"In light of the ignorance of typical political leaders and members of the general public, we might be tempted by the idea of rule by experts, as in Plato’s Republic. 18 Unfortunately, when it comes to descriptive social theory, even the experts’ knowledge is unimpressive, as demonstrated recently by the social psychologist Phillip Tetlock. Tetlock conducted a fifteen-year study in which he collected tens of thousands of predictions from hundreds of political experts concerning matters within their areas of expertise (for example, would the economy slide into recession, would the Soviet Union survive, who would win the next Presidential election, and so on). Tetlock’s finding, in brief, was that the best experts did only slightly better than chance at predicting outcomes. When asked to assign probabilities to their predictions, experts proved systematically overconfident; for example, events predicted with 100% confidence happened less than 80% of the time."
A: We're monkeys who more or less only *just* upgraded from feudalism. Things are going well considering we just need to keep iterating
B: Decisions get made that so clearly go against everyone's best interest on the regular. We can glimpse a better world and what we've build (while keeping us safe) is blocking us from moving forward.
I think the more fully we can recognise both the better, but A is a lot less represented in the current discourse, thanks for the reminder.
onically, with the WebMD example, if you want the most precise I formation without political/legal intfluence you would choose an even more specialist (or professional) resource, such as the BNF (for drugs in the UK, which definitely doesn't list all side effects; I find it odd when a patient asks me about a side effect of a drug that I've prescribed for them that they've read in the leaflet yet the side effect is not significant enough to warrant listing in the British National Formulary so when I look it up I can't find it), fpnotebook, CKS etc
So? Is it true that disproportionately many borderline personality cases are young women with lots of piercings and tattoos? What does that tell us about diagnosis and etiology?
Yes! Keeping power vs. being right is a great constraint to point out. I would maybe suggest expanding "being right" to "competence" though -- I'm willing to buy that Zvi is really good at being right, but he might not be nearly as good at hiring, leadership, working with other people, keeping schedules or any of the other things that are sometimes completely unimportant and sometimes absolutely critical to a job.
The head of the CDC job and benevolent dictator in particular probably require a LOT more in terms of managerial than predictive skills. Zvi would, I suspect, be most useful as an advisor to the benevolent dictator than the dictator himself. It's a shame that governments don't employ advisors anymore.
Cool hominid party trick: The world is abominably complex, so we call up organizational superstructures from the depths of meme hell and put them to work managing said abominable complexity. This is adaptive, apparently.
The dread blueprint of the FDA lurks latent in the collective unconscious, waiting to manifest and consume the souls of unsuspecting hunter-gatherers.
Are you referring to the articles on lorienpsych.com as your small database? I thought you meant more like a weighted average of patient ratings like when you recently compared different amphetamines.
Yes - the amazing thing is that our society works far better than we have any right to expect. (And far better than it would if 99% of reform proposals were adopted.)
And...just put a link to this post on your database (BTW, where is it?).
That *ought* to be sufficient, but of course it won't be for some people. You just have to live with that - valuable info online will help many people and hurt a few who misuse it. But it does net positive good.
The alternative is to become WebMD. They're *cowards*. They know they're destroying the utility of the information they offer, but they have shareholders to appease and lawyers to fear.
You can, and must, be braver. Probably you won't get sued over this, and if you do you'll probably win. But you're in the lucky position - unlike WebMD - of having a large fanbase who appreciate what you do and will bail you out if all goes pear shaped. I can (and do) guarantee that you'll never have to worry about financial hardship because of losing a lawsuit over this stuff - somebody in your fanbase will ensure you have a well-paying job (in the unlikely event you need help).
"What I sometimes call Marx's Fallacy is that if we burnt down the current system, some group of people who optimized for things other than power would naturally rise to the top. Wrong. People who most brutally and nakedly optimized for power would gain power; that's what "optimize" means. The interesting thing about the current system is that, after millions of very smart and altruistic people have contributed to it over generations, sometimes gaining and keeping power within it is modestly correlated with being good and right."
This is so brutally insightful and hilarious at the same time.
I agree with this analysis of "politics vs being right/doing the right thing" but it creates another obvious problem: the good of being right in the moment is bounded, but the good of being right in the future is unbounded
One of my favorite scenes from the Wire is when the newly-elected mayor of Baltimore has a problem. He needs money to bail out the public school system, which he can get from the governor, but if he does, he becomes "the mayor whose failing public schools forced the governor to get involved and the state as a whole to contribute tax dollars to bail him out," thus ending his future ambitions of becoming governor himself. And of course, once he is governor, his ability to do good things will be much greater than his ability as mayor. So he decides to let the school system continue to collapse to preserve his opportunity at future success, and future good.
I think almost every politician or person who has a political component to their job has made this calculation at some point in their lives. And in some sense its hard to say they are making the wrong call. Because if that fictional politician goes on to become President and signs a bill which ends child poverty, whose to say it wasn't worth a few thousand kids in Baltimore?
I don't have a solution to this problem, except to say that at some point everybody in every position that might even have a glimmer of power has to ask themselves "what's the absolute amount of good worth sacrificing today for the prospect of being able to do good tomorrow?" And to think really hard about the answer.
You are missing the HCQ elephant in the room (and the Ivermectin rhino too....). This makes your otherwise excellent post dismayingly close to worthless, until you can come to terms with the extraordinary number of lives lost due to the corruption of the process that Fauci, putting it enormously too mildly, allowed to occur.
I disagree with your analysis — that corruption/self-preservation explains why Dr Fauci does not always say the truth. Externalities and/or the possibility of being misunderstood are more important factors IMO. For instance, while an unknown blogger can say in March 2020 "we should all wear mask now", Dr Fauci or the NYTimes cannot say the same because they're afraid it will push people to hoard masks.
This is a really excellent post. It certainly makes me feel grateful we have a magic machine for consistently extracting pretty-good people with pretty-good ideas out of the population.
That said, there's a related issue: Why has the magic machine's performance seemingly decline over the course of the last ~80 years? Maybe it's an inevitable result of Moloch (for example, because as you optimize more aggressively, the gap between the most competitive and most good-producing ideas grows, or something). If so, too bad, guess we're screwed. If not, then we could sure benefit from figuring out what's causing the machine to function worse.
(As this post argues, any tinkering we do trying to fix the magic machine should be very cautious, because it sure would be a shame if something were to happen to it.)
So the central claim is that the experts at the top are genuinely super-capable, but corrupted by incentives. It's a fine hypothesis, but what evidence is there in favor of it? In academia at least I can point to an ocean of crappy research whose existence benefits no-one, so you can't blame the incentives. A Stanford epidemiologist attending a conference in March tweeted:
>At AHA #EpiLifestyle20 in a room full of epidemiologists and scientists and not ONE person is wearing a mask. Lots of hand washing & sanitizer stations. If 800 epidemiologists aren't wearing masks, you don't need to either!
We're talking about (tenured) people personally putting themselves in the way of harm here, can we really blame corruption and incentives?
I think you need to more seriously entertain the hypothesis of genuine incompetence.
> "A machine which takes Moloch as input and manages - after spending billions of dollars and the careers of thousands of hard-working public servants - to produce Anthony Fauci as output. This should be astonishing"
I think this is indeed a v good approximate account of our situation, & it shows both how well we usually do, considering how bad things could go, & also how tenuous is our toehold above the abyss. So far we've avoided the really terrible outcomes (the last four years of D.T. notwithstanding), but there have indeed been some terrible outcomes for some, & some very close calls -- some of which we know about (Bay of Pigs, Three Mile Island) & some we don't. There have also been some heartbreaking misses (eg the history Nathaniel Rich details in his (doubtless imperfect) _Losing Earth_). How well will our luck hold as we move on into 21st Century Climate Change -- to name just a single one of the many fun, um, opportunities staring us in the eye? I too am a reformer not a revolutionary, if I have my choice, but that choice is "all other things being equal," & more & more, they aren't.
A very good piece and essentially correct, unfortunately. The one part I'm not sure of is the implicit assumption that even mediocre expertise feeding into a political decision mechanism is better than no political mechanism.
Consider Covid. The advice that was given by Fauci and others was probably better than what most people would have done with no advice, better on average than what most would have done (although possibly worse than what you and I would have done) relying on whatever sources of information they trusted. On the other hand, the vaccines we are now getting took about a week to design and eleven months to get FDA approval. In a world without the FDA and other political constraints, a month of challenge trials would have shown that they were effective and not very dangerous, after which they could sell them to all comers as fast as they could ramp up production. It's hard to believe that that wouldn't have saved a couple of hundred thousand lives in America, a million or so in the world. So if you look at the overall effect, it may be negative. We might be better off with the same level of mediocre expertise and no levers of power for it to be applied to.
At a tangent ... . Your description of Fauci reminded me of my conclusion about Nixon. There's a bit of the tapes where he is talking with one of his people about the idea of ending the interest equalization tax, which was a restriction on capital mobility. Nixon points out that one or two people have said doing it would be a good idea (true). The aide points out that there are no votes in it. Nixon says (by memory, not a quote) "What the hell. Let's do something just because it's good for a change." My feeling was that he had spent his life getting power in order, as he saw it, to do good things, and then found that he had to use the power almost entirely in order to keep power.
The current dramatic case of people choosing the wrong expert to believe in is the election. Something like a quarter of the population, if polls can be believed, thinks Trump really won. Their experts say so — and they have concluded (correctly, although not in this case) that the official experts can't be trusted.
> Even if this flattering story is true, it doesn't scale... Compared to the median person who disagrees with the experts, the experts look pretty good.
This is an interesting way of phrasing it. I would say that the problem you're pointing at - the incentives for getting and holding power mean you get mediocre experts is, _itself_, a scaling problem.
Here are two thought experiments to illustrate my point:
a) Imagine there's a single unified government for the entire solar system, with population in the hundreds of billions of people, and levels of wealth substantially beyond what's available today. Billionaires are a dime a dozen - any pop star or athlete easily makes it there. There are people who own entire planets. Are the incentives for being honest and accurate _more_ aligned with getting and holding power, or less? How corrupt would you imagine the official councilor on health, for the entire solar system, which includes multiple alien races and sentient robots and the bio-corporations of titan?
b) Imagine that the federal government basically doesn't exist and almost all power - taxation, trade regulation, everything else - falls on the individual states. Each state has its own experts, who compete for power in _much_ smaller games. How corrupt is the state health controller for each state?
It seems to me that the more people involved in a competition, the greater the disparity between 'strategies to get and hold power and status' and 'strategies to be accurate.
The lesson of moloch, to me, is that we should try to avoid, as much as possible, massive, winner-take-all games of power. _That_ seems to be the thing that doesn't scale - i.e., it works OK when you do it with smaller numbers of people, but it falls apart as you start to add more and more people to the system.
There's likely a lower limit here too - with too few people, there's not enough of a pool of capable people to select from. With too many people involved, the selective pressures get so insane that the winners must be corrupted.
I love that you conclude that we are "insufficiently grateful" for Anthony Fauci, while belittling his free-agency in the same breath. It's surprisingly effective, and somewhat hilarious in light of the Fauci-is-Jesus-reincarnate memes that people are worryingly sincere about.
Funny that you say Trump is some sort of specifically bad outcome when the alternative was Hillary. I'd say the system failed by the time it presented either of them to us as a choice.
"Benevolent dictatorship, obviously, just get the best person in the country and let her fix everything. But everyone realizes this is easier said than done; the procedure to pick the best person is corruptible. "
One of Mary Renault's historical novels, _The Praise Singer_, is in part about tyranny, meaning, in the Greek context, popular dictatorship. One of the tyrants she describes, Pisistratus, is benevolent. In her version, he was the younger lover of Solon, who created the Athenian legal system. He tells the protagonist that when Solon finished everyone liked it, except one thing — everyone wanted a change to favor him. Solon's response was to leave Athens, because if he wasn't there they couldn't make him change it. "They keep his laws — I see to that, who could have given them laws they liked less well." The tyrant has been kept honest by the memory of his dead lover.
It's a powerful scene, but not a very reliable procedure.
> What's the best form of government? Benevolent dictatorship, obviously, just get the best person in the country and let her fix everything. But everyone realizes this is easier said than done; the procedure to pick the best person is corruptible. At one point we tried a very simple best-person-picking procedure that really should have worked and ended up choosing Donald Trump as the best person. I'm still not really sure what went wrong there, but apparently this is really hard.
I know you’re having trouble being serious here, but no, at one point y’all tried a very simple best-person-out-of-two-far-from-obviously-close-to-the-best-in-the-country-picking procedure. The prior picking-which-two-people-to-choose-from-procedure was very complicated and obscure and probably very corruptible.
You tell a hypothetical story about what would happen if Biden appointed Zvi. But what actually happened when Trump actually appointed Scott Gottlieb head of FDA? Why did he only last two years? Was he pushed out as a political cost, as in your story? Did he burn out fighting with the organization, a different but parallel story? Did he have power to control the organization, or would it not have made a difference if he had stayed for another two years?
WebMD is useless but your specific example is an unusual choice. Aspirin is relatively safe for healthy people but less so if you have liver damage. The Wikipedia article for salicylate poisoning says there were 20,000 cases in America in 2004.
Great post. The last line -- "prediction markets would still be better" -- is idiotic [1], but we can't have everything. Scott isn't perfect, but he is better than we deserve; legibly mediocre.
[1] If prediction markets were so great, we'd see them happening *somewhere*. Gambling is legal in Canada; you can't blame everything on anti-gambling laws. Where are the Canadian companies that use prediction markets to make decisions? They don't exist. You think they should? Go to Canada, start a company that uses prediction markets instead of a CEO, become a billionaire, and THEN I will believe that prediction markets work. Until then, they are just a way to signal "rationalist community good".
I'm surprised that no one has yet mentioned Zeynep Tufekci (https://zeynep.substack.com/people/299814-zeynep) as another writer who seems to have consistently been righter than the experts. She's been my go to for COVID information for quite some time, and probably the person who's most influenced my actions and what I recommend to others.
One assumption that I'm not entirely sure I agree with is this one:
"If you expose the plan to politics, the politics will drag it in the direction of being worse. Every feedback channel you open up is a way for somebody to attack you. "
Some massive assumptions involved are that:
1) You will consistently have specific illegible experts
2) These experts have satisfactory self-contained knowledge
I agree with the mechanism that politics CAN drag things in the direction of being worse because it inevitably creates channels for people to jam questionable inputs.
But politics also has the ability to create channels for people to jam in relevant inputs. Most people can't, don't, and won't develop the approach needed to balance a proposal against multiple different objectives. Part of this is maintaining power. However, Fauci isn't strictly acting as providing medical advice, but rather his pronouncements have economic & sociological impacts.
I think the essay creates a good argument for Plato's Republic, but when you keep in mind that Plato's ruling class must exercise judgment in lying, and that the web of information in a modern society is an order of magnitude more complex, etc, etc. Then... I'm not so sure. I am not sure if I want to sign off in an idealized theory even.
Let's put it another way: does Fauci & the CDC by itself have sufficient expertise to decide to lockdown the US economy? The answer is probably no, because why shouldn't the CEA (Council of Economic Advisors) get a say in that? That's a really big decision that hit's their domain. And maybe we need the APA (American Psychological Association) because this really could cause psychological stresses that the CDC may not be as expert in. And... at this point, we may also be talking about a value judgment, which requires some sources to provide those inputs (ethicists? representatives of the US population?) And when you realize all of the people who should legitimately be consulted to make a complex decision, you now have politics.
That leads to the idea that politics is a natural result of human interactions involving complex decisions. Some politics is good & some politics is bad. But it needs to exist, and it isn't so easily separated into "impure power-grabbing". Coordinating with all stakeholders is "maintaining political power", but the stakeholders legitimately exist! And pretending they can be replaced (even in theory) with the ideal dictator means that we somehow don't believe there are other stakeholders, or that we think they can be ignored. Sometimes bypassing stakeholders works!! However, it really can blow up quite horribly!! "Move Fast and Break Things" is a very productive & very destructive slogan.
All of which is to say: Scott this is a great post. Can you please tell me if Democratic Republics are good or bad? This post has implications and I just gotta know.
I like this post in general, and I agree with the central thesis. However (as someone who put nontrivial thought into both masks and coronavirus case/death estimates, see eg: https://pandemic.metaculus.com/rankings/), I disagree with the causal mechanism is that corruption caused accurate people to become less accurate in public-facing ways.
I think it's more likely that the corruption caused counterfactually competent people to be less accurate, even in their own thoughts.
Speaking of Lorien Database changes, you might want to make the UI a bit more useable on mobile. There was some charcoal-on-prussian-blue text and stuff last I checked.
I think you’ve conflated two forms of “expert,” the advisor and the policy maker.
While a policy maker is evidently subject to the sort of public pressure you describe, one would hope a president could get better advice in private, and that experts whose public facing role limits them might still be able to speak more frankly “on background” to reporters, politicians, etc.
I... do not get the sense our expert class recognizes their failures in private. It is not clear to me, for instance, that the FDA’s top leadership suspects that their refusal to consider and approve the Astra-Zeneca vaccine is effectively mass murder. And it isn’t clear to me that in private public health officials understood the threat of COVID-19 until remarkably late. That’s more worrisome than that our system puts pressure on the most public-facing roles.
Or, put another way, we are selecting almost entirely for “can handle dealing with the public” rather than “gives the best possible advice” in the filter that selects for our public experts.
My current nomination for "most trusted pandemic amateur" would be Zeynep Tufekci. I'm skeptical that Zvi always optimizes for being right, versus coming up with hot takes to make us angry about how people are handling the pandemic? (Maybe we should be angry, but I need better evidence.) Still a good source of link and questions, though.
In particular, I am wondering if anyone has good sources or explanations for how the FDA is making decisions about vaccines during the pandemic? This seems mysterious and many have questions, but I don't think I understand it yet.
"The experts successfully swooped in and saved us from all of that, figuring out which way the wind was blowing only two weeks later than competent amateurs. This was a useful service. Without the experts, things would have stayed open forever."
I expect that people would, in Bill Gate's memorable phrase, "notice the bodies piling up in the corner" eventually.
It's my opinion that Fauci has done a terrible job, but I genuinely don't know anyone else who would've done better in this nightmare. We have a crisis of expertise AND implementation.
Zvi probably also has a very permissive peer group. For normal people it's still a multi-criteria decision-making problem: maximize rightness together with minimizing slights/offense/loss of status within the peer group. Whatever experts are proclaiming also gets used as input and it gets difficult to deviate from it depending on how peers process it. At least us normals don't have to deal with lobbyists or activists that want to remove us, I guess.
I love it when multiple things I read converge around the same time even when they weren't created at the same time. I just rewatched the classic CGP Grey video, Rules for Rulers, yesterday - https://www.youtube.com/watch?v=rStL7niR7gs&ab_channel=CGPGrey, which discusses what people in power need to do to stay in power, and has a nice similar energy to this post.
It was just yesterday that I was commenting on Robin Hanson's post (though I was also thinking about a recent podcast he was on) about Experts Versus Elites*, pointing out that experts got COVID-19 wrong initially and elites had to correct them. This would have been a good post to link to.
One takeaway that I get from this is how important anonymity can be, sometimes empowering people to speak the truth which would otherwise have too great negative consequences for themselves and the systems they represent. While we probably don't want anonymous people running our country, it's definitely among the simplest ways to bypass some of the unintended forces of corruption mentioned here, and is, for example, why voting is at least intended to be anonymous (and ideally would be so in our government as well, just with cryptography to provide all of the various assurances and verification that we'd want).
Excellent post. Although it sounds like one solution would be to bring the requirements of "being in power" more in line with "being right"; e.g., by calling out, defunding, demoting, firing, cancelling subscriptions to, etc., institutions and individuals who have done markedly worse than competent amateurs and have caused harm in the process. So "taking to the streets, pitchfork in hand," but to a lesser degree.
Are the problems Scott is addressing here [just] special cases of the general issue of how to juggle social signaling with efforts at accuracy / rationality / reliability? I'm not trying to trivialize it -- I think this problem is one that repeatedly arises both here (and back in the day at SSC) and in lots of places both in the Less Wrong crowd and more generally: how to raise questions *at all* when certain questions or certain responses or certain thresholds of reliability, etc, are also *signals* that mean things like "I am on your side," "I am powerful and you should be on MY side," or "I am NOT on your side and you should fear me." I am thinking of certain social justice questions, for instance, but one could multiply examples ad nauseam. Part of Scott's point, I take it, is that questions of scale play into this issue complicating the matter.
I want to offer a small correction. You claim: "What I sometimes call Marx's Fallacy is that if we burnt down the current system, some group of people who optimized for things other than power would naturally rise to the top. Wrong. People who most brutally and nakedly optimized for power would gain power; that's what "optimize" means."
At first glance, this claim seems by definition correct. But it does not *necessarily* hold. If there is ex ante heterogeneity in attributes. For instance, if we burnt down the current system and declared that the person born on the birthdate determined by some unpredictable quantum fluctuation has all the power, it would be *difficult* to overcome your ex ante handicap (birthdate in this case).
While that's a pedantic example, the same could imaginably be true if we only allowed the most excellent and well-respected academic economists to rise to power. Even for an optimizing power-hungry sociopath, it is unlikely that they will be able to rise to the top--not enough natural talent.
You could instead say that part of what needs to be optimized (natural talent) is out of the individual's control space.
For a statistician, if the variance in the latent variable that determines power is primarily driven by non-optimizing behaviors (primarily natural talent) then optimization has less of a potential role to play.
All that said, I agree with your overarching point, for most practical situations such as Marx's, what I said is nonsense--but in the interests of steelmanning, I think the theoretical idea that one can build a society that isn't driven entirely by power-hungry optimizers is not as self-contradictory as it might seem.
> . Citizens Against Lockdowns argues that the CDC already screwed up by stressing the later-proven-not-to-exist fomite-based transmission, ignoring the needs of ordinary people in favor of a bias towards imagining hypothetical transmission mechanisms that never materialize;
I know you're making an example and trying to be funny, but this is a really important point that needs to be made:
Most of the CAL-types (at least, the normal ones who aren't walking conspiracy-theorist stereotypes) aren't mad because "they said fomites are real but fomites don't real". They're mad because the experts figured out fomites don't real six months ago _but we're still going through all the surface-decontamination safety theatre_. They aren't mad that experts got it wrong. For that matter, it's not even clear to me that they're mad at the experts at all. They're mad that the people who constantly say that they're "following the science" didn't update their beliefs when the science came in, and it's been six fucking months of doing all this ritual bullshit that has provably zero effect on anything.
Another CDC vs. Zvi difference is that the CDC is doing public communication to a mass audience, often filtered by the media / the public discourse. Whereas Zvi is writing directly for a niche audience of people who are willing & able to put in an unusually large amount of time/attention/thought to build detailed models. Doing mass communication pushes towards:
Just sharing your conclusions rather than the details of your reasoning
Giving simple guidelines/approximations (e.g. "wear a mask", "6 ft apart") rather than more details or the underlying model
Not sharing things that you're more uncertain about
Trying to keep a consistent story rather than changing your views fluidly
Zvi doing mass communication would face some challenges that the real Zvi hasn't had to face (and which a Zvi-in-a-cave setting policy might be able to dodge). Just replacing pages on the CDC website with Zvi blog posts would create some problems.
I really appreciate this analysis. I only wish you emphasized more the relevance of regulatory capture. Whereas the FDA, CDC, and NIH are supposed to regulate the pharmaceutical industry, they very often are a mouthpiece for that industry, having been captured by it.
The generic drugs hydroxychloroquine and ivermectin, for example, have been successfully marginalized, despite abundant evidence of their efficacy against Covid-19. See hcqmeta.com and ivmmeta.com for summaries of the research to date and links to the original studies—which are abundant, at this point. (Really, please do look!)
For hydroxychloroquine: "The probability that an ineffective treatment generated results as positive as the 198 studies to date is estimated to be 1 in 1 quadrillion (p = 0.00000000000000083). Early treatment is most successful, with 100% of studies reporting a positive effect and an estimated reduction of 66% in the effect measured (death, hospitalization, etc.) using a random effects meta-analysis. ... 91% of Randomized Controlled Trials (RCTs) for early, PrEP, or PEP treatment report positive effects, the probability of this happening for an ineffective treatment is 0.0059."
For ivermectin: "100% of the 37 studies to date report positive effects. Early treatment is more successful, with an estimated reduction of 82% in the effect measured using a random effects meta-analysis. ... Prophylactic use also shows high effectiveness. 100% of the 19 Randomized Controlled Trials (RCTs) report positive effects, with an estimated reduction of 72%. ... The probability that an ineffective treatment generated results as positive as the 37 studies to date is estimated to be 1 in 137 billion (p = 0.0000000000073)."
It requires a certain amount of research-savviness to evaluate studies like these firsthand, rather than relying on media reports and authority opinion. But for those who do, and who proceed to evaluate the relevant research firsthand ... whoa!
Given America's dire need for cheap, safe, and effective early treatment, I find it hard to explain the marginalization of these repurposed generic drugs any other way, than through capture of the relevant institutions and regulatory bodies.
The NIH canceled their large RCT for hydroxychloroquine as soon as it became clear that a vaccine for HCQ would be possible. In the months since, no other treatment has been allowed to steal the limelight. We're just been told to wait (and wait) for the vaccine.
I am really distressed by this. But I see no other explanation which accounts for (a) the abundant research available and (b) the official narrative about these drugs, which appallingly contradicts that same research.
> The experts successfully swooped in and saved us from all of that, figuring out which way the wind was blowing only two weeks later than competent amateurs. This was a useful service. Without the experts, things would have stayed open forever.
I don't understand how you can be as smart as you are and still think that "things staying open forever" would be a bad thing. Cities with no meaningful lockdowns and widespread noncompliance are having VIRTUALLY IDENTICAL outcomes to cities that locked down super hard. In some cases, they're having better outcomes (though my gut check is that this is just statistical noise). It is not AT ALL obvious that the 'experts' that California has delegated their critical thinking to have done anything other than perpetrate an absolutely massive crime against humanity.
Scott, the best systems for combating COVID aren't ours, but the ones in the countries that got COVID right. These are, variously, China, Vietnam, New Zealand, Mauritius, Singapore (outside migrant worker dorms), North Korea (apparently), Australia, Thailand, Taiwan, etc. Japan, the Philippines, South Korea, Cuba, Norway, Finland, Denmark, Cyprus, Mongolia, etc. did better than average. The question is not to burn the system down or not, but whether to move to a better system. I propose a government by superforecasters, because superforecasters seem like people who know what they're doing.
"There are few biologists who deny evolution, few epidemiologists who think vaccines don't work, and few economists who are outright communists."
Low bar. Let's raise it higher.
"This should be astonishing, and we are insufficiently grateful."
We got half a million deaths and counting. That's twice as good as Russia per capita, but it's worse than Brazil and far worse than the Philippines. It's impressive by precisely zero metrics.
One consequence of this kind of thinking is that its interesting to look at different incentive structures, and see where they have better results. The first comparison I thought of for this was countries with public healthcare systems. For example the nhs.uk has a similar to webmd encyclopedia of common conditions, drugs, etc. Which seems much clearer to me as a non expert. eg the page on aspirin says its mostly safe, and lists common side effects with the context that they are rare. https://www.nhs.uk/medicines/aspirin-for-pain-relief/
I don't think this is because public healthcare systems are magically better, (and please don't make this thread into an argument about m4a). My best guess for the difference is:
If you are a National Healthcare Service you have access to huge numbers of government lawyers on retainer, and massive institutional power, so the risks and potential of being sued are much lower than for someone like webmd.
Also, for webmd, someone with a headache deciding they have an obscure cancer does you no direct harm, but if you are a public healthcare provider they will be taking up your time and resources. So presumably someone did the calculation that it was better to stop those false positives and deal with the occasional complaint.
And on the individual level, a large organisation can shield people from bad incentives. Presumably these pages were written by employees of the health service (probably in committees) who reported to higher up adminstrators, not dealing directly with the end consumer. So they could be given an instruction "write clear and concise advice" without the political issues trickling down to them.
So maybe the takeaway is that as well as there being a function of being too small to care you can also be too big to care? If Fauci was the american health tsar he would give better advice, though that runs into the benevolent dictator issues. The brits don't seem to have done so great with covid advice, so maybe the protection only works when things don't get too political
My big problem with the expert response was the lying -- statements early on like "don't wear masks because they don't work", which was an obvious lie to keep people from panic-buying masks, only to then mandate masks later. Followed by "we just need to lock down for 6 weeks to flatten the curve", followed by "we'll probably still be locked down at the end of 2021". Or the refusal to talk about honest tradeoffs - "this is how many people will likely die if we don't lock down. This is how many people will probably permanently become poor if we do."
The American people did, in fact, elect an idiot game show host. However, we were given the choice between an idiot game show host, and possibly the most unlikable person to ever run for office. Hell, this time around, the Democrats chucked any candidates that were even remotely interesting, and seems so hell-bent on telling the millions of people that voted for Trump that they are ontologically bad people that need to be scorned and punished. That's how you get someone far worse but more competent in 4 years.
I agree wholeheartedly that tearing down the system and assuming that somehow people that care about things other than power will magically rule us benevolently is a pipe dream. But it seems that both the far left and far right have lost all hope that things can look less than shitty for them and their kind in the forseeable future, so they're willing to chance it. This is something that the current administration, if they were competent at all, would attempt to diffuse.
This is a great post, thanks! That experts, institutions, and other avatars of Consensus are a weighted-like-a-ball-and-chain average of the views of the rest of society is a really cool and powerful idea. So is the point that they aren't inherently dishonest or unprincipled-- it's just that the Institutional Power Genie shows up and replaces them if they get too principled or honest.
But I have three reservations.
1. Who exactly is the Institutional Power Genie that hypothetically gets Zvi fired as CDC Director? Since we live in a democracy it must represent some actual people, right? But it clearly doesn't represent you and Zvi. And it doesn't represent your Straw Rednecks either, because they're, shall we say, not Biden's core constituency. The charitable interpretation is that it represents some third group, Straw Bluenecks who think the virus is real and bad but either are risk-averse in a very particular way that makes Zvi unacceptable to them, or just get a lot of utils out of knowing that Very Serious People are in charge. The uncharitable interpretation is that the Straw Bluenecks are just Fauci in a trenchcoat-- that the institutional, power-seeking mindset has become self-perpetuating by getting itself unfairly weighted into the average. I favor the uncharitable interpretation.
2. If Zvi's proposals are in practice not implementable by real-world institutions, then how are we supposed to interpret them? After all, a lot of what he (and everyone) posts really does amount to "here's what I would do as a benevolent dictator" not "here's my advice to you, a private person, as a friend". Is he trying to change people's minds to shift the weighted average? Or is he saying the average should be weighted differently-- toward him? To the extent that he succeeds in shifting the average, how concerned should we be that some illegible blogger can shift the average just by being popular and insistent? To the extent that he doesn't succeed, how concerned should we be that his posts are just a form of outrage-mongering entertainment?
3. What does it even mean to say that someone is "illegibly good" or "illegibly bad", if it's misleading (due to illegibility) to imagine their beliefs actually being implemented? Good or bad for whom? Zvi's advice is good for us, maybe, but that seems to just boil down to saying that we agree with it. How would we convince Straw Redneck's audience that it's good for them? Wouldn't we have to appeal to consensus-- i.e. Fauci? And if so, why are we still trying to shift Fauci's position toward our own? This model, when you work it out, actually seems to support Weyl's point of view: the benefit of Fauci, insofar as there is a benefit, is that he's legible to the broadest possible range of people.
There's another aspect to what Dr. Fauci does -- He isn't there to dispense the truth to the masses, he's there to make statements to the masses that will cause them to do what is optimal relative to Dr. Fauci's plan. Within that goal, he tries to avoid saying anything that is actually false.
For instance, early in the epidemic there was a real risk that the public panic-buying masks would vacuum the supply chain empty of them right at the moment that medical people needed far more of them. So the statement was made that there was no solid evidence that masks worked against Covid, which was true but any person who attempted to analyze it would realize that masks were likely to be quite effective. Readers of this blog will probably notice that is an attempt to avoid the tragedy of the commons -- for each individual, buying a mask as quickly as possible is the optimal choice, but also preventing others from buying masks is also optimal.
aMore subtly, the optimal plan depends on where one is in the economy. As a computer programmer, the cost of an intense lockdown is personal nuisance. For a restaurant worker, it might be homelessness. Someone has to strike the balance between those competing interests, and that's a lot easier to do if you disguise that it is being done.
Everything here reminds me of this Fantastic Anachronism post last month -- despite the horrible track record of replication in social science we should still default to trusting credentialed experts because it's better than any alternative https://fantasticanachronism.com/2021/01/12/unjustified-true-disbelief/
I'm really glad you're writing on moral mazes. I've been troubled by Zvi's simulacra work for a while now: he seems to be saying important things about what happens to adversarial systems, but the end result is so bleak. I had to go back and re-read the goddess of everything else to remind myself about the power and beauty of cooperation.
I would be very grateful for a theory that harmonizes Zvi's work with Elua's power. This article is a good first draft, but I long for hope that the Elons and Zvis of the world aren't anomalies, that the quokkas can learn to protect themselves and the slytherins can become great philanthropists.
I generally agree with this, but: I don't think it's always been this bad. Back during WW2, for example, we seemed to have competent leadership. We put Vannevar Bush in charge of military research, and that worked out great. I think something has gotten worse over time. And it's worth trying to figure out what that is and why it happened.
> When the Director of the CDC asserts an opinion, she has to optimize for two things - being right, and keeping power.
I think it's really more complicated than that. I haven't been on the 'other side' of public decision very often in my life, but I have a few times through my job at a tech company, and I know these kinds of things almost look unrecognizable to an insider, compared to an outsider.
My tech company example is, say, a widely-publicized product launch to which the general public assumed a lot of fair things, the gist of which is that we were insane and had no idea what we were doing. In fact, the true explanation was completely unrelated to that -- that the target of the launch was a type of enterprise customer that was alien to the public.
I can imagine many forms of this that might happen for a Fauci public statement that take different directions than "maintaining power". For one thing, in the early days of the pandemic, it was truly not clear in the data how serious it was. A random person can say: "oh, I have a feeling this is much worse" but the CDC / whichever officials can't. They have to balance their gut against the data they have, their internal policies, and the fact that if they say something extreme it carries a lot of weight, which means shouldn't be wrong both for logistical reasons and to maintain credibility.
There's just nothing on the line for an individual making predictions. When you end up in a position of great influence, there's a lot on the line, and you have to be much more careful. Rationally, of course, it would make the most sense to overcorrect for harmful outcomes, and spend more resources than needed. But then what happens if you're wrong, and a lunatic congress comes at you for budget cuts decrying your uselessness? (And this doesn't even broad the question of what the right behavior is when you're trying to avoid completely losing Trump's ear. The "Case Against Fauci" article thinks it's not worth it, but I'm pretty skeptical.)
I'm just saying -- it's just not just being right and being in power. "Being right" looks very different for a public official in a political situation than it does for a blogger.
Those are the only economists who are correct, though. Remember Scott that in your own reading of Marx you could not seem to comprehend his basic sentences and instead deferred to Peter Singer's own (incorrect) analysis. This is politics as the mind-killer, as the Rationalists say.
Hi Scott, long time reader. Really like your writing.
But I've noticed a somewhat distressing tendency in your recent posts (by which I don't really mean just the ones on substack) and I think it's getting worse and so I will speak up. I notice an increasing frequency of things like this:
"But someone else will decide to always trust *their* friend, a guy in a MAGA cap who says coronavirus is fake and Dr. Fauci is a Satanist. "
"At one point we tried a very simple best-person-picking procedure that really should have worked and ended up choosing Donald Trump as the best person. "
"Our system of expert-having is actually much better than we deserve, given that we elected Donald Trump president."
And I have to wonder if this is necessary or even good, or if it is just dabbing on trump supporters because it feels good. Is the point you are trying to make actually best served by including a MAGA cap, or is this just exclusionary to some of your potential audience? Politics is the mind-killer.
"The whole scientific-technocratic complex is a machine which takes Moloch as input and manages - after spending billions of dollars and the careers of thousands of hard-working public servants - to produce Anthony Fauci as output. This should be astonishing, and we are insufficiently grateful."
That kinda sounds like a proof-of-work blockchain. It consumes vast amounts of resources, but that's not (just) because it's inefficient, it simply takes a *lot* of effort to defend such a system from attacks with our current technology.
Do you think that locking your household down two weeks early as a good decision? It doesn't seem like one to me, though as the pandemic lengthens it kind of seems like well, two weeks here and there, who cares? But just in the spirit of self-reflection, didn't you lengthen your personal pain at a time when the chance of infection was essentially negligible?
If you know anyone who has been tempted to despair about supposed 'failures' of the western, liberal world order and its institutions over the last year then I'd encourage you to send them this. Whether something is considered acceptably good depends on what you compare it to - what we have now is clearly far better than what we would get by burning the system down. What's more interesting is whether what we have now is better than what we had in the past - and there's evidence (from things like the way the polio vaccine rollout was much faster in the 40s than our rollout today) that there are a few aspects of the current expert-selecting institutions that seem to have regressed, or not kept up with the increasing difficulty of the problems handed to them.
I think it's worth noting that the historical comparisons aren't ever to us actually succeeding at dealing with pandemics in the past, but to things like "WWII-style" efforts - i.e. thinking that if we could just do x as well as we once did y then things would have been a lot better. This implies that if you made an institution analogous to e.g. the weapons researchers of WW2 and the governments that funded them, or NASA in the 1960s, without copy-pasting 1940s/1960s society wholesale, the outcome would have been better.
To me that suggests its institution design that's the culprit, and that there are ways to filter our legible experts to be better that we've missed out on.
But even in making that kind of comparison, we're already back in the sane, concrete realm of institution design and trying to patch up the system we have with tweaks we know might give us better experts more reliably, rather than the ethereal realm of comparing the whole of our society to some imaginary alternative.
I'm lucky enough that I have had no reason ever to use medical services in the US, so I can't comment on that. But I have had quite a bit to do with the US political system professionally , and have had a chance to see how it works (it doesn't). It's an almost uniquely dysfunctional system, massive, labyrinthine, highly politicised and personalised, where it's hard to know who is in charge of what, and decisions take forever. Irrespective of who your political leaders are, I think your country is going to have a terrible time coping. Most of the states that have coped well have strong, professional and well-managed public services which draw a strict line between career and political appointments. (Singapore is a good example). Countries that used to do that but don't any more (like Britain and to an extent France), have suffered worse as a result.
But it isn't just the political system(and of course sheer luck plays a big part as well). You have to take into account the social capital of the country and the willingness to work together (again, most Asian countries have done well for this reason, but so has New Zealand). You also have to take into account political culture, and what seems important not only to politicians but to the whole PMC and media classes. The rest of the world watched in stunned disbelief as these classes in the US descended into severe Trump Derangement Syndrome, and stayed there, even while the epidemic was ramping up. But we had problems in Europe as well. Closing frontiers was essential but not done nearly quickly enough for reasons of political culture. Partly this was the EU's neoliberal ideology of open markets and borders. Partly it was the identification of frontier controls with nationalism (hiss) and even Trump (double hiss), which meant that if you supported them you were a Xenophobic Nazi Fascist. "Viruses don't have passports" smirked one European politician, who I think is out of a job now.
On Scott's wider point, I found his remarks uncomfortably apposite because I've suffered from a couple of these dilemmas about getting information on medicines. For the last fifteen years I've been taking blood thinners (first Fluindione, which is like Warfarin, and then Rivaroxaban). I was warned about not taking aspirin as well, and then I was told not to eat certain foods (bananas for example) rich in Vitamin K, and then as I began to read around, I discovered that most foods have an effect of some kind. Since I'm intolerant of gluten, I realised quite quickly that I was as likely to die of hunger than of blood clotting, and in the end I took a personal decision to eat a normal (but gluten-free) diet, on the basis that one thing would cancel out another. I'm fine and don't show any symptoms of uncontrolled bleeding for example. But nobody tells you anything useful. Nobody says how precisely this or that food affects the fluidity of the blood, for how long or in what combination. I've been told not to take Ibuprofen. Fine, but does that mean not even once, at 200mg? Or does it mean not continuously at a higher dose? If I take one for an inflammatory problem, will I die? Nobody seems to have any idea, or if they do they are not telling.
And I've been a chronic insomniac, probably for longer than most of the readers of this blog have been alive. There are two kinds of doctors in my experience. Good sleepers, who tell you to make sure you have a dark room and get plenty of exercise and you'll be fine, and won't prescribe anything, and insomniacs, who know that insomnia is terrible for general health, and that sometimes, lying awake at 4am and wondering how you are ever going to make it through the day, you start to think that suicide may be the easier option. The latter will prescribe something. So I've been taking Bromazepam from time to time, and, as Scott's wonderful new site points out, it's been argued that even small doses will kill you. I'm still alive, and I think I'll choose my own method of dying than you. Indeed, I think the only thing you can do is to try and see what happens. In Europe the problem with medical advice isn't quite so serious, perhaps, as in the US, but it's pretty serious. I've pretty much given up hope of getting any useful advice from medical sites, and I'm much more inclined to try something that has demonstrably worked for someone else. So I eagerly took up Scott's suggestion of Theanine, and took 500mg before bed (that was the only dose I could find). I was awake half the night and finally gave up. Any other suggestions gratefully received, but I'm going to try them for myself, not just be guided by websites that produce lots of words but no actual help.
>When the Director of the CDC asserts an opinion, she has to optimize for two things - being right, and keeping power.
As an excellent example of this, look no further than Professor David Nutt. Nutt was a Fauci-esque figure in the UK: he was, at one point, an advisor to the Ministry of Defence, Department of Health, and the Home Office, as well as President of the European College of Neuropsychopharmacology.
He also had no interest in optimising for keeping allies and power, and used his government positions to go around telling theh public exactly what the science said about addiction, illegal drugs, and so on.
He was rapidly fired for this, with several senior government figures accusing him of "talking about politics rather than science" -- which, ironically, was precicely the opposite of what he was doing.
Perhaps you should use a kind of Captcha in order to access your site? Instead of identifying 'all instances of porpoises waiting at stop signs', it could be a randomised test of the ability to reason statistically?
'If you can't pass this test, here's a helpful site where you can find out how not to misunderstand what I'm giving you..'
A very strong post, without a doubt the best you've written since your comeback.
There's a profound truth to the idea that only people outside a corruption-based system such as politics can regularly find and speak truth. Constant excellence is possible only on the individual level; mix in groups of people (with inevitable bias and selfishness) with varying incentive, and what you get is a compromised mess.
To remain a priest of Poverty (whether in an economic, social or political sense) who can always wager most, or to give up the vows for real gain? That is the question, and God knows it's not an easy one.
The biggest issue to me is that our system (particularly in medicine) has very few mechanisms for punishing over caution and lots of mechanisms for punishing good faith bold action. As a result, everyone with anything to lose is in maximum CYA mode at all times.
In a pandemic (or a war, or any other complex fast moving crisis) inaction is often just as deadly as incorrect action. But our medical establishment totally dropped the ball on switching to a “war footing”, instead acting as if minor relaxations of the normally truly Byzantine approval processes were bold acts of heroism. How many lives would have been saved if the Moderna and Pfizer vaccines had been approved at the same time as Sputnik V? No one will be punished for these tens of thousands of bodies, but if the vaccine had merely saved a hundred thousand while causing a couple hundred birth defects or whatever, heads would have rolled. This is absolutely backwards.
Of course it would be one thing if Fauci were merely over cautious - but he didn’t only make “WebMD errors” of excessive ass covering. He, by his own admission, told 2 big lies (or at least things he was pretty sure were untrue): “you should not be walking around wearing a mask” and “modulating my estimate of herd immunity based on what I think people will believe”. And beyond Fauci, “Racism is a bigger threat than COVID, so protests in a pandemic are justifiable”. That was not a WebMD error.
We have medical experts to deliver medical expertise. Perhaps we accept a bit of politicking and over caution as a cost of doing business. But when experts start openly lying because, hey, they’re experts, you wouldn’t go against science, would you? That shatters public trust in expertise. Fauci maintaining power might be good for him, but it’s bad for everyone else because half the country isn’t going to listen to him, and that’s in good part his own fault.
Do you think having more prestige for working in government bureaucracy would increase the correlation between gaining power and capability? East Asian countries seem to do a decent job of having competent officials and these things are much more prestigious there.
Really amazing writeup and there are a lot of valuable points in here. I'd like to give a bit more attention to this point since no one else has:
"Or maybe I'm dumb and biased, and actually the experts are doing much better than Zvi"
I don't think that's true in general, but after reading through Zvi's blog there is an error that could have cost thousands of lives if Zvi were in charge. He thinks that at least one of the vaccines should have been open source.
If Zvi is a tech worker, then it's easy to see where he's coming from. In tech you want a thousand flowers to bloom, and the best survive. Doesn't matter if one product sucks, only the good ones will still be on the market in 3 years. Well medicine isn't like that - move fast and break things does not work.
Medicine is built upon trust - and trust is hard won and easily lost. If 5 startup vaccine manufacturers compete to produce the vaccine doses as fast as possible and just 1 of them screws up that's a huge problem. Perhaps 5% of doses are ineffective due to insufficient QA or perhaps some of the doses result in higher rates of side effects. Well then the entire well is poisoned - news stories come out, politicians on the left and right seeking easy political points lambast vaccine producers, anti-vaxxers see a golden opportunity to move public discourse. Perhaps it doesn't kill vaccine production entirely, but public trust falls, investors and regulators become wary and vaccination decreases.
And all that for what purpose exactly? The companies best able to reliably ramp up production of millions of doses are the companies that have already done that in the past. And they have done that with rigorous QA that they've been forced to develop after hundreds of failures and lawsuits. Open sourcing would not significantly increase production and it would introduce unnecessary risk.
Now perhaps I'm misunderstanding him, perhaps he doesn't really mean open source - just that any established pharmaceutical company should also be able to secure a contract to make the vaccine - no exclusivity deals. I doubt that would actually change much. All of the big players interested in vaccine production seem to already have jumped in, and they've all chosen their champion. And the reason is simple - economies of scale. It's cheaper to produce 10 million of one kind of vaccine than 5 million of one and 5 million of the another.
The problem with balancing optimization for truth and optimization for power is that such compromises might not be stable. Individual people's ability to find the truth is judged by the very institutions in which they're trying to advance. So the more institutions optimize for maintaining power, the lesser their ability to find the truth, including recognizing people with the theoretically optimal balance of truth-seeking and power-seeking. As the capacity to recognize truth degenerates into noise, power-seeking becomes ever more important until it's all that's left.
If power can feed on the corpse of truth-seeking, but the reverse is not true, sooner or later truth will be devoured.
> The second reason I'm writing this is because people keep asking me "should we listen to experts"?
I think you're missing a step here. Experts, especially in a policy realm, are applying a different standard of proof than you are.
Locking down your house, for example, made sense as soon as you thought it was more likely than not that it would be helpful. Maybe an even lower standard (a precautionary principle) could have applied, since the consequences of a covid infection could be so severe.
On the other hand, issuing this kind of directive for an entire region imposes costs on other people. If you're wrong about your own house, then you and your housemates voluntarily pay the price; if you're wrong about policy then everyone pays.
Therefore, mandatory orders do (and should!) use a higher standard of proof, somewhere between "clear and compelling evidence" and "blindingly obvious."
Where we fail, however, is that we don't allow experts to offer sub-certain judgments. Fauci can't go up and say "we think school closures are a pretty good policy, but we're not yet certain about it." In my opinion, this happens for two major reasons:
* First and foremost, society-writ-large is not comfortable with ambiguity. Some of this is cultural (see the general confusion over election predictions, where 55% is treated as a certainty); some of it is also personal because people have limited bandwidth to perform their own risk/reward analysis.
* Second, "wrong" predictions, even when advertised as uncertain, tarnish expert credibility. Some of this is an outgrowth of the first point on ambiguity; some is a response to an adversarial environment where other motivated actors (political opponents, charlatans, conspiracy theorists) amplify failures. This is where I think your analysis on the politics comes into play, as well, and the end result is to retreat into a defensive certainty.
In turn, this has two negative policy implications:
* The first is a retreat into process. "Experts" can't always speak in hindsight, so defensive certainty prompts agencies to develop rigorous policy for drawing conclusions. Even if the conclusion is wrong, as long as the right steps were followed then the experts can claim success. See for example medical approvals. We know now that the major yet-to-be-approved covid vaccines are more likely than not to be helpful, but experts won't bless that decision until the right steps have been followed to certify the proof.
* The second implication is a strong status quo bias. The status quo is no lockdown, so there must be compelling evidence to prompt one. The status quo is no masks, so there must be compelling evidence to recommend masks. The status quo is no covid vaccines, etc. And after said compelling evidence, the policy prescription becomes the new status quo, equally unshakeable. This leads to the oft-maligned "eggs are good, bad, good, bad" flip-flop, and it also gives undue power to whoever can define the status quo -- often a political choice that we as a society don't really consider.
Little here has to do with the political realities of placating stakeholders. Instead, so much of it derives from an inappropriate (or even impossible) demand for certainty, including our instinct to conflate certainty with correctness.
Maybe a comprehensive civics education should include a long module on Bayesian-style reasoning (maintaining uncertain priors and updating with new information) and the Kelley criterion for wagering (relating the size of one's stake to both expected benefit and expected variance).
Before finally giving up on trying to convince my non-SSC friends of the wisdom of Zvi, I gave this analogy (mostly about FDA approval):
There's a water delivery man who normally delivers at 1 mph. There's a drought and he manages to get his speed up to 2 mph, saving twice as many people. But I think while very unpleasant for him, he could have gone 10 mph. Should I be happy he managed to double his speed or sad at the shortfall of what I thought possible?
I think some amount of it is about what should our emotional reactions be to inadequate equilibria in the world. Is frustration and anger a good source for driving change? Or is equanimity and understanding of bad incentives?
By what standard to we judge Fauci: Zvi or Lysenko?
Alex Tabarrok has mentioned this a little bit, being thankful for Warp Speed, but still pressing for more.
Should we focus on feasible alternatives of conceivable ones?
Maybe the answer is both, but it's easier to digest in separate posts, or separate speakers. Zvi can depress us about all the things that could have been done, and remind us that there is so much more possible. Scott can cheer us up by reminding us that we did okay giving the existing constraints in place.
We actually can infere what Zvi would do if he became director of the CDC. He wrote it several times. 9/3 "Burn it to the ground. CDC Delenda Est." (https://www.lesswrong.com/posts/RF5HjYYLY5FWgtA9v/covid-9-3-meet-the-new-cdc), 9/24: "CDC Delenda Est" (https://www.lesswrong.com/posts/E8fp7BxpuzcwaMGtz/covid-9-24-until-morale-improves), 12/17: "CDC Delenda Est" (https://www.lesswrong.com/posts/Rvzdi8RS9Bda5aLt2/covid-12-17-the-first-dose). So I suppose Zvi is your friend "pitchfork in hand, trying to burn down the system". And maybe the Fallacy does not seem to be Marx's Fallacy, it's possibly some Marxists' fallacy and possibly some other ...ists' Fallacy, and maybe "..." can be "Trump" but it can also be "Rational"; after all, your own folks never optimize for power. Making some very smart people a world dictator may make the world a Well-Kept Garden from his/her perspective. But given that this whole rationality community talks a lot about the problem that it's really hard to define a utility function that does not turn out to optimize for something you did not want in the first place, I think it could be slightly more modest with regards to politics.
"Just get the best person in the country and let _her_ fix everything"
Which is a correct statement because women actually are better leaders, on average. See, this is why I like to read you, nuances like this one and the hundreds I probably don't catch.
In that trolley problem, it’s worth noting that the out of control trolley is going to round the bend and kill all six people either way. That’s a common problem in illustrations of the trolley problem.
Additional problem with installing your friend as world dictator is that a lot of the reason they succeed is there working with good information and if there world dictate a lot of people will be very interested in warped there information diet
Prediction markets are not better for policy issues like covid, climate change, poverty, etc. Markets are great for aggregating information on parochial issues like the price of gold, the present value of a corporation, and all the various baskets of these things. To have a successful prediction market you need a metric that measures what good-faith people really care about. Political issues like climate change or poverty necessarily use intermediate goals that are deficient.
Many are subject to Goodhart's law, which states that any statistical regularity will collapse once pressure is placed upon it for control purposes. For example, Mao maximized steel output in the 1950s to modernize his economy because the best economies had high steel production. It was a disaster because the prioritization of steel production over everything else ignored marginal costs which would be reflected in the profit-and-loss system used by decentralized economies.
Covid deaths are bad, but so too is shutting down the economy, preventing people from seeing friends and family. These costs are difficult to measure, so one can always dismiss them, but they include not merely psychic costs, but real deaths. We can give everyone an education, but the value of this declines the more it is put into a national policy, and many now have useless degrees and much debt. CO2 production is bad, but invariably it is measured piecemeal, ignoring fixed costs, and counting negative externalities for one process while ignoring the negative externalities for another.
If Biden proposed a new Middle East peace policy, a market in Middle East peace would be of no value. The same is true for any of our great issues of debate: racism, sexism, inequality, the environment.
This doesnt quite correlate w my experience following epidemiologists and virologists.
Late January I had a flu and Wuhan was a story already, so that was a way to enjoy a respiratory disease, by considering the chances of a pandemic. I was following 'This week in Virology' podcasts reading up on early cases of community transmission on flutrackers.com (did they catch all the contacts of this or that tourist from Wuhan? Oh no, this early case in Thailand infected a taxi driver, and this case in Japan was a guide a tour bus etc), There were R0 studies giving various numbers mostly between 2.5 and 3, and existing (flu) plans indicated NPIs are unlikely to control the spread of something >2. If anything, this turned out to be pesimistic. By Feb.4. I had bought a few months of food supplies (recently checked the date on that receipt), to be able to stay home during peak transmission. I was hoping it wouldn't become a big deal here in Europe before Spetember, if they managed to catch early transmission and warm weather being just aroung the corner, until Italy happened. Though w Iran (Feb 19th) it was clearly a pandemic, as epitwitter at the time explained. Also expected mortality was in the range of 0,5% to 2%, already.
Overall, I think listening to the chatter of experts, and immersing oneself in the discoruse was a good way to stay well informed about this pandemic. Quite possibly the official statements of government bodies were far more conservative, exobiting high normalcy bias.
The problem is worse than you describe, because in many cases what we are looking for is polymathic expertise. For the efficacy and side effects of psychiatric drugs, we need psychiatry and pharmacology and general medicine - and since psychiatrists are MDs, a good psychiatrist gets us most of what we need. But for pandemic response, we need epidemiology plus sociology, economics, education, politics and not just in the how-do-I-stay-in-power sense, and more other fields than I'm going to try and enumerate here. Some of that we need to know if a proposed intervention is even feasible - sociology tells us how long we can keep people locked down before they start going to private house parties or mass political rallies instead of bars. Some of it we need to know whether the proposed cure is worse than the disease - how much harm does a year of zoom-schooling do to the long-term life prospects of elementary, middle, and high school students?
But it's hard enough finding legible but mediocre expertise in one field, that we generally call the job done when we find a Dr. Fauci. Fauci, working within the limits of the political system we have, is a mediocre expert in one thing. The one most important thing of the past year, perhaps, but not more important than all the other things put together. Not good enough.
Polymathic expertise does exist, and it's not *that* hard to find if you're looking for it. Zvi, for example. Or Scott. But broadly legible polymathic expertise, is very hard to come by.
Scott give prediction markets lots of credit lately. Is there a good source that argue why we should give it this credit and not just treat it like another random idea that sounds nice but probably doesn't really work as advertised, like blockchain?
(If you think that blockchain is not just another trend, same question)
Re: not worrying more about addiction to Adderall than alcohol: Most people don't drink every day [citation needed]. They drink occasionally at social events. Adderall is recommended to patients to be taken regularly for some period, isn't it? If I drunk every day, I'd definitely be worried about the possibility of getting addicted.
I'm a layman who knows basically nothing about either Adderall or alcohol addiction. My point is that your comparison is not easy to interpret for someone who has never drunk regularly, and has never had a reason to investigate how much one can drink regularly without getting addicted.
What about Medscape. Appears much more professional and informative and less "political". I generally use it for medical questions. Haven't tried it for mental health issues.
“a kind of bumbling careerist with a decent understanding of epidemiology” is harsh for someone who was lead editor of Harrison’s. I mean, sheesh. Dude’s not a saint or Einstein but he seems pretty smart. I mean, not as smart as the economists who all seem to know marginally more than anybody about anything...
I agree about the general problem of power inviting corruption, but I think you're being far too kind to the current system.
Burn-it-downism actually works a lot better than I'd naively expect. The Bolshevik Revolution led to Stalinism, yes, but Stalinism involved possibly the most successful economic development program ever - mortared, of course, with a great deal of blood, but not the implosion one might have predicted. Even attempts at outright anarchy haven't ended with everyone starving to death. The bigger issue is violence, which revolution always causes and which our current systems are reasonably good at preventing. Of course, an anarchist would argue that the entire thing is built on the *threat* of force, but being held at gunpoint is still better than being shot.
But - with simulacra levels and all that - we currently have leaders that are detached from reality entirely, which is the reason for the uptick in populist sentiment, with the aura of revolution and violence that's always implied. When Congress has an approval rating below 25% for ten years straight, it really shouldn't be surprised to get a lynch mob at its gates. You talk about Fauci not being Lysenko, but Lysenko was very much the exception in the USSR; Fauci isn't Korolev, either, and I'd say Lysenko has closer analogues in the modern American public service than Korolev (something something bioethics). The Attila the Hun comparison is misleading because Attila was leader, not expert, and honestly I'm not sure Trump is all that far from an Attila in temperament. Not that we actually know much about Attila, historically. I could say something about Tamerlane and US foreign policy, but I don't really have any well-formed thoughts on that. The point is that the positive elements you're talking about are institutional successes fifty years old; I mean, even Fauci himself isn't exactly a new appointment. Possibly this is because the Cold War USA faced an actual threat in communism, and before that the Axis, and therefore had to acknowledge physical reality to fight them.
And you could say that, sure, our society's current paralysis is bad, but it's still better than risking things on any big changes because those can backfire, and basically that what we have is good enough. This is called conservatism. The problem with that, even if you're of the opinion that <gestures at planet> this is fine, it's not *stable*. Technological change leads to social change, sustainability is more than just a buzzword, relying on imports from increasingly poor countries has a natural endpoint, and if it takes ten years to repair one highway intersection our infrastructure will eventually fail entirely. We need utopianism in our discourse - we need multiple competing utopias - because otherwise we're reduced to using the literal worst expert ever as the bar our experts need to jump. And we probably need the threat of a mob with pitchforks (preferable distant, but with present dysfunction a more immediate one seems to be required) to make our elected and unelected leaders actually optimize for something besides power - in layman's terms, do their jobs.
This is a fantastic line of inquiry. But, doesn't it bump up against anti-culture war imperatives? It has all the trappings of previous ones: expertise vs. non-expertise, authoritarianism vs. individualism, and seems to also land on standard left vs. right spectrums
This isn't kind, but it's true and I think adds something specific: the two times I've met Zvi in person (at rationalist-adjacent meet-ups in NYC, three years apart) I found him arrogant and unpleasant, to such a degree that it really stuck with me. Like, loudly declaiming his opinions and derisive toward anyone who disagreed with him. I'm generally pretty unbothered by non-neurotypical people, but that didn't feel like what was going on. I think it was just a difficult, non-collaborative personality.
I haven't read his blog much so maybe he's much easier to hear in writing. But I also wonder if these traits are somewhat correlated with being right about something early and loudly and also anticorrelated with the skills it would take to affect the world outside of, say, the rationalist-adjacent blogosphere. I also worry it leads to worse, un-dampened outcomes when he's wrong. Somewhat related to the broader "competence" point that others mention.
(This is really an unkind thing to say, so Scott please feel free to delete if you think it doesn't add enough to be worth it.)
I remember reading a really good article you wrote about a book you read on Marx. It was really insightful to me, but I got the feeling that communism could be fixed by more carefully thinking about the problems on a system like that, maybe using simulations, etc. But it seems you heavily dislike communism and I would like to understand if you see a problem with the core idea of the workers owning the means of production and also the redistribution of wealth.
>When Zvi asserts an opinion, he has only one thing he's optimizing for - being right - and he does it well.
>When the Director of the CDC asserts an opinion, she has to optimize for two things - being right, and keeping power.
The senior officials in the CDC are (both reasonably, and according to their own claims,) optimizing for *saving lives* and keeping power. Sometimes, the thing they do to try to save lives is shut up about true things, or even say false things. That seems to be what happened with masks; they were trying to manage an economic supply/demand system to make sure hospitals and healthcare workers, who were at higher risk, had masks, and that the general public didn't go out and buy them. Was this a bad idea? In retrospect, obviously it was - but more critically, truth is a harsh mistress, and it seems that you can't serve two different goals, even when the first goal is honesty, and the other goal is public health.
This post really exemplifies the saying "damning with feint praise." Sure, the experts are consistently wrong but that's only because they are optimizing for something other than being right! Come on, man. I don't know why you defend these people.
>The Director of the CDC reads those same papers. But some important Senator says that if airborne transmission is announced, important industries in his state will go bankrupt. [...] So the Director puts out a press release saying the evidence is not quite strong enough to say airborne transmission definitely happens, and they'll review it further.
Not saying this isn't true, but I wonder why these pressures aren't balanced out by countervailing pressures. The correct press release could harm the important Senator's state, but should greatly benefit the average citizen, so the important Senator's objection should be counter-balanced by the support of many other Senators (who should profess their objection to fire the CDC Director). This is why we believe representative democracy has a chance to benefit citizens at all. If other Senators will not resist the Senator's pressure, it means they need his support for something else, more important. In the end, this should theoretically (ideally) work out for the best compromises to benefit everyone. And when something is very important, like covid, we should expect to see fewer decisions related to it being compromised for something else.
What is the reason it doesn't work out that way? Irrationality? Politicians representing not the interests of their constituencies but their donors/lobbies? Rigid structures, e.g. the Senator in charge of CDC appointments was decided before it was so important, and now can't be changed?Something else?
OK - so the system as a whole produces mediocre expert advice - the question is can we (whatever group smaller than the general popupation that 'we' represents) develop a system that would give us better expertise?
Feel like the driving factor for a lot of this is chance. There are a vanishing number of competent people and a vast sea of incompetents. If you have a mechanism for competence selection that is in any way imperfect, you're getting an idiot.
"Maybe expertise is a sham, and a smart guy thinking for five minutes can outdo a decade of working on a PhD."
I think about this a lot with respect to my field: urban mobility. I have a PhD in operations research / transportation, I work at an urban mobility startup, and I spend a nonzero fraction of my free time learning about urban mobility. I know a lot about this area. And yet, I know approximately nothing that you couldn't realize by looking out the window and thinking sufficiently deeply about what you see.
But most people - even in the industry - haven't realized many of the things which are realizable just by thinking hard. So can a smart person thinking for 5 minutes outdo a decade of learning? I suppose for a sufficiently smart person, yes? But sufficiently smart is a high bar. In mobility, time investment seems to serve as a substitute for something like "raw smarts" or "wisdom," a way to crowd source deep thoughts. In practice, everyone needs at least some time investment.
(Example of something that most people don't seem to have internalized, but should be obvious to us all if only we were all wise enough: for most intra-urban origin-destination pairs served by multiple transit modes, most of the modes should be approximately equally costly (including non-monetary costs) most of the time! E.g. if it were easier and faster and cheaper to go A -> B by bus than by car, people would shift car -> bus, the bus would slow down and become more unpleasant, traffic would decrease and thus car travel times would decrease, etc.)
I think not all fields are like this. E.g. no amount of staring out the window and thinking hard could reveal to me that human cells use DNA to store genetic information.
Just on a really narrow point it seems like both your recs and the recs from WebMD could be dramatically improved by adding a dime numerical score of how bad the side effects are. This allows one to both give a detailed summary of all the side effects and not require denying their importance while still giving a general sense of danger level relative to other drugs (forcing those who say this drug ruined a family members life to essentially debate the ppl claiming other drugs ruined lives they know).
> rather than being minimaxed for COVID-prediction
So this essay is great, but you've hit one of my pet peeves; you're using the word "minmaxed" wrong.
Minmaxing means something pretty specific; it doesn't just mean "optimized". For example, when I evaluate positions in a chess game, I assume my opponent makes the worst-for-me ("min") moves they can make and that future me makes the best-for-me ("max") moves future me can make. In short, it's only natural to use "minmax" to describe a situations where I'm trying to do something despite the actions of something else which I'm assuming is trying to thwart me. (Granted, technically we could call it minmaxing where there's a mathematically trivial opponent who doesn't do anything, but it's still a misleading word choice, like calling something a pentagon when you totally know it's a square and then defending your word choice by saying that one of the five sides has length zero.)
If you're trying to do something but there is no opponent (like covid prediction), the word "optimized" is a better fit than "minmaxed".
Here’s how I use medical article databases. I’ve had x problem which the small-town PCP was ignoring. So hours on WebMD and similar gave me enough background that the next time it acted up I went to the ER and said “I think this might be a hernia” (they did an ultrasound, I was right.) Did that for a few other things. We can’t all access actual experts, so I need just enough info to get through to the doctors in front of me who can treat - prescribe, order tests, etc. Those databases are great for that.
I felt like crap for a month and couldn’t get any answers from docs and hit the databases. Compared bloodwork and got “why do they say nothing is wrong when the Internet says these numbers are metabolic acidosis?” Then more googling and it was possibly a side effect of a medication, so I did a taper myself and felt a little less crappy. I will never know if I was “right” but I burned less days feeling like crap. The databases are like a 2nd amendment for medical knowledge. We can arm ourselves. That being said, more info is better.
Excellent article highlighting the sloppy content on WebMD. It's a trash shite selling snake funded snake oil. Nobody with an iota of critical thinking takes this site seriously. WebMD is literally a lousy joke of the online world.
I don't think it's really a corruption thing that led Fauci to F up. Birx clearly wanted to stay in her job and thought that her being in place was better than Trump installing a 100% yes man. For Fauci, and the rest of the CDC, FDA, and NIH I think there's this massive status quo bias that is hard to overcome. Working in pharma it is clear that regulators and doctors want to be really really confident about something before they do it, even if the risk of them being wrong is no way near as severe as not doing anything. I think that's what happened with lockdowns, masks, at home antibody tests, etc. Until you could get a sufficient amount of data to be bulletproof (for non-delusional people/denialists) they won't do anything.
I don't think anyone thinking COVID was the killer virus that the WHO made it out to be at the beginning can reasonably judge experts.
Well, maybe not. I suppose it's the same conundrum for Scott Alexander as for Fauci. If he ever openly says that COVID was a false alarm (which it was) and the response a massive civilizational fuck up the likes of which you won't find in the history of the planet until now (which it was and is), he would lose a lot of his supposedly "rational" audience.
> Expertise isn't a sham.
Tetlock begs to differ: https://spot.colorado.edu/~huemer/papers/passivity.htm
"In light of the ignorance of typical political leaders and members of the general public, we might be tempted by the idea of rule by experts, as in Plato’s Republic. 18 Unfortunately, when it comes to descriptive social theory, even the experts’ knowledge is unimpressive, as demonstrated recently by the social psychologist Phillip Tetlock. Tetlock conducted a fifteen-year study in which he collected tens of thousands of predictions from hundreds of political experts concerning matters within their areas of expertise (for example, would the economy slide into recession, would the Soviet Union survive, who would win the next Presidential election, and so on). Tetlock’s finding, in brief, was that the best experts did only slightly better than chance at predicting outcomes. When asked to assign probabilities to their predictions, experts proved systematically overconfident; for example, events predicted with 100% confidence happened less than 80% of the time."
It's useful to hold the tension between
A: We're monkeys who more or less only *just* upgraded from feudalism. Things are going well considering we just need to keep iterating
B: Decisions get made that so clearly go against everyone's best interest on the regular. We can glimpse a better world and what we've build (while keeping us safe) is blocking us from moving forward.
I think the more fully we can recognise both the better, but A is a lot less represented in the current discourse, thanks for the reminder.
onically, with the WebMD example, if you want the most precise I formation without political/legal intfluence you would choose an even more specialist (or professional) resource, such as the BNF (for drugs in the UK, which definitely doesn't list all side effects; I find it odd when a patient asks me about a side effect of a drug that I've prescribed for them that they've read in the leaflet yet the side effect is not significant enough to warrant listing in the British National Formulary so when I look it up I can't find it), fpnotebook, CKS etc
So? Is it true that disproportionately many borderline personality cases are young women with lots of piercings and tattoos? What does that tell us about diagnosis and etiology?
Yes! Keeping power vs. being right is a great constraint to point out. I would maybe suggest expanding "being right" to "competence" though -- I'm willing to buy that Zvi is really good at being right, but he might not be nearly as good at hiring, leadership, working with other people, keeping schedules or any of the other things that are sometimes completely unimportant and sometimes absolutely critical to a job.
The head of the CDC job and benevolent dictator in particular probably require a LOT more in terms of managerial than predictive skills. Zvi would, I suspect, be most useful as an advisor to the benevolent dictator than the dictator himself. It's a shame that governments don't employ advisors anymore.
Cool hominid party trick: The world is abominably complex, so we call up organizational superstructures from the depths of meme hell and put them to work managing said abominable complexity. This is adaptive, apparently.
The dread blueprint of the FDA lurks latent in the collective unconscious, waiting to manifest and consume the souls of unsuspecting hunter-gatherers.
Are you referring to the articles on lorienpsych.com as your small database? I thought you meant more like a weighted average of patient ratings like when you recently compared different amphetamines.
Best post since you came back online, Scott!
Yes - the amazing thing is that our society works far better than we have any right to expect. (And far better than it would if 99% of reform proposals were adopted.)
And...just put a link to this post on your database (BTW, where is it?).
That *ought* to be sufficient, but of course it won't be for some people. You just have to live with that - valuable info online will help many people and hurt a few who misuse it. But it does net positive good.
The alternative is to become WebMD. They're *cowards*. They know they're destroying the utility of the information they offer, but they have shareholders to appease and lawyers to fear.
You can, and must, be braver. Probably you won't get sued over this, and if you do you'll probably win. But you're in the lucky position - unlike WebMD - of having a large fanbase who appreciate what you do and will bail you out if all goes pear shaped. I can (and do) guarantee that you'll never have to worry about financial hardship because of losing a lawsuit over this stuff - somebody in your fanbase will ensure you have a well-paying job (in the unlikely event you need help).
"What I sometimes call Marx's Fallacy is that if we burnt down the current system, some group of people who optimized for things other than power would naturally rise to the top. Wrong. People who most brutally and nakedly optimized for power would gain power; that's what "optimize" means. The interesting thing about the current system is that, after millions of very smart and altruistic people have contributed to it over generations, sometimes gaining and keeping power within it is modestly correlated with being good and right."
This is so brutally insightful and hilarious at the same time.
I agree with this analysis of "politics vs being right/doing the right thing" but it creates another obvious problem: the good of being right in the moment is bounded, but the good of being right in the future is unbounded
One of my favorite scenes from the Wire is when the newly-elected mayor of Baltimore has a problem. He needs money to bail out the public school system, which he can get from the governor, but if he does, he becomes "the mayor whose failing public schools forced the governor to get involved and the state as a whole to contribute tax dollars to bail him out," thus ending his future ambitions of becoming governor himself. And of course, once he is governor, his ability to do good things will be much greater than his ability as mayor. So he decides to let the school system continue to collapse to preserve his opportunity at future success, and future good.
I think almost every politician or person who has a political component to their job has made this calculation at some point in their lives. And in some sense its hard to say they are making the wrong call. Because if that fictional politician goes on to become President and signs a bill which ends child poverty, whose to say it wasn't worth a few thousand kids in Baltimore?
I don't have a solution to this problem, except to say that at some point everybody in every position that might even have a glimmer of power has to ask themselves "what's the absolute amount of good worth sacrificing today for the prospect of being able to do good tomorrow?" And to think really hard about the answer.
You are missing the HCQ elephant in the room (and the Ivermectin rhino too....). This makes your otherwise excellent post dismayingly close to worthless, until you can come to terms with the extraordinary number of lives lost due to the corruption of the process that Fauci, putting it enormously too mildly, allowed to occur.
I disagree with your analysis — that corruption/self-preservation explains why Dr Fauci does not always say the truth. Externalities and/or the possibility of being misunderstood are more important factors IMO. For instance, while an unknown blogger can say in March 2020 "we should all wear mask now", Dr Fauci or the NYTimes cannot say the same because they're afraid it will push people to hoard masks.
This is a really excellent post. It certainly makes me feel grateful we have a magic machine for consistently extracting pretty-good people with pretty-good ideas out of the population.
That said, there's a related issue: Why has the magic machine's performance seemingly decline over the course of the last ~80 years? Maybe it's an inevitable result of Moloch (for example, because as you optimize more aggressively, the gap between the most competitive and most good-producing ideas grows, or something). If so, too bad, guess we're screwed. If not, then we could sure benefit from figuring out what's causing the machine to function worse.
(As this post argues, any tinkering we do trying to fix the magic machine should be very cautious, because it sure would be a shame if something were to happen to it.)
So the central claim is that the experts at the top are genuinely super-capable, but corrupted by incentives. It's a fine hypothesis, but what evidence is there in favor of it? In academia at least I can point to an ocean of crappy research whose existence benefits no-one, so you can't blame the incentives. A Stanford epidemiologist attending a conference in March tweeted:
>At AHA #EpiLifestyle20 in a room full of epidemiologists and scientists and not ONE person is wearing a mask. Lots of hand washing & sanitizer stations. If 800 epidemiologists aren't wearing masks, you don't need to either!
We're talking about (tenured) people personally putting themselves in the way of harm here, can we really blame corruption and incentives?
I think you need to more seriously entertain the hypothesis of genuine incompetence.
> "A machine which takes Moloch as input and manages - after spending billions of dollars and the careers of thousands of hard-working public servants - to produce Anthony Fauci as output. This should be astonishing"
I think this is indeed a v good approximate account of our situation, & it shows both how well we usually do, considering how bad things could go, & also how tenuous is our toehold above the abyss. So far we've avoided the really terrible outcomes (the last four years of D.T. notwithstanding), but there have indeed been some terrible outcomes for some, & some very close calls -- some of which we know about (Bay of Pigs, Three Mile Island) & some we don't. There have also been some heartbreaking misses (eg the history Nathaniel Rich details in his (doubtless imperfect) _Losing Earth_). How well will our luck hold as we move on into 21st Century Climate Change -- to name just a single one of the many fun, um, opportunities staring us in the eye? I too am a reformer not a revolutionary, if I have my choice, but that choice is "all other things being equal," & more & more, they aren't.
Is there some reason why WebMD couldn't include information about how common various side effects? The severity of reported side effects?
A very good piece and essentially correct, unfortunately. The one part I'm not sure of is the implicit assumption that even mediocre expertise feeding into a political decision mechanism is better than no political mechanism.
Consider Covid. The advice that was given by Fauci and others was probably better than what most people would have done with no advice, better on average than what most would have done (although possibly worse than what you and I would have done) relying on whatever sources of information they trusted. On the other hand, the vaccines we are now getting took about a week to design and eleven months to get FDA approval. In a world without the FDA and other political constraints, a month of challenge trials would have shown that they were effective and not very dangerous, after which they could sell them to all comers as fast as they could ramp up production. It's hard to believe that that wouldn't have saved a couple of hundred thousand lives in America, a million or so in the world. So if you look at the overall effect, it may be negative. We might be better off with the same level of mediocre expertise and no levers of power for it to be applied to.
At a tangent ... . Your description of Fauci reminded me of my conclusion about Nixon. There's a bit of the tapes where he is talking with one of his people about the idea of ending the interest equalization tax, which was a restriction on capital mobility. Nixon points out that one or two people have said doing it would be a good idea (true). The aide points out that there are no votes in it. Nixon says (by memory, not a quote) "What the hell. Let's do something just because it's good for a change." My feeling was that he had spent his life getting power in order, as he saw it, to do good things, and then found that he had to use the power almost entirely in order to keep power.
The current dramatic case of people choosing the wrong expert to believe in is the election. Something like a quarter of the population, if polls can be believed, thinks Trump really won. Their experts say so — and they have concluded (correctly, although not in this case) that the official experts can't be trusted.
> Even if this flattering story is true, it doesn't scale... Compared to the median person who disagrees with the experts, the experts look pretty good.
This is an interesting way of phrasing it. I would say that the problem you're pointing at - the incentives for getting and holding power mean you get mediocre experts is, _itself_, a scaling problem.
Here are two thought experiments to illustrate my point:
a) Imagine there's a single unified government for the entire solar system, with population in the hundreds of billions of people, and levels of wealth substantially beyond what's available today. Billionaires are a dime a dozen - any pop star or athlete easily makes it there. There are people who own entire planets. Are the incentives for being honest and accurate _more_ aligned with getting and holding power, or less? How corrupt would you imagine the official councilor on health, for the entire solar system, which includes multiple alien races and sentient robots and the bio-corporations of titan?
b) Imagine that the federal government basically doesn't exist and almost all power - taxation, trade regulation, everything else - falls on the individual states. Each state has its own experts, who compete for power in _much_ smaller games. How corrupt is the state health controller for each state?
It seems to me that the more people involved in a competition, the greater the disparity between 'strategies to get and hold power and status' and 'strategies to be accurate.
The lesson of moloch, to me, is that we should try to avoid, as much as possible, massive, winner-take-all games of power. _That_ seems to be the thing that doesn't scale - i.e., it works OK when you do it with smaller numbers of people, but it falls apart as you start to add more and more people to the system.
There's likely a lower limit here too - with too few people, there's not enough of a pool of capable people to select from. With too many people involved, the selective pressures get so insane that the winners must be corrupted.
I love that you conclude that we are "insufficiently grateful" for Anthony Fauci, while belittling his free-agency in the same breath. It's surprisingly effective, and somewhat hilarious in light of the Fauci-is-Jesus-reincarnate memes that people are worryingly sincere about.
Funny that you say Trump is some sort of specifically bad outcome when the alternative was Hillary. I'd say the system failed by the time it presented either of them to us as a choice.
By way of nominative determinism, it's worth noting that many people die to warfarin'.
"Benevolent dictatorship, obviously, just get the best person in the country and let her fix everything. But everyone realizes this is easier said than done; the procedure to pick the best person is corruptible. "
One of Mary Renault's historical novels, _The Praise Singer_, is in part about tyranny, meaning, in the Greek context, popular dictatorship. One of the tyrants she describes, Pisistratus, is benevolent. In her version, he was the younger lover of Solon, who created the Athenian legal system. He tells the protagonist that when Solon finished everyone liked it, except one thing — everyone wanted a change to favor him. Solon's response was to leave Athens, because if he wasn't there they couldn't make him change it. "They keep his laws — I see to that, who could have given them laws they liked less well." The tyrant has been kept honest by the memory of his dead lover.
It's a powerful scene, but not a very reliable procedure.
> What's the best form of government? Benevolent dictatorship, obviously, just get the best person in the country and let her fix everything. But everyone realizes this is easier said than done; the procedure to pick the best person is corruptible. At one point we tried a very simple best-person-picking procedure that really should have worked and ended up choosing Donald Trump as the best person. I'm still not really sure what went wrong there, but apparently this is really hard.
I know you’re having trouble being serious here, but no, at one point y’all tried a very simple best-person-out-of-two-far-from-obviously-close-to-the-best-in-the-country-picking procedure. The prior picking-which-two-people-to-choose-from-procedure was very complicated and obscure and probably very corruptible.
It's common to hear the following:
1. If group I dislike is disproportionately responsible for problem X, it's group I dislike's fault.
2. If group I like is disproportionately responsible for problem X, it's a society's fault, or the general incentive structure, etc.
I see this in the response to corona of many people here. Who is to blame for corona spreading because people refuse to wear masks? Dumb people. I personally agree fully, see https://alexanderturok.wordpress.com/2020/07/08/march-of-the-clevons/
But then, who to blame for the lack of prediction markets and challenge trials? Well, that's a problem with "society" and "incentives."
You tell a hypothetical story about what would happen if Biden appointed Zvi. But what actually happened when Trump actually appointed Scott Gottlieb head of FDA? Why did he only last two years? Was he pushed out as a political cost, as in your story? Did he burn out fighting with the organization, a different but parallel story? Did he have power to control the organization, or would it not have made a difference if he had stayed for another two years?
Maybe list all the side effects but in a font size proportional to their frequency?
WebMD is useless but your specific example is an unusual choice. Aspirin is relatively safe for healthy people but less so if you have liver damage. The Wikipedia article for salicylate poisoning says there were 20,000 cases in America in 2004.
I like the idea of cocooned experts, incorruptibly sealed while they print out opinions.
Promising instances:
• Scott and his substack
• a wealthy patron paying the living expenses of a writer
• forecasters on prediction markets that capture some vig
• the crop of very smart people who made, are making a killing [in crypto, other investors] and can pursue their own ends, e.g. fuck you money
• communities like LessWrong where status is accorded to thoughtful, considerate correctness
Great post. The last line -- "prediction markets would still be better" -- is idiotic [1], but we can't have everything. Scott isn't perfect, but he is better than we deserve; legibly mediocre.
[1] If prediction markets were so great, we'd see them happening *somewhere*. Gambling is legal in Canada; you can't blame everything on anti-gambling laws. Where are the Canadian companies that use prediction markets to make decisions? They don't exist. You think they should? Go to Canada, start a company that uses prediction markets instead of a CEO, become a billionaire, and THEN I will believe that prediction markets work. Until then, they are just a way to signal "rationalist community good".
I'm surprised that no one has yet mentioned Zeynep Tufekci (https://zeynep.substack.com/people/299814-zeynep) as another writer who seems to have consistently been righter than the experts. She's been my go to for COVID information for quite some time, and probably the person who's most influenced my actions and what I recommend to others.
One assumption that I'm not entirely sure I agree with is this one:
"If you expose the plan to politics, the politics will drag it in the direction of being worse. Every feedback channel you open up is a way for somebody to attack you. "
Some massive assumptions involved are that:
1) You will consistently have specific illegible experts
2) These experts have satisfactory self-contained knowledge
I agree with the mechanism that politics CAN drag things in the direction of being worse because it inevitably creates channels for people to jam questionable inputs.
But politics also has the ability to create channels for people to jam in relevant inputs. Most people can't, don't, and won't develop the approach needed to balance a proposal against multiple different objectives. Part of this is maintaining power. However, Fauci isn't strictly acting as providing medical advice, but rather his pronouncements have economic & sociological impacts.
I think the essay creates a good argument for Plato's Republic, but when you keep in mind that Plato's ruling class must exercise judgment in lying, and that the web of information in a modern society is an order of magnitude more complex, etc, etc. Then... I'm not so sure. I am not sure if I want to sign off in an idealized theory even.
Let's put it another way: does Fauci & the CDC by itself have sufficient expertise to decide to lockdown the US economy? The answer is probably no, because why shouldn't the CEA (Council of Economic Advisors) get a say in that? That's a really big decision that hit's their domain. And maybe we need the APA (American Psychological Association) because this really could cause psychological stresses that the CDC may not be as expert in. And... at this point, we may also be talking about a value judgment, which requires some sources to provide those inputs (ethicists? representatives of the US population?) And when you realize all of the people who should legitimately be consulted to make a complex decision, you now have politics.
That leads to the idea that politics is a natural result of human interactions involving complex decisions. Some politics is good & some politics is bad. But it needs to exist, and it isn't so easily separated into "impure power-grabbing". Coordinating with all stakeholders is "maintaining political power", but the stakeholders legitimately exist! And pretending they can be replaced (even in theory) with the ideal dictator means that we somehow don't believe there are other stakeholders, or that we think they can be ignored. Sometimes bypassing stakeholders works!! However, it really can blow up quite horribly!! "Move Fast and Break Things" is a very productive & very destructive slogan.
All of which is to say: Scott this is a great post. Can you please tell me if Democratic Republics are good or bad? This post has implications and I just gotta know.
I like this post in general, and I agree with the central thesis. However (as someone who put nontrivial thought into both masks and coronavirus case/death estimates, see eg: https://pandemic.metaculus.com/rankings/), I disagree with the causal mechanism is that corruption caused accurate people to become less accurate in public-facing ways.
I think it's more likely that the corruption caused counterfactually competent people to be less accurate, even in their own thoughts.
Fuller explanation on Twitter:
https://twitter.com/LinchZhang/status/1357839902163931136
Speaking of Lorien Database changes, you might want to make the UI a bit more useable on mobile. There was some charcoal-on-prussian-blue text and stuff last I checked.
Also is it true that formites are not a vector?
I think you’ve conflated two forms of “expert,” the advisor and the policy maker.
While a policy maker is evidently subject to the sort of public pressure you describe, one would hope a president could get better advice in private, and that experts whose public facing role limits them might still be able to speak more frankly “on background” to reporters, politicians, etc.
I... do not get the sense our expert class recognizes their failures in private. It is not clear to me, for instance, that the FDA’s top leadership suspects that their refusal to consider and approve the Astra-Zeneca vaccine is effectively mass murder. And it isn’t clear to me that in private public health officials understood the threat of COVID-19 until remarkably late. That’s more worrisome than that our system puts pressure on the most public-facing roles.
Or, put another way, we are selecting almost entirely for “can handle dealing with the public” rather than “gives the best possible advice” in the filter that selects for our public experts.
My current nomination for "most trusted pandemic amateur" would be Zeynep Tufekci. I'm skeptical that Zvi always optimizes for being right, versus coming up with hot takes to make us angry about how people are handling the pandemic? (Maybe we should be angry, but I need better evidence.) Still a good source of link and questions, though.
In particular, I am wondering if anyone has good sources or explanations for how the FDA is making decisions about vaccines during the pandemic? This seems mysterious and many have questions, but I don't think I understand it yet.
> (but prediction markets would still be better)
I'm absolutely not sold on this. Perhaps you already have an article explaining why you believe this, or plan to create one?
"The experts successfully swooped in and saved us from all of that, figuring out which way the wind was blowing only two weeks later than competent amateurs. This was a useful service. Without the experts, things would have stayed open forever."
I expect that people would, in Bill Gate's memorable phrase, "notice the bodies piling up in the corner" eventually.
For something in a loosely related vein by Zvi himself, see https://thezvi.wordpress.com/2017/10/29/leaders-of-men/
It's my opinion that Fauci has done a terrible job, but I genuinely don't know anyone else who would've done better in this nightmare. We have a crisis of expertise AND implementation.
Zvi probably also has a very permissive peer group. For normal people it's still a multi-criteria decision-making problem: maximize rightness together with minimizing slights/offense/loss of status within the peer group. Whatever experts are proclaiming also gets used as input and it gets difficult to deviate from it depending on how peers process it. At least us normals don't have to deal with lobbyists or activists that want to remove us, I guess.
> later-proven-not-to-exist fomite-based transmission
I know this is a little off-topic, but did that ever get proven? Last I knew it was in the "weak evidence that it isn't very common" category.
I love it when multiple things I read converge around the same time even when they weren't created at the same time. I just rewatched the classic CGP Grey video, Rules for Rulers, yesterday - https://www.youtube.com/watch?v=rStL7niR7gs&ab_channel=CGPGrey, which discusses what people in power need to do to stay in power, and has a nice similar energy to this post.
It was just yesterday that I was commenting on Robin Hanson's post (though I was also thinking about a recent podcast he was on) about Experts Versus Elites*, pointing out that experts got COVID-19 wrong initially and elites had to correct them. This would have been a good post to link to.
* https://www.overcomingbias.com/2021/02/experts-versus-elites.html
One takeaway that I get from this is how important anonymity can be, sometimes empowering people to speak the truth which would otherwise have too great negative consequences for themselves and the systems they represent. While we probably don't want anonymous people running our country, it's definitely among the simplest ways to bypass some of the unintended forces of corruption mentioned here, and is, for example, why voting is at least intended to be anonymous (and ideally would be so in our government as well, just with cryptography to provide all of the various assurances and verification that we'd want).
Excellent post. Although it sounds like one solution would be to bring the requirements of "being in power" more in line with "being right"; e.g., by calling out, defunding, demoting, firing, cancelling subscriptions to, etc., institutions and individuals who have done markedly worse than competent amateurs and have caused harm in the process. So "taking to the streets, pitchfork in hand," but to a lesser degree.
Are the problems Scott is addressing here [just] special cases of the general issue of how to juggle social signaling with efforts at accuracy / rationality / reliability? I'm not trying to trivialize it -- I think this problem is one that repeatedly arises both here (and back in the day at SSC) and in lots of places both in the Less Wrong crowd and more generally: how to raise questions *at all* when certain questions or certain responses or certain thresholds of reliability, etc, are also *signals* that mean things like "I am on your side," "I am powerful and you should be on MY side," or "I am NOT on your side and you should fear me." I am thinking of certain social justice questions, for instance, but one could multiply examples ad nauseam. Part of Scott's point, I take it, is that questions of scale play into this issue complicating the matter.
I want to offer a small correction. You claim: "What I sometimes call Marx's Fallacy is that if we burnt down the current system, some group of people who optimized for things other than power would naturally rise to the top. Wrong. People who most brutally and nakedly optimized for power would gain power; that's what "optimize" means."
At first glance, this claim seems by definition correct. But it does not *necessarily* hold. If there is ex ante heterogeneity in attributes. For instance, if we burnt down the current system and declared that the person born on the birthdate determined by some unpredictable quantum fluctuation has all the power, it would be *difficult* to overcome your ex ante handicap (birthdate in this case).
While that's a pedantic example, the same could imaginably be true if we only allowed the most excellent and well-respected academic economists to rise to power. Even for an optimizing power-hungry sociopath, it is unlikely that they will be able to rise to the top--not enough natural talent.
You could instead say that part of what needs to be optimized (natural talent) is out of the individual's control space.
For a statistician, if the variance in the latent variable that determines power is primarily driven by non-optimizing behaviors (primarily natural talent) then optimization has less of a potential role to play.
All that said, I agree with your overarching point, for most practical situations such as Marx's, what I said is nonsense--but in the interests of steelmanning, I think the theoretical idea that one can build a society that isn't driven entirely by power-hungry optimizers is not as self-contradictory as it might seem.
"Anthony Fauci is neither Attila the Hun nor Trofim Lysenko. "
Sure Jan: https://www.today.com/video/dr-fauci-on-coronavirus-fears-no-need-to-change-lifestyle-yet-79684677616
https://www.spglobal.com/marketintelligence/en/news-insights/trending/lfG-uPOaVGA2qlV4OSZnQg2
> . Citizens Against Lockdowns argues that the CDC already screwed up by stressing the later-proven-not-to-exist fomite-based transmission, ignoring the needs of ordinary people in favor of a bias towards imagining hypothetical transmission mechanisms that never materialize;
I know you're making an example and trying to be funny, but this is a really important point that needs to be made:
Most of the CAL-types (at least, the normal ones who aren't walking conspiracy-theorist stereotypes) aren't mad because "they said fomites are real but fomites don't real". They're mad because the experts figured out fomites don't real six months ago _but we're still going through all the surface-decontamination safety theatre_. They aren't mad that experts got it wrong. For that matter, it's not even clear to me that they're mad at the experts at all. They're mad that the people who constantly say that they're "following the science" didn't update their beliefs when the science came in, and it's been six fucking months of doing all this ritual bullshit that has provably zero effect on anything.
Another CDC vs. Zvi difference is that the CDC is doing public communication to a mass audience, often filtered by the media / the public discourse. Whereas Zvi is writing directly for a niche audience of people who are willing & able to put in an unusually large amount of time/attention/thought to build detailed models. Doing mass communication pushes towards:
Just sharing your conclusions rather than the details of your reasoning
Giving simple guidelines/approximations (e.g. "wear a mask", "6 ft apart") rather than more details or the underlying model
Not sharing things that you're more uncertain about
Trying to keep a consistent story rather than changing your views fluidly
Zvi doing mass communication would face some challenges that the real Zvi hasn't had to face (and which a Zvi-in-a-cave setting policy might be able to dodge). Just replacing pages on the CDC website with Zvi blog posts would create some problems.
I really appreciate this analysis. I only wish you emphasized more the relevance of regulatory capture. Whereas the FDA, CDC, and NIH are supposed to regulate the pharmaceutical industry, they very often are a mouthpiece for that industry, having been captured by it.
The generic drugs hydroxychloroquine and ivermectin, for example, have been successfully marginalized, despite abundant evidence of their efficacy against Covid-19. See hcqmeta.com and ivmmeta.com for summaries of the research to date and links to the original studies—which are abundant, at this point. (Really, please do look!)
For hydroxychloroquine: "The probability that an ineffective treatment generated results as positive as the 198 studies to date is estimated to be 1 in 1 quadrillion (p = 0.00000000000000083). Early treatment is most successful, with 100% of studies reporting a positive effect and an estimated reduction of 66% in the effect measured (death, hospitalization, etc.) using a random effects meta-analysis. ... 91% of Randomized Controlled Trials (RCTs) for early, PrEP, or PEP treatment report positive effects, the probability of this happening for an ineffective treatment is 0.0059."
For ivermectin: "100% of the 37 studies to date report positive effects. Early treatment is more successful, with an estimated reduction of 82% in the effect measured using a random effects meta-analysis. ... Prophylactic use also shows high effectiveness. 100% of the 19 Randomized Controlled Trials (RCTs) report positive effects, with an estimated reduction of 72%. ... The probability that an ineffective treatment generated results as positive as the 37 studies to date is estimated to be 1 in 137 billion (p = 0.0000000000073)."
It requires a certain amount of research-savviness to evaluate studies like these firsthand, rather than relying on media reports and authority opinion. But for those who do, and who proceed to evaluate the relevant research firsthand ... whoa!
Given America's dire need for cheap, safe, and effective early treatment, I find it hard to explain the marginalization of these repurposed generic drugs any other way, than through capture of the relevant institutions and regulatory bodies.
The NIH canceled their large RCT for hydroxychloroquine as soon as it became clear that a vaccine for HCQ would be possible. In the months since, no other treatment has been allowed to steal the limelight. We're just been told to wait (and wait) for the vaccine.
I am really distressed by this. But I see no other explanation which accounts for (a) the abundant research available and (b) the official narrative about these drugs, which appallingly contradicts that same research.
> The experts successfully swooped in and saved us from all of that, figuring out which way the wind was blowing only two weeks later than competent amateurs. This was a useful service. Without the experts, things would have stayed open forever.
I don't understand how you can be as smart as you are and still think that "things staying open forever" would be a bad thing. Cities with no meaningful lockdowns and widespread noncompliance are having VIRTUALLY IDENTICAL outcomes to cities that locked down super hard. In some cases, they're having better outcomes (though my gut check is that this is just statistical noise). It is not AT ALL obvious that the 'experts' that California has delegated their critical thinking to have done anything other than perpetrate an absolutely massive crime against humanity.
Come on Scott. You're smarter than this
Scott, the best systems for combating COVID aren't ours, but the ones in the countries that got COVID right. These are, variously, China, Vietnam, New Zealand, Mauritius, Singapore (outside migrant worker dorms), North Korea (apparently), Australia, Thailand, Taiwan, etc. Japan, the Philippines, South Korea, Cuba, Norway, Finland, Denmark, Cyprus, Mongolia, etc. did better than average. The question is not to burn the system down or not, but whether to move to a better system. I propose a government by superforecasters, because superforecasters seem like people who know what they're doing.
"There are few biologists who deny evolution, few epidemiologists who think vaccines don't work, and few economists who are outright communists."
Low bar. Let's raise it higher.
"This should be astonishing, and we are insufficiently grateful."
We got half a million deaths and counting. That's twice as good as Russia per capita, but it's worse than Brazil and far worse than the Philippines. It's impressive by precisely zero metrics.
One consequence of this kind of thinking is that its interesting to look at different incentive structures, and see where they have better results. The first comparison I thought of for this was countries with public healthcare systems. For example the nhs.uk has a similar to webmd encyclopedia of common conditions, drugs, etc. Which seems much clearer to me as a non expert. eg the page on aspirin says its mostly safe, and lists common side effects with the context that they are rare. https://www.nhs.uk/medicines/aspirin-for-pain-relief/
I don't think this is because public healthcare systems are magically better, (and please don't make this thread into an argument about m4a). My best guess for the difference is:
If you are a National Healthcare Service you have access to huge numbers of government lawyers on retainer, and massive institutional power, so the risks and potential of being sued are much lower than for someone like webmd.
Also, for webmd, someone with a headache deciding they have an obscure cancer does you no direct harm, but if you are a public healthcare provider they will be taking up your time and resources. So presumably someone did the calculation that it was better to stop those false positives and deal with the occasional complaint.
And on the individual level, a large organisation can shield people from bad incentives. Presumably these pages were written by employees of the health service (probably in committees) who reported to higher up adminstrators, not dealing directly with the end consumer. So they could be given an instruction "write clear and concise advice" without the political issues trickling down to them.
So maybe the takeaway is that as well as there being a function of being too small to care you can also be too big to care? If Fauci was the american health tsar he would give better advice, though that runs into the benevolent dictator issues. The brits don't seem to have done so great with covid advice, so maybe the protection only works when things don't get too political
My big problem with the expert response was the lying -- statements early on like "don't wear masks because they don't work", which was an obvious lie to keep people from panic-buying masks, only to then mandate masks later. Followed by "we just need to lock down for 6 weeks to flatten the curve", followed by "we'll probably still be locked down at the end of 2021". Or the refusal to talk about honest tradeoffs - "this is how many people will likely die if we don't lock down. This is how many people will probably permanently become poor if we do."
The American people did, in fact, elect an idiot game show host. However, we were given the choice between an idiot game show host, and possibly the most unlikable person to ever run for office. Hell, this time around, the Democrats chucked any candidates that were even remotely interesting, and seems so hell-bent on telling the millions of people that voted for Trump that they are ontologically bad people that need to be scorned and punished. That's how you get someone far worse but more competent in 4 years.
I agree wholeheartedly that tearing down the system and assuming that somehow people that care about things other than power will magically rule us benevolently is a pipe dream. But it seems that both the far left and far right have lost all hope that things can look less than shitty for them and their kind in the forseeable future, so they're willing to chance it. This is something that the current administration, if they were competent at all, would attempt to diffuse.
This is a great post, thanks! That experts, institutions, and other avatars of Consensus are a weighted-like-a-ball-and-chain average of the views of the rest of society is a really cool and powerful idea. So is the point that they aren't inherently dishonest or unprincipled-- it's just that the Institutional Power Genie shows up and replaces them if they get too principled or honest.
But I have three reservations.
1. Who exactly is the Institutional Power Genie that hypothetically gets Zvi fired as CDC Director? Since we live in a democracy it must represent some actual people, right? But it clearly doesn't represent you and Zvi. And it doesn't represent your Straw Rednecks either, because they're, shall we say, not Biden's core constituency. The charitable interpretation is that it represents some third group, Straw Bluenecks who think the virus is real and bad but either are risk-averse in a very particular way that makes Zvi unacceptable to them, or just get a lot of utils out of knowing that Very Serious People are in charge. The uncharitable interpretation is that the Straw Bluenecks are just Fauci in a trenchcoat-- that the institutional, power-seeking mindset has become self-perpetuating by getting itself unfairly weighted into the average. I favor the uncharitable interpretation.
2. If Zvi's proposals are in practice not implementable by real-world institutions, then how are we supposed to interpret them? After all, a lot of what he (and everyone) posts really does amount to "here's what I would do as a benevolent dictator" not "here's my advice to you, a private person, as a friend". Is he trying to change people's minds to shift the weighted average? Or is he saying the average should be weighted differently-- toward him? To the extent that he succeeds in shifting the average, how concerned should we be that some illegible blogger can shift the average just by being popular and insistent? To the extent that he doesn't succeed, how concerned should we be that his posts are just a form of outrage-mongering entertainment?
3. What does it even mean to say that someone is "illegibly good" or "illegibly bad", if it's misleading (due to illegibility) to imagine their beliefs actually being implemented? Good or bad for whom? Zvi's advice is good for us, maybe, but that seems to just boil down to saying that we agree with it. How would we convince Straw Redneck's audience that it's good for them? Wouldn't we have to appeal to consensus-- i.e. Fauci? And if so, why are we still trying to shift Fauci's position toward our own? This model, when you work it out, actually seems to support Weyl's point of view: the benefit of Fauci, insofar as there is a benefit, is that he's legible to the broadest possible range of people.
There's another aspect to what Dr. Fauci does -- He isn't there to dispense the truth to the masses, he's there to make statements to the masses that will cause them to do what is optimal relative to Dr. Fauci's plan. Within that goal, he tries to avoid saying anything that is actually false.
For instance, early in the epidemic there was a real risk that the public panic-buying masks would vacuum the supply chain empty of them right at the moment that medical people needed far more of them. So the statement was made that there was no solid evidence that masks worked against Covid, which was true but any person who attempted to analyze it would realize that masks were likely to be quite effective. Readers of this blog will probably notice that is an attempt to avoid the tragedy of the commons -- for each individual, buying a mask as quickly as possible is the optimal choice, but also preventing others from buying masks is also optimal.
aMore subtly, the optimal plan depends on where one is in the economy. As a computer programmer, the cost of an intense lockdown is personal nuisance. For a restaurant worker, it might be homelessness. Someone has to strike the balance between those competing interests, and that's a lot easier to do if you disguise that it is being done.
Everything here reminds me of this Fantastic Anachronism post last month -- despite the horrible track record of replication in social science we should still default to trusting credentialed experts because it's better than any alternative https://fantasticanachronism.com/2021/01/12/unjustified-true-disbelief/
I'm really glad you're writing on moral mazes. I've been troubled by Zvi's simulacra work for a while now: he seems to be saying important things about what happens to adversarial systems, but the end result is so bleak. I had to go back and re-read the goddess of everything else to remind myself about the power and beauty of cooperation.
I would be very grateful for a theory that harmonizes Zvi's work with Elua's power. This article is a good first draft, but I long for hope that the Elons and Zvis of the world aren't anomalies, that the quokkas can learn to protect themselves and the slytherins can become great philanthropists.
I generally agree with this, but: I don't think it's always been this bad. Back during WW2, for example, we seemed to have competent leadership. We put Vannevar Bush in charge of military research, and that worked out great. I think something has gotten worse over time. And it's worth trying to figure out what that is and why it happened.
> When the Director of the CDC asserts an opinion, she has to optimize for two things - being right, and keeping power.
I think it's really more complicated than that. I haven't been on the 'other side' of public decision very often in my life, but I have a few times through my job at a tech company, and I know these kinds of things almost look unrecognizable to an insider, compared to an outsider.
My tech company example is, say, a widely-publicized product launch to which the general public assumed a lot of fair things, the gist of which is that we were insane and had no idea what we were doing. In fact, the true explanation was completely unrelated to that -- that the target of the launch was a type of enterprise customer that was alien to the public.
I can imagine many forms of this that might happen for a Fauci public statement that take different directions than "maintaining power". For one thing, in the early days of the pandemic, it was truly not clear in the data how serious it was. A random person can say: "oh, I have a feeling this is much worse" but the CDC / whichever officials can't. They have to balance their gut against the data they have, their internal policies, and the fact that if they say something extreme it carries a lot of weight, which means shouldn't be wrong both for logistical reasons and to maintain credibility.
There's just nothing on the line for an individual making predictions. When you end up in a position of great influence, there's a lot on the line, and you have to be much more careful. Rationally, of course, it would make the most sense to overcorrect for harmful outcomes, and spend more resources than needed. But then what happens if you're wrong, and a lunatic congress comes at you for budget cuts decrying your uselessness? (And this doesn't even broad the question of what the right behavior is when you're trying to avoid completely losing Trump's ear. The "Case Against Fauci" article thinks it's not worth it, but I'm pretty skeptical.)
I'm just saying -- it's just not just being right and being in power. "Being right" looks very different for a public official in a political situation than it does for a blogger.
>and few economists who are outright communists.
Those are the only economists who are correct, though. Remember Scott that in your own reading of Marx you could not seem to comprehend his basic sentences and instead deferred to Peter Singer's own (incorrect) analysis. This is politics as the mind-killer, as the Rationalists say.
I think you are heavily biased by living in USA and some of your observations might not generalize elsewhere.
Consider this:
https://www.wired.com/story/how-taiwans-unlikely-digital-minister-hacked-the-pandemic/
&
https://blog.pol.is/uber-responds-to-vtaiwans-coherent-blended-volition-3e9b75102b9b
&
https://en.m.wikipedia.org/wiki/Central_Epidemic_Command_Center
Hi Scott, long time reader. Really like your writing.
But I've noticed a somewhat distressing tendency in your recent posts (by which I don't really mean just the ones on substack) and I think it's getting worse and so I will speak up. I notice an increasing frequency of things like this:
"But someone else will decide to always trust *their* friend, a guy in a MAGA cap who says coronavirus is fake and Dr. Fauci is a Satanist. "
"At one point we tried a very simple best-person-picking procedure that really should have worked and ended up choosing Donald Trump as the best person. "
"Our system of expert-having is actually much better than we deserve, given that we elected Donald Trump president."
And I have to wonder if this is necessary or even good, or if it is just dabbing on trump supporters because it feels good. Is the point you are trying to make actually best served by including a MAGA cap, or is this just exclusionary to some of your potential audience? Politics is the mind-killer.
'If a random shmuck who doesn't know anything about anything Googles "who should I trust about COVID?", Google will return Dr. Fauci's name.'
I decided to try this experiment using incognito mode. With the quotes the first result is this post, without them it is the WHO.
"The whole scientific-technocratic complex is a machine which takes Moloch as input and manages - after spending billions of dollars and the careers of thousands of hard-working public servants - to produce Anthony Fauci as output. This should be astonishing, and we are insufficiently grateful."
That kinda sounds like a proof-of-work blockchain. It consumes vast amounts of resources, but that's not (just) because it's inefficient, it simply takes a *lot* of effort to defend such a system from attacks with our current technology.
Do you think that locking your household down two weeks early as a good decision? It doesn't seem like one to me, though as the pandemic lengthens it kind of seems like well, two weeks here and there, who cares? But just in the spirit of self-reflection, didn't you lengthen your personal pain at a time when the chance of infection was essentially negligible?
If you know anyone who has been tempted to despair about supposed 'failures' of the western, liberal world order and its institutions over the last year then I'd encourage you to send them this. Whether something is considered acceptably good depends on what you compare it to - what we have now is clearly far better than what we would get by burning the system down. What's more interesting is whether what we have now is better than what we had in the past - and there's evidence (from things like the way the polio vaccine rollout was much faster in the 40s than our rollout today) that there are a few aspects of the current expert-selecting institutions that seem to have regressed, or not kept up with the increasing difficulty of the problems handed to them.
I think it's worth noting that the historical comparisons aren't ever to us actually succeeding at dealing with pandemics in the past, but to things like "WWII-style" efforts - i.e. thinking that if we could just do x as well as we once did y then things would have been a lot better. This implies that if you made an institution analogous to e.g. the weapons researchers of WW2 and the governments that funded them, or NASA in the 1960s, without copy-pasting 1940s/1960s society wholesale, the outcome would have been better.
To me that suggests its institution design that's the culprit, and that there are ways to filter our legible experts to be better that we've missed out on.
But even in making that kind of comparison, we're already back in the sane, concrete realm of institution design and trying to patch up the system we have with tweaks we know might give us better experts more reliably, rather than the ethereal realm of comparing the whole of our society to some imaginary alternative.
I'm lucky enough that I have had no reason ever to use medical services in the US, so I can't comment on that. But I have had quite a bit to do with the US political system professionally , and have had a chance to see how it works (it doesn't). It's an almost uniquely dysfunctional system, massive, labyrinthine, highly politicised and personalised, where it's hard to know who is in charge of what, and decisions take forever. Irrespective of who your political leaders are, I think your country is going to have a terrible time coping. Most of the states that have coped well have strong, professional and well-managed public services which draw a strict line between career and political appointments. (Singapore is a good example). Countries that used to do that but don't any more (like Britain and to an extent France), have suffered worse as a result.
But it isn't just the political system(and of course sheer luck plays a big part as well). You have to take into account the social capital of the country and the willingness to work together (again, most Asian countries have done well for this reason, but so has New Zealand). You also have to take into account political culture, and what seems important not only to politicians but to the whole PMC and media classes. The rest of the world watched in stunned disbelief as these classes in the US descended into severe Trump Derangement Syndrome, and stayed there, even while the epidemic was ramping up. But we had problems in Europe as well. Closing frontiers was essential but not done nearly quickly enough for reasons of political culture. Partly this was the EU's neoliberal ideology of open markets and borders. Partly it was the identification of frontier controls with nationalism (hiss) and even Trump (double hiss), which meant that if you supported them you were a Xenophobic Nazi Fascist. "Viruses don't have passports" smirked one European politician, who I think is out of a job now.
On Scott's wider point, I found his remarks uncomfortably apposite because I've suffered from a couple of these dilemmas about getting information on medicines. For the last fifteen years I've been taking blood thinners (first Fluindione, which is like Warfarin, and then Rivaroxaban). I was warned about not taking aspirin as well, and then I was told not to eat certain foods (bananas for example) rich in Vitamin K, and then as I began to read around, I discovered that most foods have an effect of some kind. Since I'm intolerant of gluten, I realised quite quickly that I was as likely to die of hunger than of blood clotting, and in the end I took a personal decision to eat a normal (but gluten-free) diet, on the basis that one thing would cancel out another. I'm fine and don't show any symptoms of uncontrolled bleeding for example. But nobody tells you anything useful. Nobody says how precisely this or that food affects the fluidity of the blood, for how long or in what combination. I've been told not to take Ibuprofen. Fine, but does that mean not even once, at 200mg? Or does it mean not continuously at a higher dose? If I take one for an inflammatory problem, will I die? Nobody seems to have any idea, or if they do they are not telling.
And I've been a chronic insomniac, probably for longer than most of the readers of this blog have been alive. There are two kinds of doctors in my experience. Good sleepers, who tell you to make sure you have a dark room and get plenty of exercise and you'll be fine, and won't prescribe anything, and insomniacs, who know that insomnia is terrible for general health, and that sometimes, lying awake at 4am and wondering how you are ever going to make it through the day, you start to think that suicide may be the easier option. The latter will prescribe something. So I've been taking Bromazepam from time to time, and, as Scott's wonderful new site points out, it's been argued that even small doses will kill you. I'm still alive, and I think I'll choose my own method of dying than you. Indeed, I think the only thing you can do is to try and see what happens. In Europe the problem with medical advice isn't quite so serious, perhaps, as in the US, but it's pretty serious. I've pretty much given up hope of getting any useful advice from medical sites, and I'm much more inclined to try something that has demonstrably worked for someone else. So I eagerly took up Scott's suggestion of Theanine, and took 500mg before bed (that was the only dose I could find). I was awake half the night and finally gave up. Any other suggestions gratefully received, but I'm going to try them for myself, not just be guided by websites that produce lots of words but no actual help.
>When the Director of the CDC asserts an opinion, she has to optimize for two things - being right, and keeping power.
As an excellent example of this, look no further than Professor David Nutt. Nutt was a Fauci-esque figure in the UK: he was, at one point, an advisor to the Ministry of Defence, Department of Health, and the Home Office, as well as President of the European College of Neuropsychopharmacology.
He also had no interest in optimising for keeping allies and power, and used his government positions to go around telling theh public exactly what the science said about addiction, illegal drugs, and so on.
He was rapidly fired for this, with several senior government figures accusing him of "talking about politics rather than science" -- which, ironically, was precicely the opposite of what he was doing.
Perhaps you should use a kind of Captcha in order to access your site? Instead of identifying 'all instances of porpoises waiting at stop signs', it could be a randomised test of the ability to reason statistically?
'If you can't pass this test, here's a helpful site where you can find out how not to misunderstand what I'm giving you..'
A very strong post, without a doubt the best you've written since your comeback.
There's a profound truth to the idea that only people outside a corruption-based system such as politics can regularly find and speak truth. Constant excellence is possible only on the individual level; mix in groups of people (with inevitable bias and selfishness) with varying incentive, and what you get is a compromised mess.
The Swedish poet Karin Boye incidentally wrote a poem on this subject, whose first stanza is both fitting and poignant. Av tvång (https://www.karinboye.se/verk/dikter/dikter/av-tvang.shtml) from Härdarna (1927):
"Jag är en fattigdomens präst
och ska väl så förbli.
Den inget har kan våga mest,
till dåd och tanke fri."
Roughly
"I am a priest of Poverty
and I shall so remain.
Who nothing owns can wager most
free in deed and thought."
To remain a priest of Poverty (whether in an economic, social or political sense) who can always wager most, or to give up the vows for real gain? That is the question, and God knows it's not an easy one.
The biggest issue to me is that our system (particularly in medicine) has very few mechanisms for punishing over caution and lots of mechanisms for punishing good faith bold action. As a result, everyone with anything to lose is in maximum CYA mode at all times.
In a pandemic (or a war, or any other complex fast moving crisis) inaction is often just as deadly as incorrect action. But our medical establishment totally dropped the ball on switching to a “war footing”, instead acting as if minor relaxations of the normally truly Byzantine approval processes were bold acts of heroism. How many lives would have been saved if the Moderna and Pfizer vaccines had been approved at the same time as Sputnik V? No one will be punished for these tens of thousands of bodies, but if the vaccine had merely saved a hundred thousand while causing a couple hundred birth defects or whatever, heads would have rolled. This is absolutely backwards.
Of course it would be one thing if Fauci were merely over cautious - but he didn’t only make “WebMD errors” of excessive ass covering. He, by his own admission, told 2 big lies (or at least things he was pretty sure were untrue): “you should not be walking around wearing a mask” and “modulating my estimate of herd immunity based on what I think people will believe”. And beyond Fauci, “Racism is a bigger threat than COVID, so protests in a pandemic are justifiable”. That was not a WebMD error.
We have medical experts to deliver medical expertise. Perhaps we accept a bit of politicking and over caution as a cost of doing business. But when experts start openly lying because, hey, they’re experts, you wouldn’t go against science, would you? That shatters public trust in expertise. Fauci maintaining power might be good for him, but it’s bad for everyone else because half the country isn’t going to listen to him, and that’s in good part his own fault.
Do you think having more prestige for working in government bureaucracy would increase the correlation between gaining power and capability? East Asian countries seem to do a decent job of having competent officials and these things are much more prestigious there.
Really amazing writeup and there are a lot of valuable points in here. I'd like to give a bit more attention to this point since no one else has:
"Or maybe I'm dumb and biased, and actually the experts are doing much better than Zvi"
I don't think that's true in general, but after reading through Zvi's blog there is an error that could have cost thousands of lives if Zvi were in charge. He thinks that at least one of the vaccines should have been open source.
If Zvi is a tech worker, then it's easy to see where he's coming from. In tech you want a thousand flowers to bloom, and the best survive. Doesn't matter if one product sucks, only the good ones will still be on the market in 3 years. Well medicine isn't like that - move fast and break things does not work.
Medicine is built upon trust - and trust is hard won and easily lost. If 5 startup vaccine manufacturers compete to produce the vaccine doses as fast as possible and just 1 of them screws up that's a huge problem. Perhaps 5% of doses are ineffective due to insufficient QA or perhaps some of the doses result in higher rates of side effects. Well then the entire well is poisoned - news stories come out, politicians on the left and right seeking easy political points lambast vaccine producers, anti-vaxxers see a golden opportunity to move public discourse. Perhaps it doesn't kill vaccine production entirely, but public trust falls, investors and regulators become wary and vaccination decreases.
And all that for what purpose exactly? The companies best able to reliably ramp up production of millions of doses are the companies that have already done that in the past. And they have done that with rigorous QA that they've been forced to develop after hundreds of failures and lawsuits. Open sourcing would not significantly increase production and it would introduce unnecessary risk.
Now perhaps I'm misunderstanding him, perhaps he doesn't really mean open source - just that any established pharmaceutical company should also be able to secure a contract to make the vaccine - no exclusivity deals. I doubt that would actually change much. All of the big players interested in vaccine production seem to already have jumped in, and they've all chosen their champion. And the reason is simple - economies of scale. It's cheaper to produce 10 million of one kind of vaccine than 5 million of one and 5 million of the another.
Thanks so much for the read. Reminds me a lot of The Use and Abuse of Witchdoctors For Life (https://samzdat.com/2017/06/19/the-use-and-abuse-of-witchdoctors-for-life/). We never really outgrow witchdoctors, we just improve their performance somewhat.
The problem with balancing optimization for truth and optimization for power is that such compromises might not be stable. Individual people's ability to find the truth is judged by the very institutions in which they're trying to advance. So the more institutions optimize for maintaining power, the lesser their ability to find the truth, including recognizing people with the theoretically optimal balance of truth-seeking and power-seeking. As the capacity to recognize truth degenerates into noise, power-seeking becomes ever more important until it's all that's left.
If power can feed on the corpse of truth-seeking, but the reverse is not true, sooner or later truth will be devoured.
> The second reason I'm writing this is because people keep asking me "should we listen to experts"?
I think you're missing a step here. Experts, especially in a policy realm, are applying a different standard of proof than you are.
Locking down your house, for example, made sense as soon as you thought it was more likely than not that it would be helpful. Maybe an even lower standard (a precautionary principle) could have applied, since the consequences of a covid infection could be so severe.
On the other hand, issuing this kind of directive for an entire region imposes costs on other people. If you're wrong about your own house, then you and your housemates voluntarily pay the price; if you're wrong about policy then everyone pays.
Therefore, mandatory orders do (and should!) use a higher standard of proof, somewhere between "clear and compelling evidence" and "blindingly obvious."
Where we fail, however, is that we don't allow experts to offer sub-certain judgments. Fauci can't go up and say "we think school closures are a pretty good policy, but we're not yet certain about it." In my opinion, this happens for two major reasons:
* First and foremost, society-writ-large is not comfortable with ambiguity. Some of this is cultural (see the general confusion over election predictions, where 55% is treated as a certainty); some of it is also personal because people have limited bandwidth to perform their own risk/reward analysis.
* Second, "wrong" predictions, even when advertised as uncertain, tarnish expert credibility. Some of this is an outgrowth of the first point on ambiguity; some is a response to an adversarial environment where other motivated actors (political opponents, charlatans, conspiracy theorists) amplify failures. This is where I think your analysis on the politics comes into play, as well, and the end result is to retreat into a defensive certainty.
In turn, this has two negative policy implications:
* The first is a retreat into process. "Experts" can't always speak in hindsight, so defensive certainty prompts agencies to develop rigorous policy for drawing conclusions. Even if the conclusion is wrong, as long as the right steps were followed then the experts can claim success. See for example medical approvals. We know now that the major yet-to-be-approved covid vaccines are more likely than not to be helpful, but experts won't bless that decision until the right steps have been followed to certify the proof.
* The second implication is a strong status quo bias. The status quo is no lockdown, so there must be compelling evidence to prompt one. The status quo is no masks, so there must be compelling evidence to recommend masks. The status quo is no covid vaccines, etc. And after said compelling evidence, the policy prescription becomes the new status quo, equally unshakeable. This leads to the oft-maligned "eggs are good, bad, good, bad" flip-flop, and it also gives undue power to whoever can define the status quo -- often a political choice that we as a society don't really consider.
Little here has to do with the political realities of placating stakeholders. Instead, so much of it derives from an inappropriate (or even impossible) demand for certainty, including our instinct to conflate certainty with correctness.
Maybe a comprehensive civics education should include a long module on Bayesian-style reasoning (maintaining uncertain priors and updating with new information) and the Kelley criterion for wagering (relating the size of one's stake to both expected benefit and expected variance).
Before finally giving up on trying to convince my non-SSC friends of the wisdom of Zvi, I gave this analogy (mostly about FDA approval):
There's a water delivery man who normally delivers at 1 mph. There's a drought and he manages to get his speed up to 2 mph, saving twice as many people. But I think while very unpleasant for him, he could have gone 10 mph. Should I be happy he managed to double his speed or sad at the shortfall of what I thought possible?
I think some amount of it is about what should our emotional reactions be to inadequate equilibria in the world. Is frustration and anger a good source for driving change? Or is equanimity and understanding of bad incentives?
By what standard to we judge Fauci: Zvi or Lysenko?
Alex Tabarrok has mentioned this a little bit, being thankful for Warp Speed, but still pressing for more.
Should we focus on feasible alternatives of conceivable ones?
Maybe the answer is both, but it's easier to digest in separate posts, or separate speakers. Zvi can depress us about all the things that could have been done, and remind us that there is so much more possible. Scott can cheer us up by reminding us that we did okay giving the existing constraints in place.
We actually can infere what Zvi would do if he became director of the CDC. He wrote it several times. 9/3 "Burn it to the ground. CDC Delenda Est." (https://www.lesswrong.com/posts/RF5HjYYLY5FWgtA9v/covid-9-3-meet-the-new-cdc), 9/24: "CDC Delenda Est" (https://www.lesswrong.com/posts/E8fp7BxpuzcwaMGtz/covid-9-24-until-morale-improves), 12/17: "CDC Delenda Est" (https://www.lesswrong.com/posts/Rvzdi8RS9Bda5aLt2/covid-12-17-the-first-dose). So I suppose Zvi is your friend "pitchfork in hand, trying to burn down the system". And maybe the Fallacy does not seem to be Marx's Fallacy, it's possibly some Marxists' fallacy and possibly some other ...ists' Fallacy, and maybe "..." can be "Trump" but it can also be "Rational"; after all, your own folks never optimize for power. Making some very smart people a world dictator may make the world a Well-Kept Garden from his/her perspective. But given that this whole rationality community talks a lot about the problem that it's really hard to define a utility function that does not turn out to optimize for something you did not want in the first place, I think it could be slightly more modest with regards to politics.
"Just get the best person in the country and let _her_ fix everything"
Which is a correct statement because women actually are better leaders, on average. See, this is why I like to read you, nuances like this one and the hundreds I probably don't catch.
In that trolley problem, it’s worth noting that the out of control trolley is going to round the bend and kill all six people either way. That’s a common problem in illustrations of the trolley problem.
Additional problem with installing your friend as world dictator is that a lot of the reason they succeed is there working with good information and if there world dictate a lot of people will be very interested in warped there information diet
Prediction markets are not better for policy issues like covid, climate change, poverty, etc. Markets are great for aggregating information on parochial issues like the price of gold, the present value of a corporation, and all the various baskets of these things. To have a successful prediction market you need a metric that measures what good-faith people really care about. Political issues like climate change or poverty necessarily use intermediate goals that are deficient.
Many are subject to Goodhart's law, which states that any statistical regularity will collapse once pressure is placed upon it for control purposes. For example, Mao maximized steel output in the 1950s to modernize his economy because the best economies had high steel production. It was a disaster because the prioritization of steel production over everything else ignored marginal costs which would be reflected in the profit-and-loss system used by decentralized economies.
Covid deaths are bad, but so too is shutting down the economy, preventing people from seeing friends and family. These costs are difficult to measure, so one can always dismiss them, but they include not merely psychic costs, but real deaths. We can give everyone an education, but the value of this declines the more it is put into a national policy, and many now have useless degrees and much debt. CO2 production is bad, but invariably it is measured piecemeal, ignoring fixed costs, and counting negative externalities for one process while ignoring the negative externalities for another.
If Biden proposed a new Middle East peace policy, a market in Middle East peace would be of no value. The same is true for any of our great issues of debate: racism, sexism, inequality, the environment.
This doesnt quite correlate w my experience following epidemiologists and virologists.
Late January I had a flu and Wuhan was a story already, so that was a way to enjoy a respiratory disease, by considering the chances of a pandemic. I was following 'This week in Virology' podcasts reading up on early cases of community transmission on flutrackers.com (did they catch all the contacts of this or that tourist from Wuhan? Oh no, this early case in Thailand infected a taxi driver, and this case in Japan was a guide a tour bus etc), There were R0 studies giving various numbers mostly between 2.5 and 3, and existing (flu) plans indicated NPIs are unlikely to control the spread of something >2. If anything, this turned out to be pesimistic. By Feb.4. I had bought a few months of food supplies (recently checked the date on that receipt), to be able to stay home during peak transmission. I was hoping it wouldn't become a big deal here in Europe before Spetember, if they managed to catch early transmission and warm weather being just aroung the corner, until Italy happened. Though w Iran (Feb 19th) it was clearly a pandemic, as epitwitter at the time explained. Also expected mortality was in the range of 0,5% to 2%, already.
Overall, I think listening to the chatter of experts, and immersing oneself in the discoruse was a good way to stay well informed about this pandemic. Quite possibly the official statements of government bodies were far more conservative, exobiting high normalcy bias.
The problem is worse than you describe, because in many cases what we are looking for is polymathic expertise. For the efficacy and side effects of psychiatric drugs, we need psychiatry and pharmacology and general medicine - and since psychiatrists are MDs, a good psychiatrist gets us most of what we need. But for pandemic response, we need epidemiology plus sociology, economics, education, politics and not just in the how-do-I-stay-in-power sense, and more other fields than I'm going to try and enumerate here. Some of that we need to know if a proposed intervention is even feasible - sociology tells us how long we can keep people locked down before they start going to private house parties or mass political rallies instead of bars. Some of it we need to know whether the proposed cure is worse than the disease - how much harm does a year of zoom-schooling do to the long-term life prospects of elementary, middle, and high school students?
But it's hard enough finding legible but mediocre expertise in one field, that we generally call the job done when we find a Dr. Fauci. Fauci, working within the limits of the political system we have, is a mediocre expert in one thing. The one most important thing of the past year, perhaps, but not more important than all the other things put together. Not good enough.
Polymathic expertise does exist, and it's not *that* hard to find if you're looking for it. Zvi, for example. Or Scott. But broadly legible polymathic expertise, is very hard to come by.
Scott give prediction markets lots of credit lately. Is there a good source that argue why we should give it this credit and not just treat it like another random idea that sounds nice but probably doesn't really work as advertised, like blockchain?
(If you think that blockchain is not just another trend, same question)
For side effects, maybe something like this would work:
10% report stomach ache
3% report itching
0.2% report their hair falling out
This gives the facts and avoids having to make value judgements like "dangerous".
Isn't the entire idea of academia that they don't have to fight for power in this way and can optimize for truth?
As such, the real failure is not in governance, but in science!
Re: not worrying more about addiction to Adderall than alcohol: Most people don't drink every day [citation needed]. They drink occasionally at social events. Adderall is recommended to patients to be taken regularly for some period, isn't it? If I drunk every day, I'd definitely be worried about the possibility of getting addicted.
I'm a layman who knows basically nothing about either Adderall or alcohol addiction. My point is that your comparison is not easy to interpret for someone who has never drunk regularly, and has never had a reason to investigate how much one can drink regularly without getting addicted.
What about Medscape. Appears much more professional and informative and less "political". I generally use it for medical questions. Haven't tried it for mental health issues.
“a kind of bumbling careerist with a decent understanding of epidemiology” is harsh for someone who was lead editor of Harrison’s. I mean, sheesh. Dude’s not a saint or Einstein but he seems pretty smart. I mean, not as smart as the economists who all seem to know marginally more than anybody about anything...
I agree about the general problem of power inviting corruption, but I think you're being far too kind to the current system.
Burn-it-downism actually works a lot better than I'd naively expect. The Bolshevik Revolution led to Stalinism, yes, but Stalinism involved possibly the most successful economic development program ever - mortared, of course, with a great deal of blood, but not the implosion one might have predicted. Even attempts at outright anarchy haven't ended with everyone starving to death. The bigger issue is violence, which revolution always causes and which our current systems are reasonably good at preventing. Of course, an anarchist would argue that the entire thing is built on the *threat* of force, but being held at gunpoint is still better than being shot.
But - with simulacra levels and all that - we currently have leaders that are detached from reality entirely, which is the reason for the uptick in populist sentiment, with the aura of revolution and violence that's always implied. When Congress has an approval rating below 25% for ten years straight, it really shouldn't be surprised to get a lynch mob at its gates. You talk about Fauci not being Lysenko, but Lysenko was very much the exception in the USSR; Fauci isn't Korolev, either, and I'd say Lysenko has closer analogues in the modern American public service than Korolev (something something bioethics). The Attila the Hun comparison is misleading because Attila was leader, not expert, and honestly I'm not sure Trump is all that far from an Attila in temperament. Not that we actually know much about Attila, historically. I could say something about Tamerlane and US foreign policy, but I don't really have any well-formed thoughts on that. The point is that the positive elements you're talking about are institutional successes fifty years old; I mean, even Fauci himself isn't exactly a new appointment. Possibly this is because the Cold War USA faced an actual threat in communism, and before that the Axis, and therefore had to acknowledge physical reality to fight them.
And you could say that, sure, our society's current paralysis is bad, but it's still better than risking things on any big changes because those can backfire, and basically that what we have is good enough. This is called conservatism. The problem with that, even if you're of the opinion that <gestures at planet> this is fine, it's not *stable*. Technological change leads to social change, sustainability is more than just a buzzword, relying on imports from increasingly poor countries has a natural endpoint, and if it takes ten years to repair one highway intersection our infrastructure will eventually fail entirely. We need utopianism in our discourse - we need multiple competing utopias - because otherwise we're reduced to using the literal worst expert ever as the bar our experts need to jump. And we probably need the threat of a mob with pitchforks (preferable distant, but with present dysfunction a more immediate one seems to be required) to make our elected and unelected leaders actually optimize for something besides power - in layman's terms, do their jobs.
My grandmother was struck by lightning... :(
This is a fantastic line of inquiry. But, doesn't it bump up against anti-culture war imperatives? It has all the trappings of previous ones: expertise vs. non-expertise, authoritarianism vs. individualism, and seems to also land on standard left vs. right spectrums
Stunning apologetic for elitist behavior that leads to actual harm.
This isn't kind, but it's true and I think adds something specific: the two times I've met Zvi in person (at rationalist-adjacent meet-ups in NYC, three years apart) I found him arrogant and unpleasant, to such a degree that it really stuck with me. Like, loudly declaiming his opinions and derisive toward anyone who disagreed with him. I'm generally pretty unbothered by non-neurotypical people, but that didn't feel like what was going on. I think it was just a difficult, non-collaborative personality.
I haven't read his blog much so maybe he's much easier to hear in writing. But I also wonder if these traits are somewhat correlated with being right about something early and loudly and also anticorrelated with the skills it would take to affect the world outside of, say, the rationalist-adjacent blogosphere. I also worry it leads to worse, un-dampened outcomes when he's wrong. Somewhat related to the broader "competence" point that others mention.
(This is really an unkind thing to say, so Scott please feel free to delete if you think it doesn't add enough to be worth it.)
I remember reading a really good article you wrote about a book you read on Marx. It was really insightful to me, but I got the feeling that communism could be fixed by more carefully thinking about the problems on a system like that, maybe using simulations, etc. But it seems you heavily dislike communism and I would like to understand if you see a problem with the core idea of the workers owning the means of production and also the redistribution of wealth.
>When Zvi asserts an opinion, he has only one thing he's optimizing for - being right - and he does it well.
>When the Director of the CDC asserts an opinion, she has to optimize for two things - being right, and keeping power.
The senior officials in the CDC are (both reasonably, and according to their own claims,) optimizing for *saving lives* and keeping power. Sometimes, the thing they do to try to save lives is shut up about true things, or even say false things. That seems to be what happened with masks; they were trying to manage an economic supply/demand system to make sure hospitals and healthcare workers, who were at higher risk, had masks, and that the general public didn't go out and buy them. Was this a bad idea? In retrospect, obviously it was - but more critically, truth is a harsh mistress, and it seems that you can't serve two different goals, even when the first goal is honesty, and the other goal is public health.
"There are no first world countries"
Can someone explain the joke? Because I don't get it.
This post really exemplifies the saying "damning with feint praise." Sure, the experts are consistently wrong but that's only because they are optimizing for something other than being right! Come on, man. I don't know why you defend these people.
>The Director of the CDC reads those same papers. But some important Senator says that if airborne transmission is announced, important industries in his state will go bankrupt. [...] So the Director puts out a press release saying the evidence is not quite strong enough to say airborne transmission definitely happens, and they'll review it further.
Not saying this isn't true, but I wonder why these pressures aren't balanced out by countervailing pressures. The correct press release could harm the important Senator's state, but should greatly benefit the average citizen, so the important Senator's objection should be counter-balanced by the support of many other Senators (who should profess their objection to fire the CDC Director). This is why we believe representative democracy has a chance to benefit citizens at all. If other Senators will not resist the Senator's pressure, it means they need his support for something else, more important. In the end, this should theoretically (ideally) work out for the best compromises to benefit everyone. And when something is very important, like covid, we should expect to see fewer decisions related to it being compromised for something else.
What is the reason it doesn't work out that way? Irrationality? Politicians representing not the interests of their constituencies but their donors/lobbies? Rigid structures, e.g. the Senator in charge of CDC appointments was decided before it was so important, and now can't be changed?Something else?
OK - so the system as a whole produces mediocre expert advice - the question is can we (whatever group smaller than the general popupation that 'we' represents) develop a system that would give us better expertise?
For 'we' as patients: https://zby.medium.com/rational-patient-community-6d3617dffcfe
Once upon a time, it produced George Marshall as an output. How did that happen? Why hasn't it (apparently) happened since?
How can I access Scott's psychiatry database?
Feel like the driving factor for a lot of this is chance. There are a vanishing number of competent people and a vast sea of incompetents. If you have a mechanism for competence selection that is in any way imperfect, you're getting an idiot.
"Maybe expertise is a sham, and a smart guy thinking for five minutes can outdo a decade of working on a PhD."
I think about this a lot with respect to my field: urban mobility. I have a PhD in operations research / transportation, I work at an urban mobility startup, and I spend a nonzero fraction of my free time learning about urban mobility. I know a lot about this area. And yet, I know approximately nothing that you couldn't realize by looking out the window and thinking sufficiently deeply about what you see.
But most people - even in the industry - haven't realized many of the things which are realizable just by thinking hard. So can a smart person thinking for 5 minutes outdo a decade of learning? I suppose for a sufficiently smart person, yes? But sufficiently smart is a high bar. In mobility, time investment seems to serve as a substitute for something like "raw smarts" or "wisdom," a way to crowd source deep thoughts. In practice, everyone needs at least some time investment.
(Example of something that most people don't seem to have internalized, but should be obvious to us all if only we were all wise enough: for most intra-urban origin-destination pairs served by multiple transit modes, most of the modes should be approximately equally costly (including non-monetary costs) most of the time! E.g. if it were easier and faster and cheaper to go A -> B by bus than by car, people would shift car -> bus, the bus would slow down and become more unpleasant, traffic would decrease and thus car travel times would decrease, etc.)
I think not all fields are like this. E.g. no amount of staring out the window and thinking hard could reveal to me that human cells use DNA to store genetic information.
Just on a really narrow point it seems like both your recs and the recs from WebMD could be dramatically improved by adding a dime numerical score of how bad the side effects are. This allows one to both give a detailed summary of all the side effects and not require denying their importance while still giving a general sense of danger level relative to other drugs (forcing those who say this drug ruined a family members life to essentially debate the ppl claiming other drugs ruined lives they know).
This article would have been ever so slightly better, more likely to convince certain people, without two unnecessary jabs at Trump.
>Without the experts, things would have stayed open forever.
My gut reaction: "Oh. Shoot the experts, then."
I would like reading this better if the punctuation were inside the closing quotation marks, where it generally belongs.
> rather than being minimaxed for COVID-prediction
So this essay is great, but you've hit one of my pet peeves; you're using the word "minmaxed" wrong.
Minmaxing means something pretty specific; it doesn't just mean "optimized". For example, when I evaluate positions in a chess game, I assume my opponent makes the worst-for-me ("min") moves they can make and that future me makes the best-for-me ("max") moves future me can make. In short, it's only natural to use "minmax" to describe a situations where I'm trying to do something despite the actions of something else which I'm assuming is trying to thwart me. (Granted, technically we could call it minmaxing where there's a mathematically trivial opponent who doesn't do anything, but it's still a misleading word choice, like calling something a pentagon when you totally know it's a square and then defending your word choice by saying that one of the five sides has length zero.)
If you're trying to do something but there is no opponent (like covid prediction), the word "optimized" is a better fit than "minmaxed".
Citation: https://en.wikipedia.org/wiki/Minimax
Here’s how I use medical article databases. I’ve had x problem which the small-town PCP was ignoring. So hours on WebMD and similar gave me enough background that the next time it acted up I went to the ER and said “I think this might be a hernia” (they did an ultrasound, I was right.) Did that for a few other things. We can’t all access actual experts, so I need just enough info to get through to the doctors in front of me who can treat - prescribe, order tests, etc. Those databases are great for that.
I felt like crap for a month and couldn’t get any answers from docs and hit the databases. Compared bloodwork and got “why do they say nothing is wrong when the Internet says these numbers are metabolic acidosis?” Then more googling and it was possibly a side effect of a medication, so I did a taper myself and felt a little less crappy. I will never know if I was “right” but I burned less days feeling like crap. The databases are like a 2nd amendment for medical knowledge. We can arm ourselves. That being said, more info is better.
Scaling problem: II. Dr. Anthony Fauci is the WebMD of people.
https://twitter.com/nntaleb/status/1246122737602646017?s=20
<blockquote class="twitter-tweet"><p lang="en" dir="ltr">LOCALISM<br>The higher you go the more incompetence/reckelessness (with small exceptions). If this virus ever teaches anything, it is localism.<br><br>You<br>Your family<br>Your town<br>Your region/county/state<br>Federal Gov: CDC<br>The UN/<a href="https://twitter.com/WHO?ref_src=twsrc%5Etfw">@WHO</a></p>— Nassim Nicholas Taleb (@nntaleb) <a href="https://twitter.com/nntaleb/status/1246122737602646017?ref_src=twsrc%5Etfw">April 3, 2020</a></blockquote> <script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>
The degree to which post-modern opinions and their resultant policy advice are trapped between:
- Who watches the watchmen?
and
- We need a philosopher king in charge of every facet of life
simultaneously fascinates and terrifies me.
Excellent article highlighting the sloppy content on WebMD. It's a trash shite selling snake funded snake oil. Nobody with an iota of critical thinking takes this site seriously. WebMD is literally a lousy joke of the online world.
I don't think it's really a corruption thing that led Fauci to F up. Birx clearly wanted to stay in her job and thought that her being in place was better than Trump installing a 100% yes man. For Fauci, and the rest of the CDC, FDA, and NIH I think there's this massive status quo bias that is hard to overcome. Working in pharma it is clear that regulators and doctors want to be really really confident about something before they do it, even if the risk of them being wrong is no way near as severe as not doing anything. I think that's what happened with lockdowns, masks, at home antibody tests, etc. Until you could get a sufficient amount of data to be bulletproof (for non-delusional people/denialists) they won't do anything.
C'mon, you praise Anthony Fauci, but complains of DFW for loose ends in Infinite Jests? No wonder people send you annoying emails...
just kidding. I know how you feel, I gave up The pale King
This reminds me of Samo Burja's case study of the scientist who back in the day pulled for a Justice League of scientists to control nuclear technology: https://www.bismarckanalysis.com/Nuclear_Weapons_Development_Case_Study.pdf
I don't think anyone thinking COVID was the killer virus that the WHO made it out to be at the beginning can reasonably judge experts.
Well, maybe not. I suppose it's the same conundrum for Scott Alexander as for Fauci. If he ever openly says that COVID was a false alarm (which it was) and the response a massive civilizational fuck up the likes of which you won't find in the history of the planet until now (which it was and is), he would lose a lot of his supposedly "rational" audience.
Good post despite the obligatory Trump bashing virtue signal