Some clairvoyance and (perhaps?) regress problems with Goldman's "Epistemic Folkways"

In his paper “Epistemic Folkways and Scientific Epistemology” (reprinted in Axtell’s Knowledge, Belief and Character and originally published in Goldman’s own Liasons: Philosophy Meets the Cognitive and Social Sciences), Goldman undertakes the project of delineating and defending both a descriptive and a normative form of what he calls “scientific epistemology.” His term “scientific epistemology” is warranted, Goldman thinks, due to what he takes to be crucial dependences on the cognitive sciences in meeting the “two missions of epistemology” (to describe our folk concepts, and to evaluate and improve upon them).
What I found particularly interesting occurs in the first section of Goldman’s paper, in which he attempts to provide a Sosa-amended reliabilist account of justification, which he then defends against a few classic counterexamples which have plagued reliabilist positions in the past.
I take his position to claim something of the following:

S’s belief that p is justified iff S’s belief is a result of processes all of which are on the epistemic evaluator’s list of intellectual virtues. If S’s belief is a result of processes, any of which are on the epistemic evaluator’s list of intellectual vices, the belief is unjustified. If the processes which engender S’s belief are on neither list, the belief is non-justified.

How does a virtue or vice get on the epistemic evaluator’s list? At this point, (obviously) there is an appeal to reliability—processes that are deemed to produce a sufficiently high ratio of true beliefs will make the evaluator’s list of intellectual virtues.
Importantly for Goldman, because we are giving at this point a “descriptive” account of the epistemic concept of justification, we should not (he claims) stray from what the folk would agree upon. And hence, a good test question for determining whether a candidate virtue makes the evaluator’s list is to suppose we ask the folk, “How does S know that p?” Answers, Goldman thinks, would include such processes as: “Because he saw it, because he heard it, because he remembered it,” etc. These are the sort of processes Goldman has in mind.

A problem I think his account doesn’t resolve, though, is why we should not include clairvoyance on the list of intellectual virtues given his criteria for what should arrive on the evaluator’s list. I don’t say this because I think clairvoyance is, in fact, a virtue, per se, but because there seems to me no clear reason why Goldman should, given his reliability criterion, be in a position to weed it from his evaluator’s list. A good clairvoyance case, which he mentions, is the case of “Norman” (borrowed from BonJour), who has the reliable faculty of clairvoyance, but who doesn’t have any evidence for or against having this faculty. Goldman thinks that the folk intuition here is that Norman’s clairvoyance-based beliefs are not justified, and so, because Goldman seems to allow folk-approved reliable processes a place on the evaluator’s list of virtues, he has some explaining to do if he is to preserve the folk intuition that such beliefs aren’t justified.
His explanation, I think, is unsatisfying. He makes an appeal to what he takes to be the folk trait of “categorical conservatism”: the folk, he think, display a preference for “entrenched” categories; “they do not lightly supplement or revise their categorical schemes.” And so, Goldman thinks, clairvoyance wouldn’t make the list because the folk are hesitant to add it, even though the ones they do add are added for the same reason that clairvoyance would be added—namely—because it is a sufficiently reliable process.
One question we should ask here is: is Goldman right that the folk would, in fact, be inclined to take clairvoyance off the list of virtues (in light of the fact that other virtues are on their for having the process of reliability that clairvoyance possesses). I’m worried that, if he is right about this, then prospects appear dim for Goldman’s appeal to reliability as a basic concept within his theory, which is how he seems intent to present it. If justification is explained in terms of intellectual virtues, and they are explained in terms of reliability (as Goldman seems to want), then reliability must be basic. But no longer is it basic if it is not capable of explaining the justificational status of beliefs arising out of a process of clairvoyance. Such beliefs should be able to be identified by recourse to reliability (for his theory to preserve the heirarchial structure he appears to want). And so, dropping clairvoyance from the list is at the expense of the clean heirarchial structure he wants in his theory.
Another problem I’d like to explore with Goldman’s position in this paper is that it plays, I think, the dangerous game of falling into regress. My concern here might be quite misguided, to warn. It is as follows: If a criterion Goldman uses to determine whether a belief is justified is whether the folk would deem it reliable (i.e. the criterion he uses to deem clairvoyance non-evaluator-list-worthy), then it seems as though he is positing this criterion as at least a necessary condition for a candidate virtue’s making the list. I wonder, though, how broadly this criterion can be applied. If Goldman weeds virtues from the evaluator’s list by virtue of the fact that the folk’s “categorical conservativism” would not allow it, then why should we not think that folk acceptance should also be applied as a criterion at the level of “theory” acceptance, as well as process acceptance. He doesn’t give a reason to think that there is a relevant factual difference between processes and theories that would justify folk acceptance as a criterion for one but not the other. But as soon as we apply folk acceptance to the level of theory acceptance, then it seems dubious to suspect that the “folk” are going to be inclined to understand justification in terms of such technical machinery as “evaluator’s lists” and “truth ratios” and (in cases like clairvoyance) even reflection on their own categorical conservatism as needed for reference within a theory of justification. And hence, the regress is: if folk intuitions can be stipulated as criteria for ruling out processes that would make evaluator’s lists, they would also most likely rule Goldman’s own theory of justification which makes references to enough machinery to be non-folk-approved at the theory level as clairvoyance would be non-folk-approved at the process level. Of course, Goldman has a way out if he can demonstrate the relevant factual difference between processes and theories that justify folk approval as a criterion for one but not the other; however, I’m not sure what such a factual difference could be.