uncertainty is often more Knightian than Bayesian which makes different things appropriate
Equivalently speaking, if you can apply Bayesianism in a general enough setting, then you can trivially reduce the epistemic uncertainty to 0 in at least some environments, so Bayesianism is superfluous.
Specifically, the problem arises because Bayesianism assumes logical omniscience, but that is tantamount to at least having a halting oracle, specifically here a machine that can at least do infinite computation in finite time, and in that case, it is trivial to know all the recursively enumerable set of logic with certainty, because we can simply enumerate the set of theorems in math, and thus we don’t have to deal with uncertainty at all, since many, many problems (not all of them) are either a strict subset of the recursively enumerable set or are equivalent to the set of recursively enumerable turing machines, so uncertainty didn’t matter at all, so Bayesianism is superfluous.
Equivalently speaking, if you can apply Bayesianism in a general enough setting, then you can trivially reduce the epistemic uncertainty to 0 in at least some environments, so Bayesianism is superfluous.
Specifically, the problem arises because Bayesianism assumes logical omniscience, but that is tantamount to at least having a halting oracle, specifically here a machine that can at least do infinite computation in finite time, and in that case, it is trivial to know all the recursively enumerable set of logic with certainty, because we can simply enumerate the set of theorems in math, and thus we don’t have to deal with uncertainty at all, since many, many problems (not all of them) are either a strict subset of the recursively enumerable set or are equivalent to the set of recursively enumerable turing machines, so uncertainty didn’t matter at all, so Bayesianism is superfluous.