Eavesdroppers and epistemic modals

Sometimes, we make claims of the form: "A might be F." For example, "Keller might be taking a nap," or "Zane might be guilty." Such claims are called epistemic modals. What makes them true, when they are, seems to be an epistemic matter. That is: what makes a claim of the form "A might be F" true is whether A's being F is epistemically possible relative to some body of knowledge. But whose body of knowledge? This is where things get tricky. Some contextualists claim that the relevant body of knowledge will always at least include the speaker's body of knowledge. A rather flatfooted version of contextualism (which MacFarlane 2011: 146) calls "solipsistic contextualism")) says the speaker's body of knowledge is all that matters. And so when Clint says, "Keller might be taking a nap," the proposition expressed by his utterance is true (on solipsistic contextualism) just in case Keller's taking a nap isn't ruled out by what Clint knows. This view might sound plausible at first blush. There are however several problems. Perhaps, for example, the relevant body of knowledge should be extended to include not only what Clint knows but also what others involved in the conversation know. But suppose it is. The problem I want to consider is one that appears to infect any form of contextualism about epistemic modals. I'll use John Hawthorne's (2007) example of an eavesdropper case here:

EAVESDROPPER: Someone is on the way to the grocery store. I hear her say: 'Susan might be at the store. I could run to her.' No party to the conversation that I am listening in on knows that Susan is on vacation. But I know that she is. Despite the fact that it is compatible with what the conversants know that Susan is in the store and that the speaker will run into her, I am inclined to judge the speaker's modal judgements to be incorrect. (Hawthorne 2007: 92)

For the contextualist, this is awkward. It seems that the person who utters 'Susan might be at the store" has said something true. After all, Susan's being at the store is entirely compatible with the speaker's body of knowledge (and the body of knowledge possessed by those in the conversation I'm overhearing). However, having overheard this conversation, I am inclined to judge the speaker's claim to be incorrect. 

But the picture that emerges is not a friendly one for the contextualist. It looks as though we have a case of disagreement: I reject what the speaker I overhear accepts. It also looks like neither the person I overhear nor I myself am mistaken: "Susan might be at the store" thus seems true relative to the body of knowledge operative in the conversation I overhear and false relative to what I know. But this sort of 'faultless disagreement' implies a sort of truth-relativism about epistemic modals. In order to avoid this result, the contextualist has got to explain away why the eavesdropper case only appears to be a disagreement (or at least a case where my denying what the speaker I overhear asserts is felicitious) when it's actually not, or alternatively, why the case is one that only appears to be faultless, when in fact either I or the person I overhear believes something false. 

No matter how the contextualist tries to get out of this puzzle that appears to motivate truth-relativism about epistemic modals, one thing is clear: the contextualist is going to have a hard time explaining just whose body of knowledge is supposed to be the relevant one in this context.

Reference:

Hawthorne, John (2007). "Eavesdroppers and Epistemic Modals," in Philosophical Issues, 17, The Metaphysics of Epistemology, 2007. 

MacFarlane, John (2011). "Epistemic Modals are Context Sensitive," in Epistemic Modality, eds. Egan, A. & Weatherson, B. (Oxford: OUP).