Two realisms enter, only one can leave
I gave my Saturday over to the UAlbany Grad Student Philosophy Conference, and I am glad I did. There were some very good papers. Props are due to the grad students who organized it. This post records a thought I had during the conference. I'll start by explaining the point in terms of Philip Kitcher's metaphysics and epistemology, but I think the problem generalizes.
PK is committed to relativism about epistemic significance but realism about scientific claims. This combination yields a pluralist realism; cf. my previous posts here, here, and here. In this replete realism, our interests determine what language we speak and the world determines what sentences are true in that language.
PK is also committed to reliabilism, the view that a true belief counts as knowledge if it results from a reliable process. Crudely, a process is reliable if it has the propensity to produce more true beliefs than false ones. This requires that there be some fact of the matter about which process type produces a particular belief token. Failing that, there will be no determinate answer as to whether the belief is produced by a reliable process or not; it might be an instance of some reliable types but also of some unreliable types. This is the generality problem for reliabilism.
The generality problem can be answered by insisting that there is some determinate process that produces each belief. Even though we can't say with any great confidence what the process is, we can appeal to a general doctrine of psychological realism (Alston) and write a promissory note to be cashed in with future cognitive science (Goldman).
This response to the generality problem requires that future cognitive science will or at least could give us a unitary answer to questions of the form "What process produces belief X?" According to the pluralist realist, however, future cognitive science will depend on our concerns and interests. There is no reason to think that identifying cognitive processes is any less interest-dependent than identifying (eg) species. So the pluralist realist seems unable to accept the psychological realist solution to the generality problem.
There seem to be three options:
1. The interests and projects of epistemologists are sufficiently precise that they strongly constrain what future cognitive science could mean by "cognitive process." There need not be a single, inevitable possibility, but the possibilities must be sufficiently constrained so as to allow for reliability judgments.*
2. The generality problem may be answered in some other way, saving reliabilism without psychological realism. (Good luck with that.)**
3. Pluralist realism and reliabilism are inconsistent. At least one must be abandoned.
* This approach tries to rein in the pluralism, at least for cognitive science. This does not really seem to be an option for PK, who is committed to a maximally pluralist realism; again, see previous entries.
** PK has probably written about the generality problem somewhere, but I can't recall what he says. Bishop and Trout claim that their strategic reliabilism completely avoids the generality problem, but I am not entirely clear on why they think that. Their approach still requires recognizing particular beliefs as the outcome of strategy types.
Mon 14 Apr 2008 10:49 AM