OP is using a lot of words to describe what seems to be the "propensity" interpretation of probability [1]
The propensity interpretation generalizes the relative-frequency interpretation to allow for single-shot "propensities" that may be constrained by (e.g.) physics, but are not intended (or required) to be repeated.
The example I've seen is the chance of a tossed thumbtack to land point-up. There are odds associated with this, but you don't need a relative frequency interpretation to quantify them.
I think it's also relevant to other non-repeatable events like "rain tomorrow" or some of the other forecasting scenarios Scott is listing. The examples he's choosing tend to kick up a lot of dust ("drinking monkey blood causes autism") and he's got a chip on his shoulder about it all, so it's not really pleasant to engage with.
Guy who has Bayes' theorem as his column description is a fan of propensity? What's going on? It's still there, I checked. "P(A|B) = [P(A)*P(B|A)]/P(B), all the rest is commentary." How many non-Jews get the "all the rest is commentary" part?
For your thumb tack example you can at least do some simulation. My reading of the article is that the author argues that it’s totally reasonable to use 5% when you have no idea but want to sound more precise than “I feel like it’s unlikely”
You can reverse engineer probabilities (or P-words if you don't want to call them that) from the bets you would make. Obviously there's a problem there if everyone's dead, but they don't have to be literal bets. Someone could estimate your P-word of doom by what you say and how you think about things/act. The article argues that it's fine for you to just come out and say it.
God I hate how poorly substack handles pages with large numbers of comments. Because the entire comment tree is reloaded synchronously from the server over and over again, I can't scroll on my desktop without major hitching and hangs, I can't scroll on my mobile without the content completely disappearing for anywhere from 2 to 12 seconds, and so I just don't read scott anymore.
The propensity interpretation generalizes the relative-frequency interpretation to allow for single-shot "propensities" that may be constrained by (e.g.) physics, but are not intended (or required) to be repeated.
The example I've seen is the chance of a tossed thumbtack to land point-up. There are odds associated with this, but you don't need a relative frequency interpretation to quantify them.
I think it's also relevant to other non-repeatable events like "rain tomorrow" or some of the other forecasting scenarios Scott is listing. The examples he's choosing tend to kick up a lot of dust ("drinking monkey blood causes autism") and he's got a chip on his shoulder about it all, so it's not really pleasant to engage with.
[1] One take, although there is a literature: https://en.wikipedia.org/wiki/Propensity_probability#:~:text....