This is a special post for quick takes by aarongertler. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
I've been enjoying the Sold a Story podcast, which explains how many schools stopped teaching kids to read over the last few decades, replacing phonics with an unscientific theory that taught kids to pretend to read (cargo cult vibes). It features a lot of teachers and education scholars who come face-to-face with evidence that they've been failing kids, and respond in many different ways — from pro-phonics advocacy and outright apology to complete refusal to engage. I especially liked one teacher musing on how disconcerting it was to realize her colleagues were "refuse to engage" types.
The relatable topic and straightforward reporting make the podcast very accessible. It's a good way to share a story with people outside the LessWrong bubble that may get them angry in a way that supports rationalist virtues.
This is a horrifying story. I am still at the beginning, because it is very long, and I probably won't finish it, but...
If I understand it correctly, a few decades ago American schools taught kids reading the proper way (which is now called "phonics"), and then some people invented a "better" way, which was basically... a combination of memorizing the text, and guessing the words you had a problem to remember read. Other strategies included trying to guess the word from the context, the picture at the story, the length of the word and its first and last letters.
And it seemed to work great... at very specific situations. If the teacher read a story to children first, their reading improved a lot. That's because small children have insane memory skills, compared to adults, so if you read them a story once, they will memorize large parts of it. If you also teach them to just look at the first and the last letter of the word and try to guess what it is, that makes the recall almost perfect. So it seems like the method was experimentally verified!
Of course, the kids didn't actually get good at reading texts they haven't read before. Which is the point of reading.
(...some more points added as I read..)
Ironically, the wrong method started with its author observing the kids who were good and bad at reading, and asking "what do the kids who are good at reading do differently?". And she concluded that the kids are reading so fast that there is no chance they could be actually reading every single letter and decoding every single sound... they must be guessing. Therefore, to teach other kids to be also good at reading, we much teach them to guess.
Which sounds kinda plausible... and also consider the fact that kids taught the wrong method are actually pretty good at reading (texts that someone else has already read to them). Would you be convinced?
Only a few years later, the invention of eye tracking showed that, actually, the good readers do read all the letters; they just do it very fast because they have automated the skill. But this only happens if you first teach them to do it the right way, and then they automate it. (If they automate the wrong way, the guessing, they will keep occasionally guessing wrong.)
To make things even more complicated, you can actually write books that the kids taught the wrong method can read fast, even if they have never heard them read before. You do that by using a very limited vocabulary, repeating the same few words over and over again in new combinations, so that looking at the first letter of the word already allows you to guess the word pretty reliably.
Of course, everything can become worse when politics gets involved. So, at some moment, some teachers who understood science tried to convince about it the president... George W. Bush. He listened to their arguments, agreed with them, and decided to fix education. But then 9/11 happened and he was distracted. Meanwhile, the fans of the guessing method used this to discredit phonics. "Do you really want to teach your kids the method that Bush endorses?" (Remember, most teachers are Democrats.) "Why don't you trust the science instead?" (Of course, everyone can call their ideas "science", and few people will actually verify that.)
And even at places where phonics was made mandatory, the proponents of the guessing method adapted by giving some lip service to phonics first, calling it one of many methods the children can use. But then, they just taught children to use the other methods (guessing from context, from word size, from the first and last letters).
Oh, it gets even better. At some moment the proponents of the guessing method invited some actual scientists to measure whether additional lessons in guessing improve kids' reading. The conclusion of the scientists was that the additional lessons improve reading of stories that are accompanied with pictures, but do not improve reading of stories without pictures. This later got quoted as "independent scientists confirmed that the lessons improve reading".
Epistemic status: Neither unique nor surprising, but something I felt like idly cataloguing.
An interesting example of statistical illiteracy in the field: This complaint thread about the shuffling algorithm on Magic: the Gathering Arena, a digital version of the card game. Thousands of unique players seem to be represented here.
MTG players who want to win games have a strong incentive to understand basic statistics. Players like Frank Karsten have been working for years to explain the math behind good deckbuilding. And yet, the "rigged shuffler" is a persistent belief even among reasonably engaged players; I've seen quite a few people try to promote it on my stream, which is not at all aimed at beginners.
(The shuffler is, of course, appropriately random, save for some "hand smoothing" in best-of-one matches to increase the chance of a "normal" draw.)
A few quotes from the thread:
How is that no matter how many people are playing the game, or how strong your deck is, or how great your skill level, I bet your winning percentage is 30% or less. This defies the laws of probability.
(No one ever seems to think the shuffler is rigged in their favor.)
As I mentioned in a prior post you never see these problems when they broadcast a live tournament.
(People who play in live tournaments are much better at deckbuilding, leading to fewer bad draws. Still, one recent major tournament was infamously decided by a player's atrocious draw in the last game of the finals.)
In the real world, land draw will not happens as frequent as every turns for 3 times or more. Or less than 2 to 3 turns, not drawing a land
(Many people have only played MTG as a paper game when they come to Arena. In paper, it's very common for people to "cheat" when shuffling by sorting their initial deck in a particular way, even with innocuous intent. When people are exposed to true randomness, they often can't tolerate it.)
Other common conspiracy theories about Arena:
Players of Battle for Wesnoth often accuse random number generator of being broken, e.g. when their unit has 3 attacks, each of them has independently 70% chance to hit, and all three attacks happen to miss. But the chance of that happening is actually 2,7%, and if a level takes twenty or more turns, and in each turn several units attack, this is likely to happen several times per level.