Your strength as a rationalist is the degree to which it takes “very strong and persuasive argumentation” to convince you of false things, and “weak, unpersuasive-sounding argumentation” to convince you of true things; ideally, in the latter case, the empty string should suffice.
Regarding volatility:
Though: maybe it just works out the same? E.g., the average of your estimates over time needs to obey Bayesian constraints?
Augenblick 2021 shows (see “Proposition 1”) the total belief movement should match the total uncertainty reduction in expectation for Bayesian updating. If a credence moves from p1 to p2 to p3, the total belief movement is (p2 - p1)^2 + (p3 - p2)^2, and the total uncertainty reduction is p1 (1 - p1) - p3 (1 - p3). The article describes many other Bayesian constraints!
Thanks! Re: one in five million and .01% -- thanks, edited. And thanks for pointing to the Augenblick piece—does look relevant (though my specific interest in that footnote was in constraints applicable to a model where you can only consider some subset of your evidence at any given time).
Great post, Joe!
Nitpick, one in five million.
Nitpick, maximum of 0.01 %.
Quoting Eliezer Yudkowsky:
Regarding volatility:
Augenblick 2021 shows (see “Proposition 1”) the total belief movement should match the total uncertainty reduction in expectation for Bayesian updating. If a credence moves from p1 to p2 to p3, the total belief movement is (p2 - p1)^2 + (p3 - p2)^2, and the total uncertainty reduction is p1 (1 - p1) - p3 (1 - p3). The article describes many other Bayesian constraints!
Thanks! Re: one in five million and .01% -- thanks, edited. And thanks for pointing to the Augenblick piece—does look relevant (though my specific interest in that footnote was in constraints applicable to a model where you can only consider some subset of your evidence at any given time).