In an earlier post, I quoted the following passage in which Williamson defends anti-contextualism about the concept of what is right or wrong:
In cases of decision-making, one context is distinguished above all others: that of the agent at the moment of action. The primary question is whether the sentence ‘It would be wrong for me not to resign’ expresses a truth as uttered in the context in which the speaker is Clare and the time is that for resigning if she is to do so at all. Call that context the agent’s context, and the proposition which the relevant first person present tense sentence (such as ‘It would be wrong for me not to resign’) expresses in it the agent’s proposition. If at some point in her agonizing Clare uses the sentence to express a proposition other than the agent’s proposition, which might fail to match the agent’s proposition in truth-value, she is no longer concentrating on the relevant practical problem. Similarly, an external commentator who uses the sentence ‘It would be wrong for Clare not to resign’ to express a proposition other than the agent’s proposition is no longer concentrating on the relevant practical problem for Clare. But if the sentence expresses the agent’s proposition in all the contexts at issue, then it expresses the same proposition in all those contexts, and contextualism fails for this case.
This passage employs the concept of what is distinguished, which context robs others of the autonomy needed for contextualism to be true about right and wrong. Williamson characterizes this distinguished context in various ways here: it is the “primary question”, it is the context that delineates the “relevant practical problem”. It is easy to gloss these three different characterizations in value terms: what really matters, given the particular situation Claire finds herself in, is which agent proposition is true; nothing else could be important enough to override the importance of this question for Claire, and anyone else should recognize this fact. So contextualism about the terms of this problem–e.g., the concepts of right and wrong–is false.
Notice what happens in epistemology if we accept such an argument. It doesn’t follow that contextualism about knowledge is false, but something very interesting does follow. Underneath every version of contextualism should be something akin to the question Claire faces. There will be some distinguished epistemic problem which defines the primary epistemic question for the person involved, and this issue will make the agent’s context distinguished in such a way that the agent’s context robs other contexts of the autonomy needed for contextualism to be true about this particular epistemic concept. Call this concept the noncontextualist concept (I’m being sloppy metaphysically here, since the issue isn’t really one about concepts at all; but it’s easiest to talk this way, so I’ll leave it for now).
Williamson’s view is that the noncontextualist concept is knowledge itself, fitting with his overall project of making knowledge itself the focus of his theorizing, both in epistemology and in the theory of action and mind, more generally. Thus, he undertakes the burden of showing that knowledge is analogous to right and wrong in the Claire example. What’s interesting from my own point of view, however, is the more general point that contextualism is not a deep epistemological theory, if Williamson’s argument is correct. It is not deep because any concept of which it is true is not the concept that really matters from an agent’s epistemic point of view. Once we identify what really matters epistemically, Williamson’s argument implies that contextualism won’t be adequate regarding it, and that makes Williamson’s argument quite important.
The only alternative for the contextualist here is to make it into an onion theory, on which no matter how many layers you remove, you find more contextualism at the next layer. The implication of Williamson’s argument is that on such an onion theory, nothing really matters epistemically. It’s the epistemological equivalent of a person with no soul, who is nothing but one façade layered on top of another.
So it is no surprise that Williamson won’t embrace contextualism about knowledge: knowledge, for him, is the distinguished epistemic concept. For others, some other property will be distinguished in this way: for Cohen, for example, it is something like justification. What contextualists ought always to tell us to complete their epistemic theory is where the layers end: what is the reallly important epistemic stuff beneath the outer layers. That is what value-driven epistemology demands.