Truth as an Epistemic Construct

BV has an interesting post on the claim that "p is true =df p would be accepted in cognitively ideal conditions." The obvious problem is that being in cognitively ideal conditions would change what's true -- a classic demonstration of the conditional fallacy. Still, it isn't hard to find a work-around for this technicality; simply ask what an ideal spectator would think is true of our world. We might frame this in terms of ideal advice -- what would my idealized self advise me to believe in my actual condition? (Surely that I'm non-ideal, for one thing.) So, one way to overcome the problem is to isolate the idealized vantage point from the actual situation being assessed, i.e. by situating the ideal observer in a different possible world.

BV preempts this response:

But if cognitively ideal conditions are never attained, but are merely possible conditions, then it is difficult to see how the actual truth of any proposition could be identified with what investigators in some merely possible circumstances accept. How can what is actually true be explicated in terms of merely possible judgments?

I take it BV is understanding these ideal agents to be merely investigating their own, non-actual, world. Even then, it doesn't seem so problematic if we've stipulated that their world is a perfect duplicate of ours in respect of the particular details under investigation. (It doesn't seem that bizarre to have the truth be fixed by the best judgment of all possible agents in identical situations.) Note that the normative status of an ideal judgment is unaffected by its existential status. So the only problem would be that their judgments concern unreal things; it's not a judgment's being "merely possible" that's cause for concern, but rather that a judgment is of the merely possible! But this is no longer a problem if the possibility under assessment is a duplicate of ours in the relevant respects, since then the discerned non-actual truth will exactly mirror the actual truth.

Of course, there's a simpler response, which is to have the ideal agents investigating our possible world. Then it's no mystery at all how their (non-actual but ideal) judgments could be relevant to what's actually true. [Granted, you must be part of a world to investigate it using empirical methods, so our ideal observers can't do that. But they can draw implications, i.e. given the totality of experimental data about the actual world, what should we conclude?]

And then there's an even simpler account, which steers well clear of the conditional fallacy by avoiding any kind of conditional analysis at all. Rather than defining epistemic normativity in terms of primitive modality ("what an ideally rational agent would believe"), we may just as well take normativity (ideally rational beliefs) as primitive -- and, if we like, derive modality from there. Epistemicists can then simply define 'p is true' as 'p is epistemically ideal', without appeal to confounding counterfactual situations.

So, I don't think BV has posed an insuperable problem for epistemicists about truth. A bigger problem, to my way of thinking, is that we can clearly conceive of possible worlds where the base facts elude all epistemic grasp. Contingent truths are made true by the world, which is objective / mind-independent / evidence-transcendent. Still, I'm sympathetic to a circumscribed form of epistemic constructivism, concerning necessary (a priori) truths. As I put it here, "philosophical truth just is the ideal limit of a priori inquiry; it does not answer to the sort of independent reality that might sensibly be considered beyond all epistemic reach." After all, it's awfully hard to see what in the world philosophical truths are meant to correspond to. (That big rock over there?) Far more natural to see it as a rational construct, or intangible ideal, that doesn't literally correspond to any mundane thing.