I’m a contextualist about ‘knows’. I think that George, speaking in a non-skeptical context, can truly utter, “Jonathan knows that Cheney is the Vice President,” even as Peter, speaking in a skeptical context, can truly utter, “Jonathan doesn’t know that Cheney is the Vice President”. (Peter may go on to clarify: “For all Jonathan knows, Bush died ten minutes ago and Cheney is the President.”)
Here’s a powerful challenge to this view. (This is pretty much the challenge given by Timothy Williamson in “Knowledge, Context, and the Agent’s Point of View”.) Peter, who says “Jonathan doesn’t know that Cheney is the Vice President”, ought to criticize me if I believe that Cheney is the Vice President, since it’s bad to believe what you don’t know. (Imagine Peter saying, “Jonathan doesn’t know it, but he is right to believe it.” That’s crazy. He says rather, “Jonathan ought not to believe that Cheney is the Vice President.”)
George, by contrast, will endorse my belief: “Jonathan is right to believe that Cheney is Vice President — after all, he knows that Cheney is Vice President.” So George and Peter are not only apparently disagreeing about whether I know; they’re apparently disagreeing about what I ought to do, with respect to belief.
According to contextualism, neither is incorrect about whether I know. If, as seems plausible, their reasoning is cogent, neither then is either incorrect about what I ought to believe. But this is crazy: what one ought to do is not only true relative to some circumstance of evaluation.
I’m inclined to the following somewhat radical — but maybe, on reflection, independently plausible — solution: contextualism about ‘belief.’ Suppose that I am responding properly to my evidence. I rule out, on the basis of memory, the possibility that Bush selected somebody other than Cheney for VP. I do not rule out on any basis, however, the possibility that Bush died five minutes ago. (How would I have known?) I am, let us suppose, ignoring this possibility, as is George. And we’ll stipulate that it’s sufficiently unlikely such that we can, in Lewis’s sense, ‘properly’ ignore it. So George can truly say, while properly ignoring Bush’s recent possible death, “Jonathan knows that Cheney is Vice President,” and also, “Jonathan believes that Cheney is Vice President.” I’ve ruled out (psychologically and factively) all the possibilities where Cheney is not Vice President, except the ones we’re properly ignoring. George praises me for believing what I know.
What about Peter? Peter is not ignoring the possibility that Bush died five minutes ago. Since my evidence doesn’t eliminate the possibility that he did, Peter truly says, “Jonathan doesn’t know that Cheney is VP.” But should he therefore criticize my belief? Only if he also truly says, “Jonathan believes that Cheney is VP.” But I don’t even pretend to have ruled out the case where Bush died five minutes ago. I was just ignoring that case. I never pretended to know it not to obtain. (Rough-and-ready test: if you asked me whether Bush had died five minutes ago, I’d reply that I had no idea.) So here’s my claim: it is false, relative to the skeptical context, that “Jonathan believes that Cheney is VP.”
Call the “relevant domain” the context-variant set of all possibilities that are not properly ignored. A subject “rejects” a possibility by psychologically ruling it out. (If a subject is misled, he may reject the truth.) An utterance of the form “S believes that p”, then, could have the truth conditions, S rejects all the not-p cases in the relevant domain.
I find this view appealing for a number of reasons, not least that it allows us to avoid the Williamson objection. However, I haven’t yet thought through what other consequences it might have downstream. Maybe it’ll cause trouble elsewhere. Any ideas?