Brian has a nice little post on a paper written by MacFarlane and Kolodny on 'ought'. He favors a kind of contextualism on which the truth-conditions we assign to a speaker's assertive utterance of 'S ought to X' in a context is going to capture the idea that:
KD: What you oughta do is relative to what is known.
Whose knowledge do we make reference to in KD? That's where the trickiness comes in. Suppose some miner's went down into mine A when it starts to rain. They radio up to the ground and say:
(1) You ought to close off shaft A.
Doing so would ensure that the water all goes down into shaft B where the miners know that no one is working. Doing otherwise will lead to disaster. Meanwhile on the ground, the guy at the switchboard doesn't know where the miners are. It's not his fault. The radio transmission didn't make it out of the mine. The board that is supposed to tell him where the miners are has been stolen. He knows they have to be in mine A or mine B without knowing which. He can choose to close off A entirely or B entirely, but in doing so all of the water would wash into the other mine shaft killing any miner that happens to be working below. He can press a C button that will close off most of the opening and that should allow most of the miners to survive. Unfortunately, he knows enough of the water will wash down the shaft to drown one of the ten that are in the mines working.
There's a strong intuition that the guy ought to cut his losses. If the guy at the switchboard thinks it over, he'd think:
(2) I ought to press the C button rather than close off A or B completely.
While that seems true (or, not obviously incorrect), we have a problem that it seems some version of contextualism is well-suited to solve. We think (1) is true because relative to what is known to the miners that seems optimal. We think that (2) is true because relative to what the guy at the switches knows, pressing the C button will maximize expected utility.
Here's what bothers me. I haven't been reading enough of the literature on contextualism. I thought (but could be mistaken) that the standard contextualist attitude towards, say, 'knows' is something like this. There will be certain claims about the connection between, say, knowledge and the elimination of counterpossibilities that will be true in every context. So, in every context it will be true that knowledge that p requires eliminating all possibilities in which ~p. What varies from one context to the next is _not_ the connection between knowledge and counterpossibility elimination. What varies are the counterpossibilities that need elimination. In normal contexts, the possibility that I'm demonically deceived into thinking I have hands will not be a possibility I'll need to eliminate. In sceptical contexts, I'll have to eliminate that possibility.
If this is right, it suggests that while we might have variation from one context to the next in terms of whose knowledge matters (the miners' knowledge? God's knowledge? the switchboard operator's knowledge?), there will not be contexts in which KD is false. The link between doing what you oughta and knowing will remain constant. Yet, that seems false. I think there are situations in which we use 'ought' where the truth of knowledge ascriptions doesn't matter. If I'm right about that and right that KD will be one of the fixed points in a contextualist account of 'ought', the contextualist account is in trouble.
* Imperfect self-defense. In cases of imperfect self-defense, one uses force on another in the belief that such force is necessary for self-defense when the party you harm is not an aggressor (e.g., mistaking a jogger with a flashlight for a mugger with a club and macing the guy).
Cases of imperfect self-defense are often offered as cases in which the agent's actions were all things considered wrong, but excusable. The agent's ignorance is an excuse. I don't see any straightforward way of accommodating the claim that ignorance often constitutes an excuse with KD since we typically think that:
DW: What you oughta do is avoid acting against undefeated reasons, excusably or not.
Now, maybe I'm just wrong to think that the contextualist cannot accommodate this. Maybe DW is true in some contexts and KD is true in others. That seems very much unlike the proposals I've seen put forward for 'ought' and seem structurally quite different than the accounts put forward for 'knowledge'. I think Goldman is right that there is a use of 'knows' on which 'S knows p' is acceptable if S believes truly that p. However, I wouldn't have thought that someone who took the relevant form of acceptability to involve true assertion to say that we ought to say that in only some contexts knowledge requires evidence, justification, the absence of Gettiering conditions.
Maybe we have a case of ambiguity. Maybe there's an 'ought' that the contextualists are right about and an 'ought' that the invariantists are right about. I really don't like that idea for reasons Thomson adduces in Goodness and Advice. (Note to self: I should really rethink those reasons since that's the book that messed me up in the first place.)
I think we're not making much progress on this problem.