news feature” article reviewing various characteristics of COVID 19, but not touching on the question of lab release beyond admitting “The origin of SARS-CoV-2 is still an open question.” Lol.

You guys are discussing “The proximal origin of SARS-CoV-2” which was published on March 17. I’ve read that article quite a few times, starting back before it was published when a draft was posted here: http://virological.org/t/the-proximal-origin-of-sars-cov-2/398

I’m not qualified to evaluate the scientific claims. I did find it interesting at the time of publication how much the article was changed from the draft. Unfortunately you can’t access the draft any more, but I noted at the time that the paper did not rule out “the possibility of a deliberate or inadvertent release” whereas “deliberate” was removed in final publication.

1. Yeah, I know there’s a ton of stuff from the scientific community, peer-reviewed and otherwise, that discounts the possibility of a lab release. I haven’t really bothered to check the specific stuff linked here.

As for that article I wouldn’t put too much in the removal of “deliberate”. I suspect it’s a combination of things.

First, “deliberate” implies intent, and intent was outside the scope of the article.

Second, there’s a fairly knowable list of ways an “inadvertent” release can happen, and that’s what they can analyze the in article. But to the extent a deliberate release is different from an inadvertent one you’re now talking specific schemes and plots and that’s not something they can really analyze.

Are they investigating a Chinese government plot? A disgruntled researcher? A US spy trying to discredit China? The moment you talk a deliberate plot you open up a pretty significant bag of worms with no real evidence for doing so.

1. Agreed, I’m not reading anything into the changes, I just remember finding them interesting but can’t recall any except that one.

The Nature article seems to be the seminal work. Is there another worth looking at?

If I understand it correctly, the Nature article finds evidence to doubt (a) that the observed characteristics of CoV2 were developed in vitro, and (b) that CoV2 was a product of design rather than selection. Is that right?

2. I’m not relying on Deigin’s authority.

I’m relying on the quality of his arguments. Deigin exhaustively references and quotes Shi Zengli’s own research, and discusses the methods of both her and other coronavirus researchers. From that evidence about actual methods in the field, I can conclude that the implied standard of the Nature paper that ‘all manipulated viruses are ideal’ is factually wrong.

I’m not saying that it was built in a lab, just that none of the evidence in the nature paper actually rules it out. And in fact, there’s no reference in proximity to those arguments ruling out lab manipulation to the methods being used by labs like Shi Zengli at all. A real argument about whether it could have been created in a lab would have looked at what those labs are actually doing, not some theoretical estimation based on SARS structure from almost 2 decades ago. (Which is what the referenced papers actually compare to).

1. Apologies for typo: Shi Zhengli

2. Actually you are relying on Deigin’s authority unless you have an advanced degree in microbiology you weren’t mentioning.

We just determined that a law prof doesn’t know how to use Bayes Theorem. Now you expect that we can properly vet a “64 minute read” medium post?

If there was a consensus that gain-of-function was likely, and Deigin was explaining why everyone thought that, then it would be a great educational article.

But as a “this is why the consensus narrative is wrong” article it’s useless.

I’ve seen things like this a dozen times before, really well researched, filled with smoking guns, looking like a slam dunk case to anyone who read them… only to turn out to be completely wrong because it relied on a bunch of faulty assumptions that a lay-person could not possibly know.

I see no reason to believe that Deigin is an exception to this rule.

1. 1. He doesn’t conclude ‘gain of function’ research was the likely origin. We know the Wuhan lab does Gain of Function research, they publish on it, but it’s impossible to decide just by looking at the virus if it’s the result of that or not, and a natural origin is easily as likely.

3. Microbiology? no. Evolutionary Biology? Yes. I’m competent to read the literature.

4. I’d note that the Nature article https://www.nature.com/articles/s41591-020-0820-9 is actually a correspondence, and wasn’t peer reviewed. Nor are any of the authors experts on Coronavirus or Gain of Function research. (Most of their interests are evolutionary biology of viruses, one of them does virus detection). They cite no paper regarding characteristics of manipulated viruses as part of their argument, and it’s not clear they have direct experience with such procedures. Deigin does cite (and quote, and copy figures out of) the gain of function literature, specifically with regard to coronaviruses. That evidence is very persuasive.

Which isn’t to say they shouldn’t make an argument if they have one; scientists don’t decide truth based on expertise, but on the quality of the argument provided. But the lack of citation to any relevant literature on coronavirus manipulation methods and characteristics isn’t a good look – they’re making a specific factual claim that is not backed up by any research they cite.

(Whereas I accept their argument that the furin binding site is a likely mutation in natural populations, because they cite research proving it is in influenza).

1. 1/4. I don’t know if there’s a peer-reviewed article arguing again an accidental lab release… that’s not really something that’s appropriate for peer review. But scientists with professional reputations on the line are making scientific arguments. I really don’t think Deigin falls into that category.

3. I have no idea what your formal education is but considering how you misapplied Bayes theorem (a different field admittedly) I’m honestly not inclined to trust your ability to digest the relevant literature either.

Overall I’m making a fairly simple calculation:
a) People who are qualified experts seem to be expressing the opinion, as much as it can be known, that the possibility of a lab release is not just “unsupported by the evidence” but to some extent refuted by it.

b) Consider if the first outbreak was close to a US embassy, would that be evidence of a bio-attack by China or the US? What about close to a University that had a researcher who studied a similar virus, or next to a military base (soliders as test subjects). There’s a lot of scenarios that would cause a similar suspicion.

c) There’s a lot of speculation about the first cases being in November or earlier, and not close to the market at all. If the first cases were elsewhere in Wuhan then this entire theory goes away.

Viruses crossing into humans from the wild is the Null hypothesis.

The proximity is enough of a reason that I want to know that qualified people are investigating. But not nearly enough to start declaring that the lab was “probably” or “likely” to blame.

1. 1/ The methods and products of virus manipulation are definitely suitable material for peer review. And in fact have been peer reviewed and published.

But the point of acknowledging it wasn’t peer reviewed is that no experts on coronavirus GoF research had any opportunity to provide feedback (and if there had been peer review, the editor would have been remiss in not insisting on at least one such reviewer).

The very idea that something isn’t suitable for peer review is silly on-face, btw.

And an argument is scientific when based on experimentation. An uncited asserted claim is not a scientific argument, even should it appear in a scientific publication. (Expertise in scientific argumentation means knowing what to cite, when, and why that citation proves the point you’re making, not that you get a free pass to assert claims without evidence).

3/ Citation most definitely needed on me misapplying Bayes.

a/ Then they should cite the basis of their claims that it being created in a lab is refuted by the evidence. Their say-so in a scientific publication (even a correspondence) isn’t good enough. Especially when Shi Zhengli’s lab (and others) has published papers that very much document methodology and outcomes which are apparently impossible, according to the authors of that Nature correspondence. (The only cited papers in the vicinity of the claim are about SARS spike protein function as SARS existed back in 2002-3 – they aren’t even about Gain of Function research).

Also, they don’t seem to be qualified experts on manipulating virus genomes. Experts on virus phylogeny? Sure. An expert on diagnostic testing? Sure. These aren’t the same things. And if they were experts, they still should have cited relevant papers to back up their claim.

b/ Let’s not confuse ‘evidence of’ with ‘proof of’. Some of these scenarios might constitute weak evidence of a biological attack, but that wouldn’t be sufficient to move the strong prior against a biological attack happening, unless there was sufficient other evidence to lead to that conclusion.

If you’re familiar with Bayes, you knew that already. (A fact is evidence for an hypothesis if it is more expected on that hypothesis than on any other hypothesis. How good of evidence it is depends on how unexpected it is given background knowledge, and how strongly it favors the hypothesis which best predicts it).

c/ I agree. The identity of patient-0 would be important evidence that could dramatically shift the range of viable scenarios. But it’s not evidence we currently have. We can only deal with facts we actually have.

I’m not saying the lab was ‘probably’ or ‘likely’ to blame, just that it’s ‘possibly’ to blame, and can’t currently be ruled out on available evidence. (The Wuhan market, on the other hand, seems to lack real evidence supporting it. But those aren’t the only two possibilities).

1. “I’m not saying the lab was ‘probably’ or ‘likely’ to blame, just that it’s ‘possibly’ to blame, and can’t currently be ruled out on available evidence. (The Wuhan market, on the other hand, seems to lack real evidence supporting it. But those aren’t the only two possibilities).”

To point out something that is beyond obvious, in the plethora of pointless verbiage that you spew; sure, anything is possible, even aliens, or that this all a false flag operation by the Trump Administration to cancel the elections.

But only a moron keeps beating the drum of things that are not excluded to claim that it’s the answer.

2. 1. An article on where the virus may of come from in the wild is appropriate as a peer reviewed article (I think these have been published). An article trying to debunk a lab release is not because you’re suddenly having to game theory things like cover-ups.

But the Nature commentary is relying on peer-reviewed research and is in arguably the most prestigious journal.

3. Half this thread was you defending the laughably awful application of the Bayes theorem. I actually went through the trouble of coding up the math so you could run the numbers yourself and see just how nonsensical the whole exercise was. That seems to have stopped you from defending this calculation in particular, but somehow you missed the point that Bayes theorem is entirely unsuited to analysis of this nature.

a/ And if they were experts on virus genomes you’d be asking for experts on virus phylogeny to debunk some other Medium post instead.

Don’t you wonder why no reputable researchers seem to agree with your interpretation of the evidence?

1. 1/ But they don’t cite any relevant peer-reviewed research for the critical claim. I’ve stated this explicitly several times, so claiming “the Nature commentary is relying on peer-reviewed research” comically misses the point.

https://www.nature.com/articles/nm.3985 is probably a paper they should be citing. It demonstrates the production of a synthetic coronavirus which is a chimaera of two different coronaviruses, and that nature correspondence’s basis of rejecting a lab origin for covid-19 would also reject a lab origin for this explicitly lab-produced virus. Some of the authors are also from the very lab in Wuhan.

3/ I recommend you look into Bayesian epistemology. It’s entirely suited to questions of belief like this.

And I’ve never explicitly defended Abramowicz’s specific calculation. But the correct way to object to the argument is to argue his priors and/or his evidence. Attacking the general method is a fallacy, and the general method is what I’ve defended.

a/ I’m not the one insisting on deferring to expert opinion solely because they’re experts. I’m just pointing out they’re not actually experts on the field and methods where you want to rely on their uncited claim.

They could be particle physicists for all I care, so long as they cited the appropriate literature, and that literature actually backed up their claims.

30. Try another way to think about this. To begin, assume the pandemic did start in or near Wuhan. Now consider a worker in the biology lab, who shows up with Covid-19. Does that make it obvious, or even likely, that the lab was the source of the worker’s infection?

What else could it be? It could be anything. There are millions of possible explanations for why anyone in Wuhan, including the infected lab worker, might be an early case. Assuming the worker contracted the virus outside the lab, he would be far more likely at the lab to get early discovery and diagnosis then a typical Wuhan resident would be to get diagnosed anywhere else. Everywhere else, almost all the early cases would be missed. Only the lab would be actively on the lookout for cases, and probably only for cases associated with the lab.

In effect, there would be a powerful bias for noticing a case at the lab, and against noticing a case anywhere else. Many people in or near Wuhan might have caught the virus, transmitted it, got sick, and even died without anyone being diagnosed with coronavirus, which no one was looking for—except at the lab.

During the early stages of the outbreak, only at the lab would an early case be likely to get prompt notice. So in pandemic time, a fairly extensive outbreak could have been sustained for many generations, and gone unnoticed, until a lab employee fell ill.

What would make it unlikely the worker contracted the virus at the lab? Almost everything. An indispensable method of biohazard laboratory management is to keep workers from getting infected. Precautions are extensive. At the lab, and only at the lab—among all other possible venues around Wuhan—is there any security against infection. Everywhere else, no one is protected at all— including the infected worker, whenever he went anywhere else.

So how does that affect the probabilities? The same way that betting a long-shot horse to win, vs. betting the field affects the probabilities. Almost always, it will be the bet on the field which pays off. Given a presumption of localized coronavirus transmission, even a biohazard laboratory worker has far more opportunities outside work to catch the virus than he does to catch the virus at work. And if that worker does catch the virus outside work, it is far more likely to be noticed at work than otherwise.

31. You shouldn’t be conditioning on it being in the neighborhood of the WDCP. That’s only one disease lab but there are many many disease labs. Also you can’t select the distance you are using (eg neighborhood vs. same city etc…) only after you observe the event. Indeed, more generally it’s wrong to single out this particular event since there are other similar things that could have happened which would have seemed equally surprising. Actually figuring out how to do this kind of computation is surprisingly difficult and must include all the evidence about the probability that such a virus would even be created by such a lab and how likely it is that it would escape. It’s far to easy to forget that it’s fallacious to merely look at the data which pushes one way.

I’m not sure I agree it’s even helpful here to compute the probability explicitly like this.