Saturday, March 15, 2008

'Ought' conversationally implies 'someone knows something or other about something'

I had a thought the other night about how the link between 'ought' and various knowledge ascriptions might involve conversational implicatures rather than entailment relations. I then discovered that Walter Sinnott-Armstrong basically worked out the view I was kicking around. On his view, when we use 'ought' in an advisory context and say 'S ought to X' there's a CI to the effect that S can follow the advice we've just given. I'm reading a paper by Bart Streumer criticizing Sinnott-Armstrong's view. I'm hoping that the work on the 'ought'/'can' relationship might be useful for thinking about the 'ought'/'knows' relationship as I'm trying to reconcile the view that ignorance is an excusing condition rather than, say, an enabling/disabling condition with certain observations about the use of 'ought' in cases where we're forced to come to a judgment knowing we do not know certain seemingly relevant facts about the consequences of pursuing one of the available options (e.g., Prichard's stop sign, Parfit's mine shaft).

An example. Jones' home has been burgled and Jones has been tied to a chair. Smith arrives. Jones does not answer the door because he is tied to the chair. Smith says:
(1) If you've been burgled, you ought to get out of your chair and call the police.

It seems natural for Jones to respond:
(2) I'm tied to the chair, so I cannot call the police.

The response seems perfectly apt. The question is whether it shows (1) to be false or not? Sinnott-Armstrong says it doesn't show (1) to be false. Rather, in asserting (1), one implies that the advice offered can be followed and (2) speaks to that implication. Streumer doesn't buy this response. He asks us to consider two further claims. Smith says:

(3) I know you cannot call the police, but that does not make it any less true that you ought to call the police.

Jones replies:
(4) Don't be ridiculous. It cannot be true that I ought to call the police if I cannot do so.

He thinks that Jones' reply is a natural one and that this cannot be so if 'ought' CI's 'can'. He writes, "if 'ought' conversationally implicated 'can', Smith would be cancelling the implicature when he says that he knows that Jones cannot call the police, and it would then not be natural for Jones to reply that Smith's claim cannot be true".

The thing is that I simply don't think (4) is natural. It doesn't strike me as an obviously appropriate thing to say and (3) strikes me as perfectly coherent.

Here's an example that I think suggests that Sinnott-Armstrong is winning thus far. Suppose Lanie is doing some home repair and really dislikes some lamps. I tell her that the ugly lamps ought to be left alone because the typical home buyer likes ugly lamps. It's all but analytic that most people have terrible taste. I say:

(5) You really ought to leave those lamps alone.

She insists on painting them, in which case I'd say:

(6) In that case you ought to paint them silver.

Now, consider two views:
(I) 'Ought' conversationally implies 'Will'
(II) 'Ought' entails 'Will'

A defender of (II) might say that when we know someone will not pursue some course of action, we say that they ought to pursue some other course of action and avoid other options. That, they say, is evidence that 'Ought' entails 'Will'. It's essentially the same evidence we get for 'Ought' entails 'Can' when we observe that upon learning that Jones is tied to a chair we take back the claim 'You ought to call the cops' and say instead 'You ought to have screamed for your neighbors to help untie you'. However, there is no one who thinks that (II) is true and in explaining the naturalness of (6) and taking back (5) will appeal to something along the lines of (I). I think this example shows that Sinnott-Armstrong was right to remind us that the fact that an agent ought to do one thing does not show that they should not do other things. Lanie really ought to leave those lamps alone, but since she will not, she really ought to paint them silver. Jones really ought to call the cops, but the reason the crooks tied him to his chair was to prevent him from doing what he ought to! That does not change the fact that he should have shouted for help since he could not untie himself.


Bart Streumer said...

Interesting point. I think you're right that there is a parallel here, and perhaps it's true that 'ought' conversationally implies 'not will not'. (It should be put in terms of 'not will not', I think, rather than 'will' -- surely, when you tell someone that he ought to do something, you're not conversationally implying that he will do it, but at most that it's not the case that he won't do it.)

But I think there's a clear difference between these cases as well. When you advise Lanie to paint the lamps silver, you can coherently continue to think that she really ought not to do this and leave the lamps alone instead. You are, as it were, bracketing what you think she really ought to do; you're giving her advice by holding fixed the things you know she's already decided to do, and by thinking about what she ought to do given that those things are held fixed. After all, if Lanie started to doubt her decision to paint the lamps, she could ask you: 'What do you think I really ought to do?', and you could perfectly coherently answer: 'What I think you really ought to do is leave those lamps alone, and that's what I thought all along.'

I think the Jones/Smith case is different in this respect. When Smith learns that Jones is tied to his chair, I don't think he would normally continue to think that what Jones really ought to do is call the police. He would instead start to think that what Jones really ought to do is something else -- something he can do, such as not panic and wait for help to arrive. After all, suppose that Smith has learned that Jones is tied to his chair, and suppose that, after Smith has told Jones not to panic and to wait for help to arrive, Jones asks: 'OK, but what do you think I really ought to do?'. It would surely be weird for Smith to answer: 'Well, I think that what you really ought to do is call the police, and that's what I thought all along'. At any rate, that sounds very strange to me.

There's a more general point here as well. If we're not going to let 'ought' be constrained by what the relevant agent can do, where are we going to stop? Can Smith coherently think that Jones really ought to single-handedly break the ropes, even if it's perfectly clear to Smith that Jones cannot do this? And if Smith can coherently think this, can he also coherently (and without being deluded about the facts) think that Jones really ought to turn into Superman, break the ropes, fly out of his house and single-handedly catch the burglars? It seems clear that Smith can't coherently think such things unless he's seriously deluded about the facts. But if 'ought' is not constrained by what the relevant agent can do (in other words, if 'ought' does not entail 'can'), why can't Smith coherently think such things -- and, more importantly, why can't such things be true?

Clayton said...


Thanks for your remarks, I'm more than a bit flattered that you've dropped by.

I think I share your worry that opponents of OIC have to explain why we cannot say that Jones ought to bust the ropes, fly down the street, and apprehend the men himself. I have no idea how one would go about drawing the line between the not entirely crazy things we'd say (i.e., you should have called the cops, but since you couldn't, you should have screamed for help at the very least) and the entirely crazy things no one says (i.e., you should have frozen time, tied their shoelaces together, carried a cop down from the precinct, and then started time again and watch as the hilarity ensued). So, I think I have some sympathy for OIC even if I'm not entirely confident in the arguments for it.

Anyway, I'm still thinking about this stuff and I've just downloaded a copy of your paper on reasons and impossibility. Hopefully I'll get a chance to read it this week.

I found the bit about bracketing really helpful and if I ever write something up on this general topic, I'll have to include it and credit you. There's one thing that still bothers me. In your exchange with Sinnott-Armstrong, I never saw you say what you thought about assertions like:

(*) You really should have X''d, but since you couldn't, I suppose you should have Y'd at least.

I take it on your view these always express contradictions, but I have to confess they don't strike me as contradictory. Do we just have a difference in linguistic intuition here? Is there a story you can tell as to how these could be natural sounding but false?

Bart said...

Hi Clayton,

Sorry for this late reply. I agree with you that the linguistic data are not going to decide this issue on their own. The arguments I give in 'Reasons and Impossibility' are to a large extent independent of the linguistic data (I think).

One reason why the linguistic data are not going to decide this on their own is that many people think there are different sense of 'ought': the 'ought' of obligation, the 'ought' associated with having most reason, the 'ought' of expectation and the merely evaluative 'ought' (or 'ought to be'). Most defenders of OIC would say that only the first two senses of 'ought' entail 'can'. And defenders of OIC can appeal to senses of 'ought' that do not entail 'can' to explain away linguistic data that, at first sight, do not seem to fit with OIC.

So take your example:

(1) You really ought to have X'd, but since you couldn't, I suppose you ought to have Y'd at least.

And compare (1) to:

(2) You ought to X, but since you can't, you ought to Y.

Though (1) may not be contradictory, I think (2) does sound contradictory. And I think this shows that the reason why it is possible to hear (1) as non-contradictory is the addition of 'really', 'I suppose', 'at least', and perhaps also the past tense. This seems to signal that the first 'ought' in (1) is a different sense of 'ought', namely, the merely evaluative sense that does not entail 'can'.

So I'm prepared to admit that it is possible to hear (1) as non-contradictory, but I think that hearing (1) in this way requires interpreting the first 'ought' as merely evaluative: it merely means that it would have been better if you X'd, or perhaps that, in a better world in which you were free from certain constraints, you ought (in the reasons or obligation sense) to have X'd.