I argue is that philosophers have had a tendency to present a kind of mystical view of the powers of reflection. Unreflective belief acquisition is seen in mechanistic terms, but when philosophers talk about reflection, it is as if reflective processes are not bound by the kinds of limitations which inevitably arise from being embedded within the same causal structure which governs unreflective belief acquisition.

My own reasons for favouring talk of natural kinds is just that I believe the best accounts of the success of scientific theories presupposes the existence of natural kinds.

I do think that it is legitimate to talk of goals and functions in nature, and that these things can be made sense of in naturalistic terms. There is nothing at all contrary to naturalism in the idea of goal-directed systems.

I believe, that empirically informed approaches to the question have issued in more illuminating answers than the old armchair approaches. But I think that it would be a terrible mistake to give up on addressing normative questions in epistemology.

Externalists reject any such view. I think that the idea that we can tell, simply by way of reflection, whether our beliefs are justified, is deeply commonsensical. More than that, the idea that responsible epistemic agents ought to reflect on their beliefs, and hold them only if they somehow pass muster, is utterly natural.

Here, as in so many other cases, however, it turns out that a very commonsensical idea looks far less attractive when one examines some of the experimental work which is not available to us from the armchair.

I do think that an understanding of contemporary work in the cognitive sciences has a profound effect on how one views the workings of the mind. It doesn't work the way we pretheoretically think it does. Such an understanding, of course, should have a large effect on one's views in philosophy of mind, but also in epistemology.