Scribal Roundup #1
Thoughts on things I've seen out there
1] Scott Alexander writing about selection bias.
He compares accusations of selection bias in online Internet surveys vs. selection bias in many psychological studies, where the population is Psych 101 students. An interesting point, but I think there is a difference, depending on which online survey you are talking about. For instance, if you share survey or poll within a specific community that aligns with your goals, then some people are motivated to answer and to share the survey with like minded people, and it will reflect that motivated group’s opinion.
For him, it wasn’t that. He was trying to fact-check this tweet:
and consulted the 2020 Slate Star Codex survey, which had collecting information about depression and so on, as well as asking if they had had a mystical experience. Even though only only 13% of the SSC Community define themselves as “Rationalists”, whatever that means to them. (Though enough outside observers thought otherwise that this merited its own post.) But there is something unique about the attributes of this population and opt-in community that makes me think that you cannot extrapolate to the population at large, about this particular question. Just as if you asked Psych 101 students about their attitudes towards Scientology, therapy, or psychotropic drugs.
2] A Review of ‘Escape From Model Land’ in the WSJ. An excerpt:
The problem is that Model Land is easy to enter but difficult to escape. Having built “a beautiful internally consistent model,” Ms. Thompson writes, it can be “emotionally difficult to acknowledge that the initial assumptions on which the whole thing is built are literally not true.”
There are all sorts of ways that models can lead us astray. A small measurement error on an input can lead to wildly inaccurate forecasts—a phenomenon known as the Butterfly Effect… Far more problematic are what Ms. Thompson calls “unquantifiable unknowns”—things that are left out of a model’s calculation because they can’t be anticipated, such as the unexpected arrival of a transformative technology or the abrupt collapse of a robust market… Beyond the inherent inability of models to account for the unaccountable, models also reflect the biases of their creators. We may be inclined to regard models as objective expressions of truth, yet they are deliberately constructed interpretations, imbued with the values and viewpoints of the modelers—primarily, as Ms. Thompson notes, well-educated, middle-class individuals.
I agree with this idea that there are inherent limitations of models, but that people at large don’t realize the limitations and sometimes grant it too much authority.
3] At the New Yorker, The Case for Wearing Masks Forever. This was shared and criticized online for being too pro-masks. But likely by people who didn’t read past the headline. Read Emma Green’s tweet thread to how she really feels.
4] In honor of Chanukkah, a 2004 paper: Dreidel Lasts O(N^2) Spins. It begins:
Dreidel is a popular game played during the festival of Chanukah. Players start with an equal number of tokens, and contribute one token each to a common pot. They then take turns spinning a four-sided top, called the dreidel, and depending on the side showing up, the spinner does one of the following:
Nisht (N) : Nothing.
Ganz (G) : Takes all the tokens in the pot.
Halb (H) : Takes (the smaller) half of the number of tokens in the pot.
Shtel (S) : Donates one token to the pot.Whenever the pot is empty, all the players ante up, i.e., donate one token each to the pot. Players lose, and go home, when they are required to donate a token to the pot, but cannot. The last survivor wins. The winner also goes home.
Feinerman [2] and Trachtenberg [3] investigated the fairness of a simplified model of dreidel. Zeilberger [4] conjectured that the expected number of spins in a game of dreidel between two players starting with n tokens each is O(n^2 ). Later, Banderier [1] conjectured that even in a multi-player game, the expected number of spins before one of the players goes home is O(n^2 ). We show that the expected duration of a game of dreidel where the players start with n tokens each is O(n^2 ), irrespective of the number of players.
In my Object Oriented Software Design class, I have them run Monte Carlo simulations of Dreidel to empirically reproduce the results of the Feinerman paper on Dreidel fairness. Maybe I’ll have then reproduce this result as well.
5] All of the Twitter Files are worth checking out, but see this entry at The Free Press. Much of the material is upsetting — how free speech and discussion about important public health issues was suppressed in the public sphere, including by political actors.
6] Finally, of my recent posts on this Substack, I’d recommend this one:
