Friday, February 25, 2022

Matryoshka

 Those are the little Russian dolls, nested into each other.

Reading people's comments about the truckers, I am wondering how much one issue depends on others that are not fully stated.  That's actually not a good analogy with the dolls, but it is what I think of.  (Update:  I liked it better the longer I went.) I am fond of noting that if two events that are 70% likely are stacked on each other, the resulting probability is less than 50%.  If the events are 80% likely, it takes three of them to get to 50-50, and if one event looks like if follows from another about 90%, then another very likely event then another, you get more slowly to a probability of 50%, but you get there, even with very likely chances. It's just one of the things that makes future prediction or current analysis difficult.

I was thinking of this in terms of explaining what is up with the Canadian trucker, but remembering the principle reminded me of my analysis of the probable IQs of Barack Obama, both Bushes, John Kerry, Donald Trump, etc. In all cases with people of any accomplishment there can be distraction, because even getting into an Ivy League school or writing a book or making a lot of money likely depended on significant other factors. IQ is something, but it ain't everything, not by a long shot. In all cases I tried to clear away low-information events and get to a very few key points. (Sometimes that point might take a long explanation, but it should be clearly conceived and identified.) This is because I know that even high-percentage likelihoods, if they are stacked on top of each other, rapidly become less likely. Because of this bit of biographical data it is 90% likely that Obama is at least this but not that.

It also comes up in alt-history alt-future evaluations, and in wargaming, or in looking at the effects a policy will have on the economy.  Every step of the analysis might be quite good, so stringing it together as an argument, an attempt to prove an hypothesis, seems very plausible, even probable.  But there might be too many rolls of the die in there. If you consider the Matryoshka dolls, even if each one contains a doll that is 80% its own size, it doesn't take long to get to much smaller dolls. And if you look closely at all dolls, as with all reasons that drive similarity, each level in is not really identical to its shell. Once you are many shells in there is a simplified version, which could if reversed give rise to many variants by the time it reached the outermost shell.

That said, I think you can parallel the two sets of dolls WRT the truckers. 

People do not like protests blocking traffic when it is a cause they disagree with or are neutral about. A younger friend mentioned that she, a libertarian, and her husband, a Democrat, were trapped in a car for hours with a screaming three-year-old because of a BLM highway closing. Neither has forgotten it, and the group lost support that could have been considerable in places like Massachusetts. 

Update: One of my original main points got left out.  This attitude certainly affected my disapproval of the George Floyd/BLM/Trayvon/Ferguson/National Anthem etc protests.  I tried to regard each as objectively as I could, but the substrate was always "Except what you are insisting should be fixed is actually not much broken. The police are not really targeting young black men. It is widely believed but poorly evidenced. Therefore you are advocating for changes that cannot actually happen, but only addressed with false symbolism."

Even with a cause they agree with, it matters whether they personally encounter it or not. And continued support is likely to be more grudging. I think this is true of disruptive protests of all kinds, not just traffic. College students, from other states or other parts of the state, become notoriously unpopular for disruptive protests which they can avoid the effects of by return to their nice campuses.  The truckers, who work on the roads and therefore think of the roads as their territory in a sense, are nonetheless not from the locations of the disruptions.  They are from all over. We tend to miss this aspect when looking at another country, or a distant part of their own. To us, they are all equally just Bolivians, or Russians, or Canadians. We think of them as outsiders to us, but intuit that they should regard each other as "all in this together." They don't, and we don't when it is us.

Those  who believe that vaccine mandates are a reasonable response to an epidemic are less likely to be sympathetic to a protest against vaccination by outsiders. Those who believe vaccine mandates are an unwarranted intrusion on freedom are more likely to give approval.

Next Matryoskh: Those who believe that the vaccines don't work as well as advertised - and certainly those who think they are useless or dangerous - are more likely to think of them an unwarranted intrusion. Those who believe they are an effective tool for reducing many effects of an epidemic are more likely to approve of mandates.

Next: Those who have more trust in the general expertise of the standard-definition experts are more likely to think vaccines effective, as their impression is that those are strongly supportive of their safety and effectiveness. Those who believe that standard-definition experts are generally overrated anyway are more likely to not only be suspicious of their advocacy, but sometimes even reflexively rejecting of them. 

A side mention here is that those who believe the standard experts are overrated are very likely to accuse those who believe them as trusting the government. While there are likely some who just trustingly go along with what people in authority tell them is right, I think this accusation is mostly just convenient bullshit. I see the accusation frequently, and always from people who can't be reasoned with. This is one of the points that moved me from initial skepticism about a lot of the scientific explanations - without prejudice, as this was new territory for all of us, and many modified their recommendations as new information emerged - to more acceptance of the idea they were slowly getting it generally right, not getting farther off the mark. Their critics got crazier and crazier, doubling down when proven wrong. I don't know anyone who believes the vaccines are good because the government told them so, or even that the CDC and FDA told them so. People go to the CDC for numbers and for understanding what regulations they personally are going to have to implement or work around. They have all gone looking for various other authorities they trust, ranging from stuff they themselves find persuasive reasoning to recommendations from people whose opinions they trust. 

We are quite a few dolls in on both sides here. Are there an unlimited number, so that if we could see that far in we could find a single off-on switch in each of our personalities that slowly drives the design of each doll as we move outward? I had a lengthy series over a decade ago of May We Believe Our Thoughts (MWBOT), and there has been further neurological discovery since then. I will give a quick caveat to be careful, as some things that were cutting-edge research in the 2000s, prompting a fair bit of theorising and rejecting of previous research, are now themselves rejected because they did not replicate.  In particular, the neurology of choice keeps upending.

Next dolls in: Which experts we trust has a lot to do with how well we understand the credentials they are bringing to their statements. It is not that we have those credentials ourselves, not even junior versions of them, but how we we understand what goes into them. I expect a person with training in NT studies to have a fair knowledge of documents and early controversies. A person who is a successful consultant on church growth might or might not. I would have to know more. And vice versa.  A plumber may have worked with many electricians during installations and know a great deal.  Or the plumber may think he knows, but 25% of his knowledge of electricity is common myth. In medicine, a surgical nurse will likely know a great deal about contagion at close range, but whether she knows much about epidemiology is uncertain.  She will at minimum have some foundational medical knowledge, more than the average bear.  But that may be it.

We have zero expectation or requirement that the school superintendent or school board know much medicine.  Therefore it matters how much we trust their ability to get good advice.

This can get tricky. In current human genetics debates, training in a number of fields might be helpful, including a statistician who does not know much cell biology. Such a person may know many things that people with PhD's in microbiology don't. Science editor Nicholas Wade's book was widely condemned by people with excellent credentials.  But the primary few experts in the field pointedly refused to sign on to the condemnation. (Quick. Guess the politics!) The baseball statistician Bill James repeatedly made the claim that even though lifelong scouts and coaches might understand many things he did not about how a swing developed in a player or how to motivate angry or tired athletes, many of them did not know some ordinary percentage moves that players of table games did. And he in particular knew a lot of things they did not.  He turned out to be spot on. 

In medicine, I listen attentively to what a neurologist might say about psychiatry. An endocrinologist, sometimes. A surgeon likely knows much less than I do, even with the medical degree and training. Because Covid does actually touch on many fields and knowledge of it is interdisciplinary, we end up with online conversations where second-level information actually does become important. "Her nutritional theories were rejected twenty years ago, but are gaining wider acceptance now." Are they? Or does her website just collect a few approving comments a year by a person of some vaguely related credential? "None of the people in that study were epidemiologists." Is this particular question actually an epidemiological one?

With less-formal or less-recognised credentials, this gets even harder. As a person who made my living with less-formal credentials, who also thinks that some of the formal credentials are actually a hindrance to understanding, I have a ready sympathy for the possibility that such ground-up rather than top-down learning might be a lot better. (Or that the ground-up person might also have top-down learning that they did on their own.) Yet I also know people who came up exactly as I did who are prize fools, and you can't tell them anything. The formally credentialed people were generally better, and in some aspects, incomparably better. Just is.

Thus understanding the credential behind the assertion matters, and matters in the next doll down. None of us likes being told what to do, and the source of authority of the orderer matters to us.  When traffic grinds to a halt, we trust the woman who is is making knowledgeable-looking hand gestures redirecting us off onto another road.  We believe she has knowledge we do not, and if she is additionally a policewoman or fireperson or has a construction vest and hard hat, we figure she has some official authority as well.  That's usually fine for us.  It is frustrating when you want to say "But officer if you just let me duck up to that little turn there on the left a bit, I can be out of everyone's hair completely." But we live with it. Or most of us do. There are some of us who always think they know better and want to prove it. Other orders we filter through a highly-varied set of evaluative tools - and it is at this point that all sorts of other stuff finally emerges.  It turns out that a lot of vaccine-avoiders just don't like needles. When asked if they'd like the shot while they are in for an unrelated surgery, a lot apparently say yes. So cutely calling the vaccine a "jab" or a "stab," and displaying a big needle prominently on the posters meant to encourage them to show that we consider it a routine procedure we aren't afraid of turn out to be very stupid strateies for talking people into it. Whether the listener likes us or likes our manner, with all the demographic and cultural baggage that includes turns out to matter as well. Accusations of all sorts of bad motives in those Other People begin to come out. "They always want to tell us what to do." "Of course they think that because that's where their money comes from." "They just don't like Trump." "They just hate Biden." These draw power from the fact that in some cases, they're true. For some people, not the federal government is a deep-down doll of its own which builds every shell around it over and over again in an utterly predictable way.  Sometimes, we knew right from the start what Al was going to decide, whatever he was saying out loud.

We get one doll down and find that some people are always thinking "I just want to be left alone and bothered as little as possible, so I got the shot," while others just don't like being told what to do ever, about anything. Reasoning with them just makes them get their backs up.

At the stage of the argument where it has gone on long enough lots of people settle in to assuming that everyone who disagrees with them must be operating from a single or very few deeply interior dolls. After all, this doll looks an awful lot like the ones outside it, all the way to the final exterior.

Where I started is where I will end. All the 70, 80, 90% chances were reasonable in the chain. Each doll does look like the one above it and the one below it. Yet taken together, they are not a convincing explanation for the behavior of a large group of people.  Even if they are a convincing explanation for the behavior of some people, they don't cover enough variation.  I started uncovering a set of dolls about who supports the truckers and who doesn't, and it looks good at first.  I certainly thought so, which is why I pursued it. Yet I wasn't too many dolls in before I thought "This doll doesn't look entirely like the outside one, does it?"

5 comments:

  1. "I am fond of noting that if two events that are 70% likely are stacked on each other, the resulting probability is less than 50%. If the events are 80% likely, it takes three of them to get to 50-50..."

    I'm not sure what you are trying to say here, but I think you got some numbers mixed up. Yeah, if the probability of two events is each 0.7, then the probability of both is 0.7*0.7=0.49, or less than 50 percent. But for two events with a probability of 0.8, the probability of both happening is 0.8*0.8=0.64, or considerably HIGHER than 50 percent. Am I misinterpreting what you are saying?

    ReplyDelete
  2. You can't do effective marketing/persuasion without making some attempt to understand what your potential customers are like and what they view as important. The tendency to view all of one's opponents are sharing an extensive set of characteristics ("all conservatives are ignorant" "all leftists are lazy and don't want to work") is a big inhibitor to effective persuasion.

    ReplyDelete
  3. Uncle Bill - you're calculating the probability of two *independent* events occurring where increasing the probability of the event increases the overall probability. The classic example is drawing a colored ball from a container, returning it, and drawing a second time. As you increase the percentage of balls of a given color you increase the probability both draws will be that color. The rule for dependent events is that the probability is less than the probability of similar independent events, i e. rolling a 5 *and* a sum of 8 with two dice can only happen 1 way out of 36 combinations even though there are 4 ways to sum 8 and 2 ways to roll a 5. You do usually see the probabilities used to determine that the events are dependent but the reverse assumption works as well.

    ReplyDelete
  4. James has it, very efficiently. Two independent events at 80% is 64%. A third independent event is 80% of that, 51.2%. I said "it takes three of them," which was clear in my mind, but not be quite so obvious on the page.

    ReplyDelete