June 20, 2004

Shifting Standards

If I wanted to make a deep point out of this post it would be as follows:

Grant me, for the moment, that the tense structures of our language embody at least a mild presupposition that the time at which an utterance is uttered is the same as the time at which it is heard. (Nobody seems willing to grant me that, which may or may not undermine the ensuing argument.)

Then suppose that on different occasions we may need more or less supporting evidence in order to be justified in acting on a certain belief. These are the familiar airport and bank cases--the worse the consequeneces of false belief, the better your evidence needs to be in order to act.

Now also suppose that our language embodies an assumption that the stakes are pretty much the same for everyone involved and under discussion in a certain conversation. This doesn't seem completely outlandish--but anyway, just grant me that.

Then, if "know" entails "have good enough evidence for, given the stakes," we would expect "know" to behave kind of oddly in cases in which the stakes aren't the same for everyone involved. So in the airport case, when you ask, "Does that guy know whether the plane stops in Chicago?" you are effectively asking (among other things) "Does that guy have good enough evidence for believing that the plane stops in Chicago that we are justified in acting on that belief, given what we have at stake"--where "we" includes me, you, and him. And if the stakes are different for him than they are for me and you, we will be pulled different ways--both toward affirming that he knows (given his stakes, his evidence is good enough) and denying it (given our stakes, his evidence isn't good enough).

And I think that's why the debate between contextualists and invariantists concerning these cases gets hairy. The word "know" is being put under more strain than it was designed for--being pulled in two directions at once.

(That might be more convincing if my original example worked better! For a more respectable model, look at Anil Gupta's discussion of people who think "up" is both a direction in absolute space and a normal to the earth's surface: "Meaning and Misconceptions," in Language, Logic, and Concepts, ed. by Jackendoff, Bloom, and Wynn, MIT Press, 1999.)

Posted by Matt Weiner at June 20, 2004 01:59 PM

Interesting issue. First, a small point. Why do you need the controversial claim about the semantics of tense? Isn't it enough that in the cases you want to consider it is true that the time of utterance and the time of reception are roughly simultaneous?

Second. Although I agree that the debate in here gets hairy, I am not sure that this has a deep philosophical explantion, viz., that the concept of knowledge is being put under more strain than it can handle. [Note: You put this point in linguistic terms and I have deliberately changed it to make the conceptual/linguistic contrast apparent.]

It seems to me, rather, that the explantion may just be a shallow matter of communicative clarity. Specifically, if we are contextualists and we don't adhere to a principle of (partial) semantic blindness, then the tension in our knowledge attributions may just arise from a run-of-the-mill sort of tension in making multiple uses of a polysemous word in a single context clear.

Posted by: marc at June 21, 2004 09:43 AM

"Roughly simultaneous" will do fine. The controversy comes over my claim that the language can get put under some strain when the time of writing and the intended time of reading aren't even roughly simultaneous. It's the strain that's controversial.

I think I pretty much agree with what you're saying about polysemy, if I'm understanding your point. My claim is that the reason "know" doesn't look polysemous is that it's usually OK to treat it as though the meaning is constant throughout a usage. (I'd like to be able to rephrase that in a way that's equally palatable to sensitive invariantists, but I'm not sure how right now.) When the people involved in the conversation have different stakes, full explicitness requires making the shifts in standards clear by operators like "by these standards" (Ludlow has a nice list of such operators); if we don't use those, we get anomalous formulations like the ones contextualists and invariantists want to argue over.

My wacky view is that the concept of knowledge goes no deeper than these problems do. That is, I think that the concept of knowledge itself is useful primarily as a clear and economical way of summing up several desiderata (truth, appropriate levels of justification, etc.) So there's nothing more to the concept of knowledge than what's required for communicative clarity. I have a long rambling paper on this that I'm working on (specifically, I'm working on trying to break down some of it to submissible length).

Posted by: Matt Weiner at June 21, 2004 04:03 PM