The criterion for assessing truth must be known to use it.
This seems like the bad assumption to me. (This is quite a bit different from your explanation of the criterion in the linked article, so I’m commenting on this one separately.)
For example, take a foundationalist bayesian epistemology, where Solomonoff induction is considered supreme. The justification of a degree-of-belief rests in (a) the observations so far, (b) the complexity of the hypotheses, and (c) the bayesian math connecting the two to produce a posterior.
The observations (a) have to be known in order to be used, but the reasoner does not need to have any explicit knowledge of the criterion (b and c) in order to use them correctly. It can simply be built into the way you reason. So within this frame, the criterion does not have to itself be known in order to be used.
More generally, if one supposes that there is a good criterion, it seems like all that is necessary (in order to arrive at normatively correct beliefs) is to apply the criterion correctly. It seems there is no additional reason why one must know the criterion explicitly.
So, no circular reasoning arises.
(But note that I’m saying this more to see what reaction it gets, rather than as a proposed solution to the criterion problem. The criterion problem seems to be pretty vague, so I’m only claiming this as a possible response to the specific version here.)
I think this is just sloppy writing on my part. When I wrote this I was trying to iron out some points of confusion and disagreement I had been addressing in comments on previous posts. I think this 3-part framing in this post is probably misleading and should be ignored.
I was trying to find a crisp way to explain why I think the thing being pointed at by the problem of the criterion is everywhere and relevant to a large number of problems. I think this particular presentation is bad.
This seems like the bad assumption to me. (This is quite a bit different from your explanation of the criterion in the linked article, so I’m commenting on this one separately.)
For example, take a foundationalist bayesian epistemology, where Solomonoff induction is considered supreme. The justification of a degree-of-belief rests in (a) the observations so far, (b) the complexity of the hypotheses, and (c) the bayesian math connecting the two to produce a posterior.
The observations (a) have to be known in order to be used, but the reasoner does not need to have any explicit knowledge of the criterion (b and c) in order to use them correctly. It can simply be built into the way you reason. So within this frame, the criterion does not have to itself be known in order to be used.
More generally, if one supposes that there is a good criterion, it seems like all that is necessary (in order to arrive at normatively correct beliefs) is to apply the criterion correctly. It seems there is no additional reason why one must know the criterion explicitly.
So, no circular reasoning arises.
(But note that I’m saying this more to see what reaction it gets, rather than as a proposed solution to the criterion problem. The criterion problem seems to be pretty vague, so I’m only claiming this as a possible response to the specific version here.)
I think this is just sloppy writing on my part. When I wrote this I was trying to iron out some points of confusion and disagreement I had been addressing in comments on previous posts. I think this 3-part framing in this post is probably misleading and should be ignored.
I was trying to find a crisp way to explain why I think the thing being pointed at by the problem of the criterion is everywhere and relevant to a large number of problems. I think this particular presentation is bad.