Karl Friston on “immunological dark matter”

We’ve been comparing the UK and Germany to try to explain the comparatively low fatality rates in Germany. The answers are sometimes counterintuitive. For example, it looks as if the low German fatality rate is not due to their superior testing capacity, but rather to the fact that the average German is less likely to get infected and die than the average Brit. Why? There are various possible explanations, but one that looks increasingly likely is that Germany has more immunological “dark matter” – people who are impervious to infection, perhaps because they are geographically isolated or have some kind of natural resistance. This is like dark matter in the universe: we can’t see it, but we know it must be there to account for what we can see. Knowing it exists is useful for our preparations for any second wave, because it suggests that targeted testing of those at high risk of exposure to Covid-19 might be a better approach than non-selective testing of the whole population.

Speculative, and here is the full article, mostly about other matters epidemiological.  Via Michelle Dawson.

The contents of this article may well be wrong, as many on Twitter are suggesting, that link makes two renowned researchers/commentators or here is Kai Kupferschmidt.  I am less impressed by passing potshots at the author and more interested in reading a short blog post.  I say write out 500 words (with links) comparing different methods of accounting for the unexplained residuals and heterogeneities with Covid-19 (not it is not all policy, and how much of it is sheer luck and/or timing?).  Also discuss the circulating notion that previous exposure to (some) other coronaviruses, or perhaps weak doses of Covid-19, might be giving some individuals partial immunity (an idea being batted around but not verified as far as I know.  And what do we know about how much geographic isolation might matter (a concept cited by Friston in contradistinction to the Twitter claims that he is simply talking about ghosts)?  Then tweet the post, you still can start a blog for free and write just a single post, more over time if you wish.  I would very gladly link to it.


Seems plausible. We don't yet know why women and (possibly) smokers seem to be somewhat protected. Any number of genetic or lifestyle factors could vary between populations and have a similar impact on morbidity and mortality. We know so little about this disease. My main hope is that we don't find out that a large percentage of people infected have serious or ongoing sequelae.

Only plausible if you ignore the fact that the difference between Germany, the UK, and Italy is very unlikely to be due to the fact that Germany has more immunological “dark matter”. The regional differences in Germany itself in terms of infection rates is very noticeable, in part due to Carnival being celebrated in Catholic regions, not Lutheran ones. Leading to the natural idea that the virus make a distinction based on faith.

The European financial crisis - how long ago that seems now - broke down roughly along civilizational lines. The Protestant North was mostly doing alright while the feckless Latin (and Greek Orthodox) South was its usual feckless, dishonest, corrupt self.

So it seems that Covid-19 is also breaking down along confessional lines. As Max Weber might well have predicted.

My money would be on the infection being associated with those that do not practice particularly good personal hygiene. Uptight Protestants may well have that going for them.

It is the Catholic south of Germany that is currently the economic powerhouse of the German economy.

And then there's the fact that the Germans don't do that cheek-kissing thing.... unlike the French Spaniards and Italians.
Swedes don't do it either.

Neither do the UKers.

Maybe there's some other explanation for that.

Doesn't the data for men and women show an equal likelihood of infection but a greater mortality rate? This guy is suggesting that Germans are less likely to get it period. Seems unlikely, but the answer will most likely be found using Swiss data. I'm going to guess a mix of better public health response and people following guidelines more.

Also, this is nothing like dark matter. While the concept in physics is scientifically dubious, scientist have at least some measurement of dark matter and models that incorporate dark matter into predictions of physical phenomena. All this guy is saying is that one group had more infections than another and he doesn't think we know why.

Also, I don't want to put boundaries around who can same something about another field, but damn if Tyler didn't link to a neuroscientist talking about dark matter in epidemiology. And for the love of God, if I see someone else basically asserting that SIR/SEIR models are the only ones used in epidemiology, I'm gonna direct the rioters to their office.

Tirol is Italian and Austrian, and borders Bavaria and Lombardy, the two regions with the worst outbreaks in their respective country. Any dark matter explanation for the difference between Germany and Italy would require a mechanism to explain the difference between Munich and Milan.

And provides another perfect moment to repeat this - Fix this now. I mean you!

How is Germany isolated - when it shares a border with 9 other countries?

Americans and their „ piss ant knowledge „ of maps. Anyways, I would describe most of the low numbers here in Germany to luck. The fact that Italy was affected 1st & severely allowed Germany to save itself. And the blind luck no large gatherings were happening in Feb.

This is not scientific but older more at risk populace of Germany lived through WWII, dictatorship, Russian invasion in the East, Anglo-Saxon in West. Then 40ish years of being a split country.

Perhaps ^^ those factors could play a factor in mental dexterity or due to historical presidency - Germany was spared a break. Also Germany got money into its citizens hands in early April. That couldn’t have hurt - financial stress does lead to more health complications.

Okay okay enough ranting


"Americans and their „ piss ant knowledge „ of maps. "

The author is British.

I don’t believe this theory. The testing in Gangrlt showed that plenty of Germans could get infected.
I think it’s probably due to the methodology in isolating the vulnerable like people in nursing homes. That’s what makes a big difference to fatality rates

* the testing in Gangelt

I don't really believe it either but in defense would suggest it doesn't need to be a general factor among who German population. You only would "need" this "immunological dark matter" to play a significant role at entry point to suppress early growth (late Feb-early March growth) when case totals small.

Because they don't slobber all over each other's faces like the Italians do.

You don’t know anything about Germans. In the village they did seriological tests a month or so ago they had an event where it was common that people kiss each other on the mouth. Do your research so you don’t look like a dumbass.

Except the article was comparing the UK fatality rates to the German ones. Though considering the level of discourse here recently, maybe you think the British slobber all over each other's faces.

Not in public please

Culture matters? Two Belgian researchers find that differences in the way people, and in particular different age groups, interact can explain part of this variation. Based on interaction matrices for Germany, Italy, Belgium.
They also mention another study, that attributes part of the intra-European differences to genetic variations.
Blogpost (in Dutch): https://reluctanteconomist.com/corona/ciao-of-guten-tag/

Link to the original article:

Here's a piece I wrote:



"Sweden’s permissive approach for society as a whole is the right one" - sounds a rather bizarre statement to me, since it's a losing strategy. I would think that *winning* is the right strategy, if w're going to argue for a strategy. Still, I guess, if we have to lose, we might as well lose with style.

However, as noted months ago by Angela Merkel, most people are going to be ended up infected by this virus in the foreseeable future regardless. The Swedish model is a response to a non-controversial and very mainstream perspective. It is also a longer term strategy, in a sense - comparing Finland, Denmark, and Sweden in a year is better than making a comparison today.

Yes, it is a losing strategy, based on the idea that winning is not possible in the first place. No wonder so many Americans are enamored by such a model.

" most people are going to be ended up infected by this virus in the foreseeable future regardless"
This is, of course, not what Merkel said.

It was good that the people fighting the original SARS were not as dumb as the morons who today are suggesting herd immunity for a respiratory virus for which the concept of herd immunity is inapplicable because it doesn't induce long-term immunity.

„60 bis 70 Prozent der Bevölkerung könnten infiziert werden“

Assuming no vaccine or treatment, of course, as she said in her speech to the nation at the beginning of March.

Now, if only you were smart enough to know the difference between "could" and "will".

The original SARS did not have asymptomatic spread.

Yes, and?
The fact that controlling something is harder doesn't mean it's impossible.
Neither does it mean we should give up trying.

Seems like a weird way to say "we don't know"

It means the exact opposite. They clearly said:

"This is like dark matter in the universe: we can’t see it, but we know it must be there to account for what we can see. "

Or just another example of excuse seeking in a country where the pandemic led to a high number of deaths compared to Finland, Greece, Germany, etc. Want to guess the 'dark matter' involved? Using well tested public health measures to handle a pandemic.

Ill be more impressed by his work if he starts explaining the difference between more countries without having to invent novel hypotheses for each country.

The main differences are luck/kismet/fate/etc., timing (as distinct from luck - Finland or Sweden had longer to prepare after seeing what was happening in Italy and Spain), and a desperate desire to find an excuse for the public failing of a nation's health authorities.

Germans aren't more isolated. Average population density is higher than in the UK. Even London isn't that dense, it has less than 6k people per sq. km.
Some isolated regions in East Germany were hit relatively less hard though, but that is the same for low density regions in the UK (Wales...).
You'll have to distinguish between infections and fatalities here. On infections, German regional governments dithered less. Social distancing was complete by mid March, brutally enforced by rigteous citizens. I know of someone who was heavily insulted on his commute just because he didn't wear a mask. UK state capacity and stability has fallen recently anyway, with all their Brexit-related petty struggles.
On deaths, it's probably mostly good isolation of elderly care homes. The three year required apprenticeship for care workers - I am told the exams are quite hard to pass for the non-attentive - helped keeping out the unqualified, and having well paid well trained workers who don't superspread.
Better hospital infrastructure may have contributed. Germany's system of competing health insurers and partly private hospital environment open to some competition may have led to less complacency and fewer jaded workers than at the NHS.

> Germans aren't more isolated. Average population density is higher than in the UK

Germany population density: 230 / km^2
UK: 272 / km^2
England: 432 / km^2

The only sizeable German state denser than England as a whole (population: 56M) is North Rhine-Westphalia (poulation: 18M; density: 525 / km^2). The other denser states are all small city-states (Berlin, Hamburg and Bremen).

Isn't "dark-matter" an euphemism for "too expensive for proper research"?

It is more along the lines of observable reality and our models not matching, so time for a bit of phlogiston or ether to be sprinkled like fairy dust until someone comes up with a way to explain what is going on.

No -but it could be code for not blaming COVID for deaths d/t the infection but not from ARDS. As an example- if a patient with COVID pneumonia dies on from an MI- how you write the death certificate will greatly affect the COVID death rate.

The proportion of the German population that has (smoking and smokeless) nicotine intake is approximately 33.1%. That for the UK is approximately 17.7%

Nicotine takers are known to have a reduced Hazard Ratio of death from Covid-19; this from a recent UK study based on 17 million health records and over 5,000 deaths with/without Covid-19 linking.

This information supports the view above of commenter anonromcomdotcom: "and (possibly) smokers seem to be somewhat protected".

Keep safe and best regards

Maybe smokers breath smells gross so people don't want to lean in and get too close to their face.

When there was an outbreak of "Legionnaires' disease" in the 70s, the cause was identified "Legionella exist naturally in water and moist soil. They have been found in creeks and ponds, hot and cold water taps, hot water tanks, water in air conditioning cooling towers and evaporative condensers, and soil at excavation sites." "An outbreak of this disease in Philadelphia in 1976, largely among people attending a state convention of the American Legion, led to the name "Legionnaires' disease. Subsequently, the bacterium causing the illness was named Legionella pneumophila and the name of the illness was changed to Legionellosis." The symptoms are equivalent to Covid.
"Is this a new disease?
No. The bacterium was first identified in 1976, but earlier cases have been confirmed as far back as 1947."

The Mayo Clinic site is also worth a read.

Perhaps, dark matter is another name for unseen bacteria or virus in certain places or contact with water (rather than beer) from certain sources or personal habits not easy to identify.

"Neuroscientist Karl Friston, of University College London, builds mathematical models of human brain function. Lately, he’s been applying his modelling to Covid-19, and using what he learns to advise Independent Sage, the committee set up as an alternative to the UK government’s official pandemic advice body, the Scientific Advisory Group for Emergencies (Sage)."

Friston: "Conventional models essentially fit curves to historical data and then extrapolate those curves into the future. They look at the surface of the phenomenon – the observable part, or data. Our approach, which borrows from physics and in particular the work of Richard Feynman, goes under the bonnet. It attempts to capture the mathematical structure of the phenomenon – in this case, the pandemic – and to understand the causes of what is observed. Since we don’t know all the causes, we have to infer them. But that inference, and implicit uncertainty, is built into the models. That’s why we call them generative models, because they contain everything you need to know to generate the data. As more data comes in, you adjust your beliefs about the causes, until your model simulates the data as accurately and as simply as possible."

Maybe Cowen quoted the "immunological dark matter" part of the interview of Friston because it captures just how ridiculous "modeling" has become. Yes, the issue here is the misuse of "modeling" to draw real or, more likely, imagined inferences, an offense sometimes committed by economists, sociologists, and many others. The inference drawn about Germans and "immunological dark matter" is of a piece with inferences drawn about Jews and blacks. This isn't "research", it's intellectual fraud.

US had a 0% fatality rate for SARS1 back in 2003. 27 cases and zero deaths, whilst Canada experienced a fatality rate of 17.5%.

Current Germany fatality rate of 5% for SARS2 is hardly “low”... South Korea is still under 3% I believe, and Singapore below that if I recall correctly.

The tone of Carl Bergstrom's posts has made it clear for a while that he is just as ideologically committed to his own position as he accuses others of being, I suspect for the usual reason of attacking the current UK govt. Two things: first, whatever the headline writer chose, Friston is proposing something most people recognize by now, i.e. some kind of pre-existing resistance, either innate or from previous exposure to other coronaviruses. Therefore, when we were arguing over the IFR, we should maybe have been arguing over some kind of "exposure fatality rate", as some people, especially children, do not become infected even after close contact with people who are sick. For herd immunity purposes, it is the non-susceptible proportion that matters, not whether the susceptibility is due to covid antibody or some other source.

Second, because Friston's initial modeling was theory-light, but used good assumptions about processes and susceptibility proportions, he managed to write the below paragraph in early April (the date on the linked paper is 7th May, but it was on arxiv on 9th April):

"reports of new cases in London are expected to peak on April 5, followed by a peak in death rates around April 10 (Good Friday). At this time, critical care unit occupancy should peak, approaching—but not exceeding—capacity, based on current predictions and resource availability. At the peak of death rates, the proportion of people infected (in London) is expected to be about 32%, which should then be surpassed by the proportion of people who are immune at this time. Improvements should be seen by May 8, shortly after the May bank holiday, when social distancing will be relaxed. At this time herd immunity should have risen to about 80%, about 12% of London's population will have been tested. Just under half of those tested will be positive. By June 12, death rates should have fallen to low levels with over 90% of people being immune and social distancing will no longer be a feature of daily life.”

Bergstrom seems like someone burdened with some heavy mental issues. I feel for the guy, he’s not all there and nearly everything he’s stated has been proven wrong. He’s like that white knight guy that sent those secret messages to this blog about Japan’s “worst crisis since the war.” Bergstrom ripped on Micheal Levitt a lot and then deleted all those tweets— that was pretty much confirmation that he’s just a twitterati put for attention and not an adult. He’s not someone to pay attention to other than for the Lols, which there are many every day.

On this topic you could link to this article which takes a careful look at the Australian and South Korean case fatality rates, both of which are way lower than those seen in US or Europe. Australia CFR is estimated at 1.4% and Korea at 2.4%. Rules out age demographics as explanation. Possible explanations are higher rates of antibiotic resistance ( many patients ultimately die of sepsis) or perhaps the older generations in Korea carry consequences of being children during the time of the Korean War.

Colorado has finished a 56K sample-size antibody test. Nearly 8% of the state has been infected, or nearly 500K people. Divided by death, you get 0.23% IFR.

For those under 60, the IFR is 0.03%. For those under 70, the IFR is 0.07%

We've also recently learned that many are completely immune to the virus because of another corona virus they've had in the last few years. That could be the mechanism that is protecting most kids.

More interesting is that when you take blood from random people, and then introduce the virus into that blood, in 33% of the random samples the blood goes to work fighting the virus EVEN THOUGH the blood didn't have detectable levels of antibodies.

In short, the worst is very likely over. If after all this rioting the hospitals aren't overwhelmed, then that is the end of the virus. Back to work.

*More interesting is that when you take blood from random people, and then introduce the virus into that blood, in 33% of the random samples the blood goes to work fighting the virus EVEN THOUGH the blood didn't have detectable levels of antibodies. *


Link below. They claim that people have a range of antibodies that appear to "neutralize" the sar-cov-2. These antibodies latch onto the virus and can block the virus from entering cells.

They claim that everyone's body can generate these responses in varying degrees even though it doesn't generate a large (or measurable) antibody response. They identified "elite" responders that produce a high number of neutralizing antibodies. And from that, they hope to find the B cells that make them and then clone the antibodies from those cells. Thus, we could all appear to be "elite" responders with a bit of help is their thinking.


It's an antibody test, so those are mostly false positives and you can't read anything into it.

"If after all this rioting the hospitals aren't overwhelmed, then that is the end of the virus."

This, however, is insightful. The rioters/looters are a decent test of reopening. If we don't have a clear surge in a month or so, that's a very positive sign.

That was meant to be in-reply-to phinton @10:25, just above.

*More interesting is that when you take blood from random people, and then introduce the virus into that blood, in 33% of the random samples the blood goes to work fighting the virus EVEN THOUGH the blood didn't have detectable levels of antibodies. *


> It's an antibody test, so those are mostly false positives and you can't read anything into it.

Antibody tests today have specificity of 99 to 100%. Colorado measured nearly 8%. That's not "mostly false positives". Sensitivity are around 90%. So, roughly 10X more likely to report a false negative than false positive.

With our lack of knowledge and lack of even just decent data , it's too early to throw up our hands and resort to dark matter explanations.

Over the last couple of months there've been easily a couple of dozen explanations -- with supporting but highly incomplete data -- for why some group has a higher infection rate or higher fatality rate than some other group.

We haven't come close to finding an explanation yet (except for age and co-morbidities), and I'd say we need at least a couple of years of data before we can think that we've exhausted the research avenues.

And even two years might not be enough; how many successful studies on panel data can you think of, that had only two years of observations?

Or think of the Phillips Curve, which turned out to be wrong but at least those researchers were looking at a couple of decades of data. It took additional years to prove the lack of validity of the notion of a fixed Phillips curve.

Viruses move faster than economies do so I'm not saying we have to wait a couple of decades to get definitive research results on Covid-19. But we aren't close to having enough data -- reliable and accurate data -- to do a good research job.

Imagine if economists said we don't know why Taiwan's economy has grown more quickly than Vietnam's. It must be economic dark matter!

nobody should care what i say all that much, but these last few months have taught me that Greg Cochran is a foolish old man, Nassim Taleb is dishonest and afflicted much more than he thinks he is with intellectual vanity (that happens a lot to recreational mathematicians): these last few months have also demonstrated that I underestimated Tabarrok and Cowen, and that poor guy who blogs at 538 has upped his game (I did not underestimate the 538 guy, I am just saying he upped his game).

An awful lot of older intellectuals disgraced themselves in the last few months, typically intellectuals who had a fantasy world in which they believed they understood mathematics and statistics and nobody else did. Well guess what lots of them were wrong. My best guess is that it was a mixture of a senile desire to be, in one's old age, the 4 star general one had imagined oneself to be many years before, in an unexpected struggle, as well as a side effect of reading on the internet the twitter feeds of people who agreed with one's point of view, and like an ignorant young person thinking that the compliments in one's echo chamber was proof of one's preeminence ----- and God knows what else it is that makes people into dangerous imbeciles simply because they are ARROGANTLY AWARE that they know a little more calculus than the average bear.

Other challenges await us. Do better next time.

and yes I said was instead of were, on that seventh line. i did that on purpose.

trust me, I know how to express what I want to say.

or don't trust me, tell me I don't, amuse me.

I am more a "linear algebra" guy than a "calculus" guy, but trust me, I know what my limits are, and I know the limits of the charlatans who know less about one or the other than I do.

you have no idea how much contempt I have for the phonies.

I mean, I like them as human beings, but you do not want to know what i know about how dishonest they are. think about it (and remember i was just kidding, i do not feel contempt for anyone, God loves us all, including you,,,, don't ever forget that)

That being said, there is nothing simpler in this world - and I say this as someone who knows a lot about linear algebra, and an awful lot about the fallacies and absurdities of 20th century academic philosophy (sad words, when the fact is we ALL KNOW GOD LOVES US, philosophy is a silly little hobby in comparison - trust me, if you had ever been fortunate enough to read my explanations of where the "phenomenologists had gone wrong, of course

the funny thing is this, the important thing is not whether God loves us or not, the important thing is this ====DO WE CARE

Seriously, you don't want to know what I know about the HUGE DIFFERENCE between people who care and people so do not care,

Do not try and understand..

The "dark matter" is named "timely reaction". Here's the graph that explains it all pretty well


You might want to read Maxwell Ramstead's defence of Friston on Twitter: https://twitter.com/mjdramstead/status/1267789052322942977 or even Friston and Deenan Pillay's more in-depth explanation of what he meant: https://medium.com/@karlfriston/immunological-dark-matter-b48e20bba9ea

Comments for this post are closed