Saturday, May 02, 2020

Tunnel vision

From a Facebook exchange

Hays
I've been skimming your posts on the pandemic, so there's no doubt much that I've missed. Issues you already addressed.

i) On the one hand there's the danger of knee-jerk suspicion and popular rejection of expert opinion. A conspiratorial mentality.

ii) On the other hand are some opposite extremes. An occupational hazard of expertise is tunnel vision. For instance, I used to watch an ER special about patients with urgent life-threatening conditions, but they were dangerous misdiagnosed because the specialization of the physician blinded him to possibilities outside his field of expertise. It was only when the patient was referred to different specialist that his life was saved through a correct diagnosis.

iii) By the same token, a professional mindset can disarm critical judgment as experts in one field of specialization automatically defer to experts in a different field of specialization. 

iv) In the USA, the dominant paradigm for dealing with the pandemic is containment. Minimizing the infection rate. Shutting down infection vectors.

Experts who operate with that paradigm may pay lip-service to concerns about the economic collateral damage or indefinite suspension of civil liberties, but so long as they regard the pandemic as an existential threat, their acknowledgement of other harms is a throwaway concession. They think the pandemic must be contained by any means necessary, whatever the cost in terms of economic collapse and the abridgment of civil liberties.

v) I'm also struck by the fact that we have two competing paradigms: Containment policies aim to minimize the transmission rate while herd immunity policies aim to maximize the transmission rate. My point is not to assess their respective merits. It's premature to judge their comparitive success, failure, or tradeoffs. 

But it's very strange to have two competing paradigms championed by experts that mandate opposite social policies. 

vi) We also know that the same policies have disparate impact in different localities depending on other variables like population density, mean age of population, bad health habits (e.g. chain smoking). 

vii) It doesn't help their credibility that public health experts have drastically downgraded their doomsday scenarios.


Jeremy  
Which public health experts have drastically downgraded their doomsday scenarios? Do you mean the people who said that if we take these measures we will have far fewer deaths who then said several weeks later that we have lowered the expected death count by doing those measures, and then people went viral with fake news pretending that they changed any of what they had said? There has been a huge amount of that sort of thing. Not so much of anyone actually changing their estimates.

Hays 
To take one example, I was alluding to this presentation: 


He says the worst-case scenarios are being revised downwards, and that's because we overreacted. I don't see how that follows:

i) Were the worst-case scenarios based on doing nothing? Did the worst-case scenarios not factor in ways to counteract the pandemic? 

ii) Even if (i) is the case, has do-nothing herd mentality policies led to a worst-case scenario in terms of the death toll? 

ii) Different policies have been implemented. Surely attributing the downward revision of the worst-case scenarios is not independent of the policy. The question is whether some polices are more efficacious than others. Just doing something isn't what lowers the worst-case scenario. It has to be something that makes a difference. Something that's effective to a certain degree.

Surely we don't wish to promote a circular confirmation theory where overreacting is its own justification, and if something appears to work, that's because we overreacted. There are multiple variables, and the we don't have the counterfactual as a point of contrast to test it against. 

iii) More to the point, different localities have been using the same policies with dramatically different results. So how much is attributable to the policies?

As we know, there can be other factors like population density, medium age of the population, high rates of tobacco consumption.

iv) Another thing I had in mind is early projections from the Imperial College, where if we carry on as normal, 2.2 million Americans will die, if mitigation measures are taken, 1 million Americans will die, and if suppression measures are taken, 100,000 Americans will die.

Jeremy
I don't think it's circular reasoning to make a prediction, apply a policy that tests that prediction, and discover that what you predicted happened. Maybe there are other variables to test to see if other factors are at play, but it's not circular reasoning to observe that what you predicted is what happened. That's part of the scientific method, even if it's not all of it. You can't do all the steps at once, and sometimes doing proper controls is immoral, because we're not doing studies where we deliberately infect people or try to achieve disastrous results. We have to go with the best evidence we can get. But no scientific argument is a perfect deductive airtight piece of reasoning. Not having ruled out every possible variable is not the same as being circular.

The ridiculous argument I keep seeing, e.g. from Bill Bennett, is that because the non-restrictive scenario predictions didn't happen it must not be all that bad, so we shouldn't have implemented the measures that kept it from being bad. That's worse than circular. It's contradictory. As a friend of mine said, it's like taking the parachute off halfway down because it's kept you alive so far.


Hays
How is it a ridiculous argument if the worst-case projections for less restrictive measures turned out to be wildly inflated projections? 

And bad compared to what? Bad to compared to the predictions?

Again, if the predictions were exaggerated, then what's the warrant for saying it's the more drastic measures that keep it from being worse–if less restrictive measures didn't have the dire consequences? What's the empirical standard of comparison?

Moreover, I don't agree that if it's perceived to work, then that ipso facto justifies overreaction. Some cancer patients die, not from cancer, but from side-effects of cancer therapy. We could eliminate death from side-effects of cancer therapy by eliminating cancer therapy. That would work. But it would hardly be legitimate to use that result as circular justification for banning cancer therapy because the overreaction worked. Yes, it worked, but at what cost.

Furthermore, there's still the question of what works. That's a comparative assessment. Do lockdowns and mass house arrest work? Compared to what? What about testing, tracing, and quarantine for the actually infected?

How well does containment work if most of the infected population has mild symptoms that don't require treatment? 

Does that work better than testing, tracing, treatment (if necessary) and quarantine for those develop a life-threatening condition? 

How does treating everyone the same whether their infected, asymptomatic, or able-bodied even if infected represent the best strategy?

And what about the herd immunity alternative? I'm not suggesting that's a silver bullet. That has tradeoffs. But what if, until the development of a vaccine, herd immunity is the primary protection against recurring pandemics? Don't containment policies impede herd immunity? Should we take drastic measures to impede herd immunity in the population?
  
Jeremy
Where are you getting evidence for inflated predictions if there isn't anywhere that has implemented the no-restriction scenario that those predictions were dealing with? Are you saying Bennett's argument is not as ridiculous as it obviously seems (i.e. the parachute example) because in some scenario where no one implements any restrictions hardly anyone still dies? Where is the evidence for that? We have clear evidence of much higher infection rates and much higher death rates for places that have had less restriction or have delayed putting restrictions in place well past the time they needed to if they were to avoid the high rates. We have no evidence of a place with no restrictions that has similarly low levels as what happens with restrictions. Do you just get to make data up if you don't have it?

Hays 
Many countries have avoided lockdowns/curfews. Some focused on testing/tracing, treatment (if necessary) and selective quarantines for the seriously inflected.

Then you have the less restrictive measures of some Scandinavian countries. Admittedly, that has yet to be fully tested, and they initial spike, but that's part of the strategy. The question is where they plateau.

It's also the case, as I noted before, that localities using the same containment policies have dramatically disparate results, so you can't chalk it all up to the policies.

I'm not the one who brought up Bennett's argument. Don't impute to me an argument I didn't use, then proceed to use that as the substitute for my own stated position. 

Actually, from what I've read, there's something of an inverse rational between infection rates and fatality rates. In some countries a high perceptive of the population appears to be or have been infected, yet the fatality rate is fairly low, which suggests the fatality rate is much lower than predicted if you have higher percentages of infected but proportionality much lower death tolls. That dilutes the correlation. 

From what I've read, the high fatalities usually involve high-risk groups due to age and comorbidities. 

Jeremy
OK, but having testing/tracing or selective quarantines is not the same as doing nothing, which is what the higher numbers were projecting. So again, why think there is any data showing that those numbers were too high? No one tried doing nothing, so we don't have any data to contradict those numbers. Then you say I'm not being a grownup for pointing out that there is no data for such a case. You can't claim the 2.2M deaths figure is exaggerated, because we have no data of cases where we did nothing. That's the only way to support such a claim.

As for Bennett, you quoted my statement of Bennett's argument and then proceeded to argue against how I critiqued it. You questioned my claim that it is ridiculous. How is it inaccurate to characterize that as defending Bennett's argument?

Hays
No, there were different projections based on whether we carry on as normal, enact mitigation measures, or go to suppression measures. BTW, have these turned out to be accurate?

Jeremy
So again, why think there is any data showing that those numbers were too high? No one tried doing nothing, so we don't have any data to contradict those numbers.

Hays
The herd immunity experiment in Scandinavia. Admittedly, that has yet to run its course, but the same can be said for containment policies. 

Jeremy
Then you say I'm not being a grownup for pointing out that there is no data for such a case. You can't claim the 2.2M deaths figure is exaggerated, because we have no data of cases where we did nothing. That's the only way to support such a claim.

Hays
Adjusted for population, is that's what's happening in Scandinavia? 

Jeremy
As for Bennett, you quoted my statement of Bennett's argument and then proceeded to argue against how I critiqued it. You questioned my claim that it is ridiculous. How is it inaccurate to characterize that as defending Bennett's argument?

Hays
Bennett is not my spokesman. I never made his argument my frame of reference. I presented my own arguments and examples.
  
One side of the argument:


Of course, he has his own bias, and I realize one can find an expert to champion any side of any cause, but this should be part of the scientific discussion rather than having public policy stampeded by a lockdown/mass house arrest model as the only response that merits consideration.

1 comment:

  1. When it comes to COVID-19, smoking is not a bad habit but a good habit. So say the French:
    https://www.dailymail.co.uk/health/article-8246939/French-researchers-plan-nicotine-patches-coronavirus-patients-frontline-workers.html?ito=email_share_article-masthead

    As for group differences, my money is on Vitamin D deficiency as a likelier suspect.

    ReplyDelete