Thanks to everyone for sharing their intuitions about the cases in my previous post. Here I’m going to back up and say something about my interest in the cases. Among linguists, the canonical view about modal expressions like “might”, “may” and “must” is that they are quantifiers over possibilities where the domains of quantification are contextually restricted. The view is an extremely powerful one; if correct, it provides a simple, highly unified explanation of a wide variety of language use. Recently, the canon’s neat story has come under attack on two fronts, in its treatment of bare epistemic modals (BEMs) and bare normative modals (BNMs). (A bare modal statement is a modal statement that doesn’t contain a restrictor phrase like “in view of my evidence” or “in view of what the law requires”.) I’ve got a manuscript defending a general, flexible contextualist account of bare modal statements and an application of that account to BEMs. (Here: http://www.unl.edu/philosop/people/faculty/dowell/dowell.shtml) Now I’m working on defending an application of that account to BNMs. (NB: My apologies in advance; this post is both long and oversimplified.)
The puzzle cases for contextualism about BNMs focus on apparent disagreement across contexts. On the canonical contextualist account, modal statements are doubly relative; they have two parameters that context must provide a value for. On the view I’m defending, one of those parameters is set either by a contextually determined body of information or set of circumstances, the other by contextually determined standard.
Here's the basic puzzle case for apparent disagreement across contexts shifting the value for the information parameter:
Imagine a doctor deliberating about which of three drugs, X, Y, or Z to prescribe a patient to relieve the symptoms of her skin irritation. Doctor’s limited information suggests that X and Y each will provide complete relief, while drug Z will provide only partial relief. In light of this, she asserts:
Imagine a doctor deliberating about which of three drugs, X, Y, or Z to prescribe a patient to relieve the symptoms of her skin irritation. Doctor’s limited information suggests that X and Y each will provide complete relief, while drug Z will provide only partial relief. In light of this, she asserts:
Doctor: (X) “I ought to prescribe either X or Y.”
Suppose a consulting physician has more information about the patient’s medical history and each of the drugs than Doctor does. In particular, Consultant knows that while it is true that, absent any interfering drug already in the patient’s system, either X or Y would provide complete relief, it is also true that Patient is already taking drug W which, together with one of either X or Y, is certain to kill the patient. (Consultant knows that, given that the patient is taking W, exactly one of X or Y will kill her and the other completely cure her, but she does not know which drug would have which effect; each is equally likely to be the lethal drug.) Z, in contrast is certain to provide some relief and certain not to have any negative side effects. Given this, Consultant replies
Consultant: (Z) “No, it is not the case that you ought to prescribe either X or Y; you ought to prescribe drug Z”.
We can sum up the expected outcomes, given the information of the Doctor and Consultant, respectively, this way:
Drug X Drug Y Drug Z
Doctor: 1.x complete cure 1.x complete cure 1.x partial relief
Consultant: .5 x Death + .5 x complete cure .5 x Death + .5 x complete cure 1.x partial relief
The question is: How can a contextualist capture the sense we have that Doctor and Consultant are giving different and incompatible answers to a common question? If each of (X) and (Z) is relativized to the speaker’s information, then Doctor and Consultant are not disagreeing and Consultant’s assertion is not an answer to the same question as Doctor’s. But since Consultant’s “no” is felicitous, Doctor and Consultant do seem to be disagreeing. Moreover, Consultant seems to be advising Doctor by aiming to give a superior answer to the very question Doctor aims to answer with (X).
The puzzle rests on an assumption about which value a contextualist account holds that context selects, namely, that context selects for each of (X) and (Z) the information the speaker possesses at the time of utterance. If that’s right, then Doctor and Consultant are neither disagreeing, nor addressing a common question.
On the flexible, contextualist account I favor, which value for each of the parameters is contextually selected is determined partly by a speaker’s extension-determining intentions, i.e. publicly manifestable intentions to let some feature of the context settle the restrictions. I understand the intentions here dispositionally, as a disposition to recognize some feature of the context as restriction-settling.
On the flexible, contextualist account I favor, which value for each of the parameters is contextually selected is determined partly by a speaker’s extension-determining intentions, i.e. publicly manifestable intentions to let some feature of the context settle the restrictions. I understand the intentions here dispositionally, as a disposition to recognize some feature of the context as restriction-settling.
The beginning of wisdom then is to note exactly what is puzzling about the case, given the presumed contextualist treatment. Let t3 be (roughly) the latest time at which Doctor can effectively prescribe a drug. Let t1 be the time at which Doctor asserts (X) and t2 the time Consultant asserts (Z). On the contextualist treatment presumed in the puzzle’s statement, a speaker is asking herself “which act is best performed at t3 given what I know now?” Interest in that question would be surprising, because Doctor finds herself in a context of deliberation and, as Finlay and Bjornsson note in their manuscript on this topic, deliberators tend to be news-sensitive. I think of news-sensitivity dispositionally, as a readiness to revise one’s judgment about what one ought to do in light of new information. Given that deliberators, at least when they are rational, are typically news-sensitive, it would be odd if Doctor were interested in which act would be best at t3, given what she knows at t1, only to abandon her interest in that question, should new information come to light prior to t3. If Doctor is news-sensitive, it is better to understand her as intending to speak to (something like) the question “what ought I to do at t3 given the information I’ll have then?”
On my account, in the puzzle case, Doctor and Consultant are disagreeing about how to answer that very question. Let’s assume that Doctor has at t1 a reasonable opinion about which body of information that that is, namely, that it’s the information she possesses then. Then (X) is warranted, though false.
On my account, in the puzzle case, Doctor and Consultant are disagreeing about how to answer that very question. Let’s assume that Doctor has at t1 a reasonable opinion about which body of information that that is, namely, that it’s the information she possesses then. Then (X) is warranted, though false.
This analysis generates a prediction about cases of hindsight evaluation that I wanted to test with the first of my two cases in my previous post. The prediction in that case is that we won’t think of better-informed, hindsight evaluators as warranted in disagreeing with deliberators who did the best they could with the information available at the time of action. And that prediction was rather overwhelmingly borne out by the reactions that I got from Pea Soup commentators and others (some philosophers, some careful ordinary speakers). A strong majority found either Sp to be the best response or tied with Gp. Among the remaining group, there was a majority preference for Gp or the view that all except Dp were acceptable. No one ranked Dp or Fp as better than both Sp and Gp. An even larger majority found Dp unacceptable and, most of those found Fp also unacceptable.
This is just what we should expect on the account I defend. “I disagree” clearly expresses an attempt to reject the whole modal claim in (Z) and (Z’). “That’s false” is less clear, but can be heard in the same way. “That’s a shame”, in contrast, is an expression of regret that may signal a context-shift, as it does in Gp. Sp explicitly expresses agreement with the whole modalized claim in (Z) and (Z’), as is warranted on my account. Interestingly, many who stuck up for either Dp or Fp did so by insisting that they are semantically appropriate, though either silly or false. This is also the right thing to say on my account, which distinguishes between norms of assertion that govern linguistic appropriateness and those that govern warrant. Added bonus: solipsistic relativism predicts that Dp is not only appropriate, but warranted and true. So, it conflicts with this response pattern.
This is just what we should expect on the account I defend. “I disagree” clearly expresses an attempt to reject the whole modal claim in (Z) and (Z’). “That’s false” is less clear, but can be heard in the same way. “That’s a shame”, in contrast, is an expression of regret that may signal a context-shift, as it does in Gp. Sp explicitly expresses agreement with the whole modalized claim in (Z) and (Z’), as is warranted on my account. Interestingly, many who stuck up for either Dp or Fp did so by insisting that they are semantically appropriate, though either silly or false. This is also the right thing to say on my account, which distinguishes between norms of assertion that govern linguistic appropriateness and those that govern warrant. Added bonus: solipsistic relativism predicts that Dp is not only appropriate, but warranted and true. So, it conflicts with this response pattern.
The Pea Soup comments on my second case helped me see that my characterization of the case needs improvement. The case that I described was both too vague and too friendly to the contextualist. Thanks to everyone for their help with both cases.
Hi Janice,
As you know, Gunnar and I are generally friendly to this line of defense of contextualism. A couple of problems, though, that motivate us to look for a complementary solution in our paper:
1. Doctor’s interest can’t strictly be in what she ought to do given the information she’ll have at t3, because this interest wouldn’t give her any reason to seek additional information, or any reason not to deliberately avoid acquiring additional information. Hence we talk (vaguely) about an interest in what she ought to do given the best information she can acquire by t3 (which we recognize is also not quite right).
2. On either account of the motivating interest of deliberation, there’s another problem. You echo Kolodny & MacFarlane, writing, ‘Consultant seems to be advising Doctor by aiming to give a superior answer to the very question Doctor aims to answer with (X)’. But of course, if Consultant chose not to make his information available, then it is neither information that Doctor will have at t3, nor is it information (let’s suppose) that she can acquire by t3. The result of this is that Consultant’s answer is only true, and only superior to Doctor’s answer, IF Consultant chooses to make his information available. (If Consultant had minded his own business, then Doctor would have got the answer to her question right by herself!) Gunnar and I think that this is seriously problematic for the claim that better-informed advisers are trying to give better answers to the same contextualist ought-questions that deliberating agents ask themselves.
This leads us to think that the contextualist has to accept that appropriate advice does not always answer the same question that the deliberator asks. As we argue, our interest in the truth of any information-relative ought-proposition is generally derivative from more fundamental interests in the things we care about. In light of this, it seems to us very plausible that deliberators will abandon interest in some ought-propositions for others, as new information comes available.
Hi, Steve,
You’re just right that Doctor’s interest “can’t strictly be in what she ought to do given the info she’ll have at t3. That’s a place where my post was oversimplified for the sake of brevity. The fuller discussion is in my paper which I’ll try to summarize (but I’m also happy to send you my rough draft, just let me know if you’d like it).
Remember that at bottom what fixes whether a piece of information is in or out of the domain-restricting set is the speaker’s intentions AND that I’m understanding the speaker’s intentions dispositionally. So, the question to ask about some piece of information is whether the agent, in being made aware of it, would recognize it as information she intended to take into account.
I like your suggestion that deliberators are news-sensitive and I’m am cashing that out in my framework. On my view, news-sensitivity is a matter of having certain intentions which are dispositions to respond to information. So imagine that Doctor knows that there is a specialist on Drug W that can easily be reached before t3 and Specialist can tell her whether Drug X or Drug Y is the one that lethally combines with W. The most natural thing for Doctor to say in response to Consultant’s (Z) is
(W) “No! In that case we should consult Specialist. She will tell us whether we ought to prescribe X or whether we ought to prescribe Y.”
The reason why this is the most natural thing for Doctor to say, on my account, is that deliberators are news-sensitive in just the sense that I’ve described. Doctor recognizes Specialist’s information as just the kind of information she intends to take into account.
On your second point….I’m not getting the worry. Doctor is deliberating about what to do before a time after which acting is unless. So, she doesn’t intend to take into account information she is unable to acquire (I’m imaging you’re thinking of a case in which only Consultant knows that Patient is taking W.) If Consultant just stands there, keeping the information about W to himself, while Doctor is talking out loud about which drug she ought to prescribe and he thinks to himself “No, she ought to prescribe Z” then it seem right to me to say that they are addressing different questions. Doctor is deliberating about what she ought to do, given the information she intends to include, information it is within her powers to make available before t3, and Consultant is deliberating about what Doctor ought to do, given that Patient is taking W.
I think this is related to Steve’s comment. I’m wondering if the right aim is relativized to what she’ll actually have at T3, as opposed for instance information she should have by T3, or information she could easily have at T3. OTOH, it seems to me that some sort of contextualist story could be told to make these the facts relevant to the truth of the judgements in the stories told.
On a different note, on this account the judgement is doubly relative — to standards as well as facts. Is there an argument that the first sort of relativity is essential? If you’re already a relativist about normative standards, that is obviously going to be needed. But I guess I suspect that to defend this kind of relativity you’ll have to defend at least some mild relativism about appropriate standards of evaluation and that may be an issue that goes beyond just semantics, though the answer will effect the semantics.
Thanks, Mark! I was thinking that my answer to Steve’s first point above addresses your first point. Or maybe I’m missing something?
About your second point: I’m not sure I’m seeing what you have in mind, but I’ll say a bit more about what I have in mind and you can see whether it helps. On the canonical view, and on the view I’m defending, modals like “must” make a *single* contribution to determining a proposition in *all* contexts in which they appear. They’re all quantifiers over possibilities. Context is what is going to distinguish the different propositions generated by e.g. “Dave has to be in in office” by determining different domains of quantification for them. (There are at least two uses we might make of that sentence, an epistemic one, and a normative one.)
But there aren’t just the epistemic and normative uses that need to be given a single account. There are the different normative uses, e.g. legal and moral. What distinguishes these? The answer, on my view and Kratzer’s, is different standards, legal or moral. SO…and this is important for me….my account is NOT committed to what you might call “moral relativism”, the view that different moral standards are selected in different contexts. I myself don’t think such a hypothesis fits well with ordinary speakers’ understanding of what they’re saying when they make moral assertions. I think that ordinary speakers take themselves to be intending to appeal to an absolute moral standard, the content of which they have substantive opinions about, but not to intend for those opinions to restrict the domain of quantification for their modal. (I also discuss this in my paper in connection with a second puzzle that has been raised against contextualists by various people, including Weatherson, Finlay and Bjornsson, and Schroeder.)
Janice,
Just on the second point, since I want to think more about the first:
OK, I get it. I did consider that interpretation, where moral standards are one candidate, but that there are not multiple moral standards. But I thought you might mean something more, since I think some people with similar views do mean more.
I guess I think that what I said about relativism goes just as much for absolutism — whether you need to use context to fill in a specific moral standard or whether you can do without that depends on whether there is in fact only one such standard. I think that depends on something besides what speakers take themselves to be talking about. Suppose for instance that there were in fact differing eligible moral standards none of which was objectively privileged. Then I take it that even if the speakers all took themselves to be speaking about the one objectively correct set of standards, that would by itself be insufficient to make there judgements all relative to the same set of standards.
Still, I think a contextualist view somewhat like this one can leave that open, so that there is a room to use features of the context to pick out a particular set of moral standards if as it turns out there are several eligible candidates. So that’s not an objection so much as a comment. My thought really is that some things depend on moral/normative metaphysics.
Janice,
Now on the first issue. Can you say a bit more about the way in which the dispositons of the doctor differentiate her intentions? One way to think of it would be, here’s a fact that if she became aware of it she would want to take into account at time t3. But that seems to include in principle unobtainable facts. Another way would be to include only facts the doctor is now disposed to go find out. That’s what your example in answer to Steve suggests, though I’m not sure that is the intention. That seems too restricted, since doctors can be lazy and weak willed about getting the answers to questions they in some sense want to find out the answers to.
Janice,
So we’re basically in agreement about the first point. (Although I wouldn’t want to identify the speaker’s intention with a disposition). Notice, by the way, that there’s presumably more going on in your Specialist case than you let on. When Doctor says, “We should consult Specialist”, this ‘should’ is presumably qualified by something like ‘given the information available to us’.
On my second point: Gunnar and I take this to show that the concern motivating deliberation and advice can’t simply be to get the right answer to a specific information-relative ought question. For Consultant isn’t really doing anything to help Doctor answer such a question by making his information available. (It’s like ‘helping’ someone build a block tower by knocking down what they’ve done and then offering to help them put it back together). For example, suppose that without Consultant’s information, it is very easy for Doctor to determine what she ought to do, but that with Consultant’s information it becomes very hard to know what she ought to do, so that the likelihood of Doctor getting it right is lower if the additional information is provided. Then it seems that Consultant really does assist Doctor in reaching the right answer to her question by withholding his information, and not by providing it. The upshot is that it seems to be a mistake to think that advisors must simply be trying to help deliberators reach the right answer to the question they are deliberating over.
I can’t resist adding that I offer a unifying semantics for ‘ought’ of the kind that Janice favours in my paper ‘Oughts and Ends’, and like Janice I think that restriction of worlds by something like standards (in my case, by ends) is what generates the normative senses. But unlike Kratzer (and I think Janice), I try to accomplish this with just one argument-place. The resulting view endorses a weak form of relativism of the kind that Janice seeks to avoid.
Oh, cool–loads of comments. I’m about to eat, so I’ll only get to mark’s first post.
I agree that whether there is more than one moral standard depends upon more than what folks take themselves to be up to when they make more claims; it depends on how many correct standards there are, if there are any. BUT I do think that in the case of some ordinary expressions, the extensions of those expressions, IF there are any, depend partly on speaker’s intentions, (as well as what there is) and that this is true in the case of quantificational expressions (modals and quantifiers over individuals) and for demonstratives (which seems to be roughly Kaplan’s thought–his view is the inspiration for mine). So, I want to say that the domains for the modals used in ordinary, moral discourse is given by speakers’ intentions and speakers do intend to be picking out a single, absolute moral standard. (Again, this is just a claim I’m making here, arguments are off the table. I’m just trying to get my semantic proposal on the table for now.) So, if there is in fact more than one moral standard, then ordinary moral assertions involve presupposition failure and are neither true nor false. So, my view is compatible with a kind of error theory about the semantics; that’s as it should be, since it’s a semantic proposal, not a metaphysical one.
ok…off to dinner…more later….
yay! modals!
Hey, Mark,
I’m not sure what you mean by “unobtainable fact”. Maybe you mean something like, ‘in principle unknowable facts’? I don’t think that deliberators have intentions to take into account in principle unknowable facts….But probably you didn’t mean that.
My idea is that deliberators have intentions to take into account a certain body of information, may have a substantive view about which body of info that is, but in any case, that view is fallible and not determinative of which body of information the relevant body is. So, for example, consider unexpected sources of information. It’s unexpected, so not part of the information that, in the deliberator’s view, will be information that she means to take into account. But it may nonetheless be part of that information if, when learning of the information she says/thinks something like “oh! I was wrong! I ought not to prescribe X or Y! I ought to prescribe Z!’. Here, I’m suggesting, she recognizes the information about Patient’s taking of W as the kind of thing she meant to include in the relevant body of info. (I think (and argue in both my ‘ought’s and epistemic modals papers) that this is just what is going on in the case of contextually determined domain restriction in the case of quantifiers over individuals. This unitary account of the quantifiers is a nice feature of my view, I think.)
So, I guess the answer to the second option you offer is, no, it’s not info she’s disposed to find out since, as you say, deliberators can be lazy. The question is, when you in hindsight realize that some info was relevant, and you could have found it out, but were too lazy to you collect it, do you think to yourself “still, I did what I ought to have” or do you think “Doh! I screwed up!”. I think even the lazy among us tend to think the latter, if it really was laziness that explained our failure to have the info.
Hi Janice,
Actually I did mean at least weakly in principle unknowable. I brought these sorts of facts up to rule out one sort of disposition as relevant to determining the facts in question. If we don’t want to countenance unknowable at the time of acting facts as relevant it rules out one way of filling in the dispostions the proposal takes into account. For example, if the relevant disposition is one to take facts with some specific content into account when given access, then even facts unknowable at the time of acting might be counted as intended. An agent might have the disposition to take facts with that content into account when given access to them (even if they are inaccessible to the agent). I figured you would not want that so I figured that you would not want the dispositions to be those sorts of dispositions.
The other alternative candidate for relevant dispositions that I could think of had the problem you and I agree on for the reasons stated in your last paragraph (at 1:57). So I was looking for another option.
There’s a way of reading that last paragraph as offering a dispositon to consider relevant in hindsight as the answer to my question about which dispositions matter. But I’m not sure that is intended as a description of the relevant dispositions, as opposed to evidence that the second way of specifying the dispositions that constitute intentions is inadequate. I’m guessing you mean it as the latter. But then I’m not sure which dispositions are the right ones to constitute the intentions in question.
To put my background worry on the table: My suspicion is that you’d be better off not going with the actual dispositions of the agent to determine what the right sort of facts are. That’s partly because I think people can be disposed not to recognize stuff that is relevant given the content of what they are thinking and saying.
Steve,
On your last point first: You’re right; I think there are two argument places. I think that for the same reasons Kratzer enumerates in her seminal paper “Modality”. (You need them both to solve certain puzzles and to handle comparative locutions, such as “…is better than….” and “…is more likely than…”.) I also don’t think standards are always set by ends, but that’s another story.
About the advice issue: It seems to me that we’re not quite connecting, but I’m not sure I can put my finger on exactly where. It sounds to me as if you’re thinking that Consultant’s information isn’t already in the body that restricts the domain of Doctor’s original assertion until after Consultant provides it. But I’m denying that because I think the most natural understanding of Doctor’s intentions puts it in there before Consultant provides it. So, for example, you say “suppose that without Consultant’s information, it is very easy for Doctor to determine what she ought to do, but that with Consultant’s information it becomes very hard to know what she ought to do, so that the likelihood of Doctor getting it right is lower if the additional information is provided.” I don’t know how to understand that without the assumption that Consultant’s info isn’t already in the relevant body. If it is, the probability can’t shift by Consultant’s providing it–whatever it was, it’s the same before and after Consultant speaks.
Steve and Mark,
Let me say a bit more about why I say Consultant’s info is already in there and in particular about how I understand the relevant intentions, since this seems to be unclear (e.g. to Mark in his post). The inspiration for my view is just Kaplan’s account of denotation-determination for demonstratives. I should add that if you like Kripke/Putnam on names and natural kind terms, you’re already stuck with referential intentions and need a way of cashing them out. You also need them if you like Donnellan on referential uses of definite descriptions and you need intentions playing a role in determining what’s said and what’s implicated if you like Grice on anything. So, the components of my view are straight off the shelf, totally familiar, everyday items in the philosopher of language’s toolkit.
The idea that we formulate the contents of these intentions to ourselves explicitly before speaking is just a non-starter. (If we had the contents of our intentions that determine how we, non-experts, manage to talk about elms and beeches with “elm” and “beech” explicitly before us every time we use those words, “the Meaning of ‘Meaning'” would have been a power-snore.)
Given that, the natural thing to do is to cash out the contents of our referential intentions dispositionally. The content of the relevant disposition is not hard to spell out. To see how this is done, start with a totally natural case of domain restriction for the use of a quantifier over individuals. Suppose Sally asserts “Every student was at the meeting”. Feeling confused (or engaging in what passes for humor among philosophers), George askes “What?! Even those students not in residence?” An entirely natural thing for Sally to say here is “No; I meant every student in residence”. (The example is from von Fintel and Gillies.) It is just implausible to say that Sally explicitly formulated to herself the restrictor “in residence” (or any of the many other restrictors she intends) before speaking. But she is right to claim that it is part of ‘what she meant’ all along. George’s query helps her see this.
The relevant disposition in the case of deliberators is of precisely the same kind. The information that is in the restricting body is given by what the speaker would recognize as in there, NOT, as Mark suggests, whatever she would recognize as relevant. There may be relevant info that a speaker has practically no way of knowing before action becomes necessary, as in the dramatic reenactment case. We can see that we don’t think such info is in the base-restricting body from seeing what appropriate hindsight evaluations would be, e.g. “that’s a shame” not “I disagree”. The challenge here for anyone who would reject this natural story is to tell a more plausible one without at the same time rejecting much of philosophers’ and linguists’ best understanding of how language works.
That’s a quick, somewhat oversimplified summary of my view on the role referential intentions play and how I think they should be cashed out dispositionally. I’ve got a whole paper on understanding referential intentions as dispositions, called “Empirical Metaphysics”. (It’s a reply of sorts to the Jackson/Chalmers semantic program.) It’s in one of the BSPC Phil Studies volumes (and also on my webpage, link above). I’ve also got a discussion of this in my manuscript on epistemic modals, also on my webpage, for those interested in a more detailed formulation.
Janice, sorry I’ve been away from the discussion for a few days…
This discussion is fragmenting into a lot of different strands:
1. Actually I agree that ‘ought’, unlike most other deontic modals, has two argument places–but I think they take (a) a modal base and (b) a set of alternatives, not a standard/ ordering source. (That’s just because ‘ought’, unlike other modals, is essentially comparative). I don’t see your point about ‘better than’: we don’t have similar forms for modals (‘oughter than’, ‘muster than’, etc.) Regarding Kratzer’s other reasons for positing an ordering source: I don’t think they bear close scrutiny. Positing an ordering source doesn’t provide a satisfactory solution to the problems it’s designed to solve (ranked possibilities, counterfactuals), and there are alternative ways of solving these problems without an ordering source. That all needs argument of course–I offer some argument in my paper ‘What Ought Probably Means’, but there’s another paper here I mean to write.
2. I totally agree that the meaning of an ought-claim is determined by the intentions of the speaker, and I also have no problem with the kind of referential intentions you describe. The problem is exactly what the content of that referential intention is. The point does not depend upon the relevant information changing at the time that Consultant decides whether or not to make his information available, I think.
At t2, Consultant gets to decide whether or not his information will be available to Doctor. Does his decision make a difference to the relevant information-base for Consultant’s judgment? We need not say that the information-base changes at t2 depending on the decision. We can say instead that the information-base is fixed at t1 by whatever Consultant will actually decide at t2. Would you then agree to this:
If at t2 Consultant decides not to make his information available, then Doctor’s judgment was never sensitive to Consultant’s information. Then there’s nothing uncooperative (on the picture Gunnar and I are opposing) about Consultant’s decision.
If at t2 Consultant decides to make his information available, then Doctor’s judgment was always sensitive to Consultant’s information.
If Consultant realizes this, wouldn’t he be right to conclude that whether or not Doctor’s judgment was (always) sensitive to his information depends on whether or not he (now) decides to provide it? So in the scenario I offered, in which Doctor is more likely to reach the right answer if her question was not sensitive to Consultant’s information, wouldn’t Consultant make it the case that Doctor is more likely to reach the right answer to her question if he doesn’t offer his information (by making it the case that the question she asked at t1 was the easier question)? So he would help her have a right answer by making it the case that her question was the easier one?
If you agree with all that, then I think my objection stands. Otherwise, what do you disagree with?
If you reject the suggestion that whether the truth of Doctor’s claim depends on whether Consultant chooses to make his information available, then it seems to me that you’re abandoning the definition of the relevant information as information that is available to Doctor. But then how is the information defined? You say that ‘the information that is in the restricting body is given by what the speaker would recognize as in there’. But if we take that seriously, your contextualism is in big trouble. For example, surely Doctor would consider relevant any information about which of drugs X and Y would kill the patient and which would cure her. But if her judgment is sensitive to THAT information, then it could never be correct for her to judge that she ought to prescribe Z.
3. Finally, on intentions as dispositions: surely there’s a middle option between thinking that referential intentions reduce to dispositions, and thinking that we have to articulate the exact content to ourselves in language. Our referential intentions at t1, it seems to me, are surely a function of what we had in mind at t1, rather than a function of what we would do or say at times after t1. I guess I should read your paper!
Quick correction: I see I misinterpreted your response to Mark. You’ve already rejected the suggestion that what is in the informational base is determined by what the speaker would recognize as relevant. So I’m not sure how you’ll seek to avoid my original objection.
Hi, Steve,
Thanks for the further comments. I’ll order my comments on your comments by following your numbers.
1. About the comparatives: I’m not following your point about our not having a term “oughter than”. We both accept that modals are quantifiers over possibilities. Given that, we should both accept that they are like our other quantifiers, like “all” and “some” and “more than”. There is also no comparative “aller than”. “More than” is the comparative for quantifier over individuals. On the modals side, everyone who accepts that modals are quantifiers (even non-contextualists) agrees that “must” and “might” function like “all” and “some” respectively, but also accept that there are comparative modals like “is more likely than” or (using my example of a normative comparative) “is better than”.
About Kratzer’s arguments for ordering sources—they don’t depend on counterfactual cases; maybe you’re thinking of Lewis’ use of a Kratzer-like framework for counterfactuals? I also don’t know what puzzle you’re alluding to when you talk of a ‘problem’ about ‘ranked possibilities’. Ranking possibilities is her solution to the puzzles about practical inferences and the Samaritan Paradox and also is required to handle comparative locutions. If you see another way of doing all that without an ordering source, I’d be interested to hear it!
2. Cool! So, I’ve convinced you that referential intentions are the way to determine restrictions on modal bases. You don’t discuss what determines those restrictions in any of your papers on modals and, it seems to me, any account that doesn’t address that question is ad hoc. (My original arguments for my view are in my paper, “A Defense of Canonical Contextualism about Epistemic Modals”, which you can find on my webpage here:
http://www.unl.edu/philosop/people/faculty/dowell/dowell.shtml ).
About your case: Whether or not the Consultant decides to make the information about Patient’s taking W available to Doctor does not affect whether or not Doctor is sensitive to that information.
Let me explain why I say this. To do that, I need to back up and say more about how I understand the relevant dispositions. My view about dispositions is utterly boring and standard. Here goes: Take fragility. At T1, a certain glass vase is fragile. That means that, at T1, it’s got some property in virtue of which it will break, if I drop it against a hard surface. Its fragility is a disposition it has already at T1, before we drop it or whether we drop it later.
So, too with our referential intentions. We have something we mean to say while we are saying it. This is indeed ‘something we have in mind’. Having certain intentions is a mental feature of the person who has them. The question is: How does having this feature distinguish the person who has them from someone who doesn’t? Where the intention in question involves a kind of new-sensitivity, my answer is: It makes them disposed to react in certain ways to new information, to recognize it as in or out of the body of information they meant to take into account at the time of their utterance. This feature of a person, their intention, is a feature they have at the time of their utterance, independently of whether and how that disposition is later triggered. It’s just like fragility. Whether or not the vase is later dropped is neither here nor there with respect to the question of whether the vase was fragile at T1.
It seems to me that there’s an idea here behind your question that is common with one behind Mark’s: Why wouldn’t deliberators make it easier on themselves by having more modest intentions? Why not have intentions that refuse to recognize certain information that is pretheoretically relevant to the question of what to do when doing so would make saying something true easier, on the view I’m defending? The answer is just a contingent fact about human beings: We don’t make it easier on ourselves. We just are news-sensitive in just the ways we think a theory needs to capture—e.g. we aim to take into account relevant information that it is foreseeable we will want to gather (e.g. by consulting other experts) and so too with unexpected sources of relevant information.
3. So….I completely agree that “our referential intentions…are…a function of what we had in mind at T1, rather than a function of what we would do or say at times after T1”. To think that appeal to dispositions blocks me from saying this is to presuppose something about dispositions that I reject. (See my comments under #2; having a disposition to Q at T1 shouldn’t be confused with Q-ings after T1, for example, being fragile at T1 shouldn’t be confused with breaking after T1. Having a disposition at a time is what explains later manifestions of that disposition. And, of course, an object can have a disposition at a time, even if the circumstances of the disposition’s triggering are never realized.
Thanks for the explanations. I stand corrected on dispositions. I’m still skeptical that intentions can be reduced to dispositions, but I realize that’s because of my own view that (occurrent) mental states are mental actions. Let’s put that aside, and also the ballooning issues about Kratzer’s modal semantics (Kratzer, by the way, appeals to the ordering source to accommodate counterfactuals). I’m a little surprised that you say I “don’t discuss what determines those restrictions in any of your papers on modals”, but it may be that I’ve just been assuming that it’s referential intentions, and have neglected to say it. Thanks for the admonition!
What I’m still not satisfied about is your treatment of the Doctor/Consultant case. The problem isn’t at all that I think referential intentions should be simple. (After all, I’m inclined to identify the intention here as something like ‘relative to the best information that it’s possible for the deliberators to acquire prior to the point in time at which further enquiry starts to have diminishing utility’.) Rather, I’m concerned that referential opacity not be used as a cover for fuzziness about reference. So, with apologies for pressing this worry so stubbornly:
What is the relevant intention in this case? In the original post you identified it as concerning the information Doctor will have when she tries to act. I pointed out that this can’t be right, but I don’t think you’ve given us a concrete alternative yet: instead you’ve talked about the information that Doctor is disposed to find relevant. But (as I thought you conceded to Mark) that can’t be right. Because presumably Doctor is disposed to find relevant ANY information that could help her make an objectively better decision, including information that is completely unavailable to her. That means that it would never be correct for her to judge that she ought to take what she knows to be the second-best option (drug Z) because she just doesn’t know which option is best.
That’s why Gunnar and I talk about ‘available’ information, and I’ve been projecting the same view onto you, although looking back I can’t see that you’ve endorsed it. Indeed, now you say that Consultant’s choice not to make his information available to Doctor does not affect whether Doctor’s judgment is sensitive to that information. But now suppose that instead of Consultant we have Greedy Heir, who knows what Consultant knows but WANTS the patient to die (and suppose that Doctor knows this). Surely Doctor’s deliberative question ‘Which drug ought I to prescribe?’ is not sensitive to Greedy Heir’s information.
We agree, I think, that deliberative ought judgments are looking for conclusions that are within the deliberator’s power to reach (perhaps with a bit of help and good luck). That seems to require that they are relativized to information that it is within the deliberator’s power to reach. But then it does make a difference to whether Consultant’s information is included whether he is willing to make it available. And that takes us back to our original problem: Doctor may be more likely to get the answer to her question right if Consultant’s information is not available, than she is if his information is available. So if we supposed that deliberation and advice are both fundamentally concerned with getting the right answer to the same ‘ought’ question (which is what Gunnar and I think should be given up), then it seems that Consultant would be more cooperative if he kept his information to himself.
Hi, Steve,
Sorry to have taken so long to get back to this. Two quick points. The first is that if I said anything that suggests I’m reducing referential intentions to dispositions, I misspoke. I don’t need to take a stand on their relationship accept to say that having a referential intention with a certain content confers on the speaker a disposition to use and evaluate their usage of a modal sentence in conformance with an certain pattern.
The second point is just a bit of background about the literature on modals. Somewhat surprisingly, not only has no one besides me argued in print for the claim that referential intentions determine modal restrictions, no one else offers any prediction-generating, metasemantic proposal for how those restrictions are determined. In my paper “A Defense of Canonical Contextualism about Epistemic Modals” I offer this oversight as a diagnosis for why some of the puzzles about EMs appear puzzling.
(You can find the paper here: http://www.unl.edu/philosop/people/faculty/dowell/dowell.shtm )
So, if you’re interested in reading anything on different proposals for how context selects domain restrictions for the quantifiers, you need to go to the literature on quantification over individuals, since, as far as I now, there isn’t any in the modals literature, besides my EM paper.
Ok, now back to the Doctor case and identifying the content of a speaker’s referential intention. To see how my view handles this case, I first need to get a bit more of the view on the table. On my view, referential intentions confer on us dispositions whose triggerings allow us to identify the nature of the dispostion and thereby form a evidentially warranted hypothesis about the contents of the intentions themselves. (Anyone interested in knowing more about my view about the relationship between dispositions and referential intentions, should have a look at my paper “Empirical Metaphysics” in the Phil Studies BSPC volume that came out in 2008 (also on my webpage) in which I provide a detailed spelling out and defense this account of their relationship.)
We need to distinguish between a speaker’s disposition to use and evaluate their usage of a modal sentence and what their intention, given an actual context, puts in the domain-determining restriction. I’m thinking of news-sensitivity as a kind of general disposition that deliberators have, one that alone doesn’t settle what information plays a domain restricting role. So, that Doctor is news-sensitive to Consultant’s information does not mean that Consultant’s information is automatically in the domain-restricting set. That disposition can only help give us information about what information plays that role, together with a context. As I’m understanding your Consultant example, this issue is whether his information is within Doctor’s power to reach. I completely agree that is a crucial factor in determining whether a piece of information is in or out of the domain-restricting set, as I hope I made clear in one of my replies to Mark. Yes, yes, no one has an intention to take into account relevant information they couldn’t possibly have before acting. Cases in which some information is beyond a deliberator’s power to reach are cases in which that information can’t play a domain-restricting role. So, news-sensitivity is a matter of being prepared to take into account relevant information when it is available.
Notice: Without a concrete spelling out of the notion, “availability” here is just as much a fudge term as “referential intention”. Fortunately for me, it is easy to spell out availability in terms of speaker’s intentions, given that I’ve got a concrete proposal for spelling out how we come to discover their contents. Without such a spelling out, you’re stuck with the objections to that notion raised by von Fintel and Gillies and by MacFarlance to that notion when DeRose relied on it in his contextualist proposal for epistemic modals.
Rats. I’ve actually got to run now. But more on consultant late (my time 🙂 ) today….
Ok, I’m back. There seem to be two questions here: What is the content of Doctor’s intention? and How is Consultant’s providing his information cooperative on my account?
Remember that, on my account, a plausible specification of the content of anyone’s referential intention requires information about the triggerings of the disposition having that intention confers. As with any disposition, a concrete, warranted hypothesis about its nature requires information about more than a single triggering. (Imagine trying to develop a plausible, highly specific hypothesis about the nature of a particular glass’s fragility from dropping it once.) So, in order to have something very concrete and plausible to say about the content of Doctor’s intention, we need to fill out the case more; we’d need to know, for example, how would Doctor evaluate what she said in hindsight in various different scenarios?
But that doesn’t mean that we can’t say anything plausible about the content of her intentions from our knowledge of what she said, the context in which she said it, and general information about what speakers in such context are like. (I.e. in the usual, Gricean way.) We can say, for example, that Doctor is in a context of deliberating about what drug to prescribe for Patient, that deliberators tend to be news-sensitive, and that doctors typically are aiming to do that which can be expected to best promote their patient’s health, when deliberating as a doctor about what to do.
What about Consultant? Would it be more cooperative, on my account, for him to share his information with Doctor, or to withhold it? The answer, on my account, is pretty straightforward. Consider the context and what information Consultant has about it. He knows that Doctor’s evidence supports that either X or Y would provide a complete cure of Patient, while Z would provide only a partial cure. Consultant also knows that Patient is taking W and that taking W in conjunction with exactly one of X or Y, is lethal. He also knows that W is not lethal when combined with Z; Z, with W, is certain to provide a cure, as before.
His knowledge of Doctor’s context also supports that Doctor is aiming to do that which can be expected to best promote Patient’s health. Cooperating with someone else involves doing what helps supports their aims. So, the question of what Consultant should do, provide the information or not, in order to be cooperative, is just the question, what should Consultant do in order to do what he can expect to best promote Doctor’s aim of promoting Patient’s health? or: Does providing the information about Patient’s medical history better promote Patient’s health or does withholding the info do that?
On my account, we answer that question by looking at the worlds in which Consultant provides his information and the worlds in which he doesn’t, each of which is also compatible with the rest of the information consultant has, including the information he has about Doctor’s aims. It’s clear, I hope, that the worlds in which Doctor has Consultant’s information are worlds in which she can be better expected to promote Patient’s health than the worlds in which she lacks that information–after all, she has a 50% likelihood of killing Patient in the latter, and none in the former. This means that, on my account, the cooperative thing for Consultant to do is share his information. (fyi: you can run the same argument, even if the only info that Consultant has that Doctor doesn’t is that Patient is taking W.)
Thanks, Janice. You’ve now committed yourself to the answer that if Consultant chooses not to make his information available, then the truth of Doctor’s claim isn’t sensitive to it. I think this is the right thing for you to say; that’s what the news-sensitive contextualism that Gunnar and I embrace says, too.
In fact I agree with almost everything you say in the last two posts. But I think that it undermines some of the stances you took in the original post. Here’s why.
You say that Doctor ‘is aiming to do that which can be expected to best promote Patient’s health’. But this is problematic, because ‘can be expected’ and possibly even ‘best promote’ are also information-relative. Bear in mind that Doctor KNOWS that drug Z can be expected NOT to best promote Patient’s health; either X or Y does that. So if her aim was as you say, it would seem that by choosing Z she takes the one option that she KNOWS will fail to realize her aim.
So it seems we have to interpret ‘that which can be expected to best promote Patient’s health’ in such a way that it picks out the best choice relative to information–presumably defined in a news-sensitive way. But if we say that, then we still need an explanation of why Consultant would help her with this aim by making his information available. For (as has already been said many times above) Doctor is set to successfully accomplish this aim without his intervention.
Your reply to this, I think, is:
the worlds in which Doctor has Consultant’s information are worlds in which she can be better expected to promote Patient’s health…
But I think this equivocates; it is only relative to our/Consultant’s information that she can be expected to better promote Patient’s health with Consultant’s information, and we’ve already agreed that it is not Doctor’s aim to act in the way that is best given somebody else’s, unavailable information.
But here you’re getting very close to what Gunnar and I argue in our paper. Not to make assumptions on Gunnar’s behalf, I would suggest that Doctor has two relevant aims: to cure Patient (an aim that can be realized to different degrees) and to avoid killing him (an overriding aim). Consultant’s information would put her in a better position for pursuing these two aims, and this is why Consultant cooperates if he makes his information available, but fails to cooperate if he does not make it available.
The crucial point here is that these aims CANNOT BE IDENTIFIED with the aim of finding the answer to some information-relative ought-question. For once you’ve granted us that information that is not made available to Doctor isn’t included in the domain for her enquiry, there is no ought-question that Doctor could be asking such that she is more likely to get the answer right if Consultant makes his information available than she is if he does not.
The lesson we draw from this is that the contextualist should say that deliberating agents have only a contingent and instrumental interest in the truth of any ought-propositions. They have more basic interests in the various values or things that matter to them. And we think that this is a necessary part of a satisfactory contextualist explanation of how advice connects with deliberation.
Having advocated a news-sensitive strategy of our own, we are perfectly happy with the details of your account of what the relevant ought-questions are. But I think that the above considerations undermine some of the things that you said in your original post, as follows.
You motivate your version of contextualism by suggesting (as Kolodny & MacFarlane do in their attack on contextualism) that to accommodate the fact that Consultant is advising Doctor (in a cooperative way), the contextualist needs to show that Consultant aims to ‘give a superior answer to the very question Doctor aims to answer’. I hope to have shown above that this is false: it is sufficient that Consultant is trying to put Doctor in a better position to promote her aims of curing Patient and not killing him. This is what really matters to her, and if Consultant helps her with this, then it is irrelevant to her whether or not he helps her answer her original ought-question correctly.
So while I agree with you that
If Doctor is news-sensitive, it is better to understand her as intending to speak to (something like) the question “what ought I to do at t3 given the information I’ll have then?”
I disagree with you when you say,
it would be odd if Doctor were interested in which act would be best at t3, given what she knows at t1, only to abandon her interest in that question, should new information come to light prior to t3.
Since her interest in any ought-proposition is only instrumental, it would make perfect sense for her to abandon interest in it when her situation changes so that its instrumental value passes to a different proposition. This is the main solution that Gunnar and I explore.
To be perfectly clear (I hope!): I agree with you that deliberating agents can be expected to be asking news-sensitive ought-questions, given their interests. But (a) I don’t think that news-sensitivity by itself is a satisfactory explanation for how advice connects with deliberation (here I agree with Kolodny & MacFarlane), and so (b) I think that the need for a satisfactory explanation of this connection does not support news-sensitivity over any other contextualist account, as I take you to have suggested above.
One final, less important point: in your second-to-last post, you wrote,
Yes, yes, no one has an intention to take into account relevant information they couldn’t possibly have before acting.
While this is true, I think it is false to conclude that ‘ought’ in the mouth of a deliberating agent is never relative to unobtainable information. As Elizabeth Barnes pointed out to me recently, there are some contexts where it is perfectly normal to say something like, ‘There’s simply no way for me to know what I ought to do.’ Her suggestion as I understand it, which Gunnar and I think is right, is that our standards for sufficient information often depend on how much is at stake. If not much is at stake, then we might be perfectly happy with the information we have. But if it is a matter of life and death (Elizabeth’s example involved giving drugs that would save/kill one’s child’s), then we might reject even the best information we can acquire as sufficient for a judgment about what we ought to do.
Apologies for the monstrously long post!
Thanks, Steve. Your comments have helped me better to understand why you and Gunnar take the line you do.
Some points before I pick up the main thread of your post.
1. Please remember that these are blog posts, not a paper, so sometimes I’m simplifying a bit for the sake of brevity. If you haven’t already, I’d like to recommend that you read my paper on epistemic modals on my webpage. There I develop an overall, single contextualist treatment of all the modals and you can get a clear idea of my view in more detail than I can give in a blog post. In the posts here, I’m focussing on how my view handles a particular case (the Doctor case) and you can see how the view handles a much broader array of cases, e.g. Barnes’ case, from that paper.
2. On Elizabeth Barnes’ point: My ‘yes, yes’ remark wasn’t meant to cover all cases, only the one under discussion, one I’d been assuming we agreed was an ought-statement whose truth was relative to the speaker’s information. I agree with her that her sample sentence is felicitous in some contexts and disagree with her suggestion about why that’s so. “I simply have no way to know what I ought to do” can be felicitous when the value for the first of the two parameters are circumstances, not information OR when the agent’s intentions pick out a body of information that goes beyond her information. The former kind of case would be a so-called ‘objective’ ‘ought’, on my view. I haven’t discussed those here, since the Doctor case is a Jackson case and it’s generally agreed that those are ‘subjective’. (I don’t like that terminology because it suggests ambiguity where I think there is context-sensitivity.) The second kind of case is familiar from the literature on epistemic modals (originally from DeRose’s paper). I discuss this kind of case in my EM paper. Suppose (adjusting his example to make it an ‘ought’ case) Patient has an incredibly paternalistic doctor. Doctor runs various tests to determine the cause of Patient’s symptoms, but he absolutely refuses to share the information with her. Instead, he simply tells her what treatment he has chosen for her. After exhausting all her options for getting his information, Patient explains the situation to her friend. She concludes, “so you see, I simply have no way of knowing what I ought to do. Maybe Doctor knows, but I can’t be sure that he has my best interests in mind.”
(For more on how this kind of case is treated in my account, see my “Canonical Contextualism about EM” on my UNL webpage.)
3. I should be a bit more precise about what I’m thinking about the issue of the Consultant providing his information. If the Consultant doesn’t provide it, but Doctor in a hindsight evaluation in which she has the information recognizes that there was, prior to action, some way that she recognizes as practicable that she could have obtained it, and on those grounds recognizes it as part of the information she intended to include in the domain restriction set, then the Consultant’s information was in the relevant set. When I say that it’s wasn’t, I am thinking of a case in which, in a hindsight evaluation, Doctor does not recognize that information as having been in the domain restricting set. This is evidence for her original intention to exclude it.
So, I’m assuming that we both mean to be talking about a case in which Doctor has NO WAY to possess the information Consultant has prior to action’s becoming necessary. As I’ve said, IF a speaker’s use of a modal expresses a proposition whose truth is sensitive to her information (as we have in the Doctor case) AND IF there is NO WAY some piece of relevant information could be known by her (where ‘could’ is an ability, given circumstances modal) by an investigation she recognizes as practicable (i.e. as determined by her intentions), THEN she is not best understood as intending a modal whose domain is restricted by that information.
This feature of the case is important to keep in mind here. This feature makes it different from the original one I discussed in important ways and we need to be careful not to shift between the two. (The original case is one in which the Consultant does provide his information and the issue there was whether or not that information was in the domain-restricting set. I argued that, (roughly) if Doctor recognizes it as the kind of information she intended to take into account all along, then it is in.)
Let’s keep our eye on the larger picture here: My view is a flexible contextualism where the domain restriction for the use of a modal, when it is restricted, is determined by speaker’s intentions for what, in context, she would recognize as determining the values of two, *separate* parameters. In the case of modals used normatively, one of those parameters takes either circumstances or information as value, the other a standard of some kind.
Right now, we’re not thinking of a case in which there is no fact of the matter yet about whether Consultant will provide the information. We’ve stipulated that he isn’t. That (together with the long thing above) puts it out. If there *isn’t* a fact of the matter yet, then there isn’t a fact of the matter yet about whether it’s in the domain-restricting set. (Compare this with a similar case involving quantification over individuals: Suppose I say ‘every student in my course will get a D or better” prior to the end of the add/drop period.) The case where there is no fact of the matter, though, is a different case, not the one we’re talking about.
In the case we’re talking about, where the information Consultant has is completely unavailable (in my sense of ‘unavailable’) to Doctor, I think Doctor is best understood as not intending a domain restricted in part by that information. What about Consultant? Steve’s comments make me think that I didn’t express myself carefully enough in my last post, so I’ll try again, more carefully.
We’ve got two questions. 1) What should Doctor do? where the worlds in the domain over which ‘should’ quantifies are determined by Doctor’s intentions (we’re imagining Doctor putting that question to herself, that’s why it’s her intentions that matter) and 2) What should Consultant do, provide the information or not? here I’m imagining that the speaker is me, *I* am assessing what would be cooperative for consultant to do. So, it is my intentions that partly determine the values for the two parameters.
These are different questions. So, it is no surprise that the sentences used to answer them are different ‘should’s in the sense that they have a different value for at least one of the two parameters. To answer the second question, I said in my last post,
“On my account, we answer that question by looking at the worlds in which Consultant provides his information and the worlds in which he doesn’t, each of which is also compatible with the rest of the information CONSULTANT has, including the information he has about Doctor’s aims.”
Yes, Consultant’s information is not the same as the Doctor’s, so the domains in the modal sentences answering the two questions are different. But that’s ok, because the questions are different.
Some things to notice here: I am the speaker of “Consultant should provide the information about Patient’s medical history to Doctor”, but the information selected, I’ve said, is Consultant’s. That’s perfectly fine on my view, because on my view, there is no problem with the speaker intending to select a body of information that is someone else’s. (I discuss this in my epistemic modals paper.) A second thing to notice is that, because my view has two parameters, not one, they can vary independently of one another. Here, what makes the statement a ‘should’ of cooperation is that the value for the standard parameter is the same as Doctor’s, i.e. worlds in the domain are ranked in accordance with how well Patient’s health is faring in them.
My view differs from Steve’s and Gunnar’s in that I don’t think we need to say that we abandon our interest in the truth of a bare normative modal statement whenever we receive new information. The perverse Consultant who refuses to provide his information is a pretty unusual case. In the typical case, where the information isn’t in principle unavailable, we can say that the information is already included, so long as the speaker’s intentions puts it in. We learn what those intentions are by observing a speaker’s reactions to new information, either before or after action becomes necessary. In the original case I discussed, we can imagine Doctor saying, in response to Consultant’s information (Z’) “you’re right; I ought not to prescribe either X or Y; I ought to prescribe Z”. This is not most naturally heard as Doctor changing the topic on herself, as Gunnar and Steve would have it (as I understand them), but as correcting an earlier (perhaps warranted) judgment as false. Or take another, typical kind of case: Doctor receives the information about Patient’s taking W after she’s prescribed Y, but realizes that she could have easily obtained that information if she’d consulted Consultant before writing the prescription. Suppose she says (as would be natural) “I shouldn’t have prescribed Y, I should have prescribed Z”. Here she recognizes Consultant’s information as in, even though she didn’t have it prior to acting, because it was the kind of information her intention determined as ‘available’ (in my sense.)
Two points in response. The first is that our view isn’t what you say it is. You write,
My view differs from Steve’s and Gunnar’s in that I don’t think we need to say that we abandon our interest in the truth of a bare normative modal statement whenever we receive new information.
We don’t think there’s any need to say that either; we also accept a news-sensitive view. But we think (i) news-sensitivity isn’t by itself enough to explain the connection between deliberation and advice, and (ii) that when the full explanation is recognized, we can see that even a news-insensitive contextualism (whether right or wrong) has the resources to explain this connection, and so where I took issue with your original post was the suggestion that the need to explain this connection requires news-sensitivity.
The second point is to try again to explain the thinking behind (i). I agree with the following:
I’m assuming that we both mean to be talking about a case in which Doctor has NO WAY to possess the information Consultant has prior to action’s becoming necessary…[and so] she is not best understood as intending a modal whose domain is restricted by that information.
We also agree that there are two different questions, ‘What ought Doctor to do?’ (as asked by Doctor), and ‘What ought Consultant to do?’ (as asked by us or Consultant), and that this is no problem for contextualism. But this misses my point. The claim that I’m taking issue with is the claim that to understand how Consultant is advising/helping Doctor, we have to understand Consultant as answering the same question that Doctor asked. In my last post, I was taking you to have suggested that this common question could be identified as ‘what can be expected to best promote Patient’s health’, and so I pointed out that we seem to need to give this description two separate readings. You replied by agreeing that there are two separate questions here, but asking what’s wrong with that. What’s wrong with that is that it isn’t a solution to the original problem.
The fact that only a rare and perverse Consultant would choose not to provide his information is beside the point. The problem is to explain WHY he would be perverse to do so (and hence why this is rare). The problem doesn’t exist from Doctor’s point of view, only from Consultant’s. From Consultant’s point of view, Doctor already has the right answer to the question she asks, so long as his information isn’t available to her. Yes, it’s true that if he chooses to make his information available, then Doctor never had the right answer to her question and he corrects her by providing his information (but notice: even if he provides her with what he knows is FALSE information!). But how is this cooperating? Plausibly, to cooperate with somebody’s effort is to act in a way that you think makes it more likely that the effort will be successful than if you did something else instead. Since Doctor is already successful as long as Consultant doesn’t interfere, it seems that interfering isn’t cooperating.
We agree that it would be perverse of Consultant not to provide his information. But even news-sensitive contextualism can’t explain why that is, so long as it identifies cooperation simply as providing the right answer to the deliberator’s question.
That’s the claim that I’m pressing you to give up. And you may implicitly have already done so, although you have not acknowledged it, when you say
What makes the statement a ‘should’ of cooperation is that the value for the standard parameter is the same as Doctor’s, i.e. worlds in the domain are ranked in accordance with how well Patient’s health is faring in them.
I have a dilemma for you. (A) If what you mean to say is that Consultant is cooperating with Doctor because his ‘ought’ is indexed to the same standards/values as the Doctor’s, even though it isn’t indexed to the same information, then Gunnar and I would agree, and point out that this is to surrender the claim that cooperative advice necessarily seeks to answer the SAME question that the deliberator asks, which has been my point all along.
(B) However I suspect this is not what you mean: that you’re contrasting the question of what Doctor ought to do with the question of what Consultant ought to do, which leaves it open that Consultant’s assertion about what Doctor ought to do involves the same sense of ‘ought’ as Doctor’s claim. But if you choose this horn, then notice that again you’ve given an account of what makes Consultant’s interference cooperative that doesn’t appeal to Consultant answering the same question that Doctor asked.
I wonder if the reason we keep talking past each other is that you don’t see the question of what makes Consultant’s interference an instance of COOPERATION a very important one? But it is important, because it played a central role in your original argument for why news-sensitive contextualism has to be right. If news-sensitive contextualism can’t answer this question adequately, then it may actually be inferior to either relativism or an insensitive contextualism that includes some device for answering the question adequately.
To be clear, Gunnar and I think that the best view is a news-sensitive contextualism combined with a pragmatic explanation of cooperation. I am trying to urge on you the need for our pragmatic explanation of cooperation.
Thanks for the comments, Steve. I think we’re still talking past each other a bit, but I think I better understand where. And I better understand your view.
Part of my problem is that I’m not seeing the problem your view is brought in to solve. I agree that, if there is a problem, it’s an elegant solution. But I’m thinking that if we accept the right flexible contextualism, there isn’t a problem.
Let me say a bit about where I’m now seeing as the issue between us as lying.
I want to separate out two issues. 1) In the original case, Consultant provides her information and then asserts “So, you ought not to prescribe X and Y; you ought to prescribe Z” This is in response to an earlier assertion by Doctor “I ought to prescribe either X or Y”. It’s common ground, I take it, that we have the intuition that Consultant is offering a conflicting answer to a common question. The challenge was to find a way for a contextualist to capture this.
2) Imagine Consultant deliberating about whether to provide his information. He’s wondering whether it would be cooperative or not to provide it. Should he provide it? Or, suppose he refuses to. How do we want to assess his action? When we’re wondering whether it would be cooperative or not, are we inclined or disinclined to say “Consultant should provide his information to Doctor”? I think we are inclined to say that he should and thereby recognize that doing so would be the cooperative thing to do. In a case where Doctor is stipulated to not have any way of getting that information, how can contextualist capture that intuition?
I think these are separate issues. (I’m not sure what you think about that.) Since I think they’re separate, I’m treating them separately.
On the first: One of the things that I really like about your and Gunnar’s notion of news-sensitivity is that I think developed in the way I want to and combined with the flexible contextualism laid out in my epistemic modals paper, I think it is easy to say how, in the original case, Doctor and Consultant are giving rival answers to a common question. The story is roughly this: We find the information that restricts the modal in Doctor’s original utterance by finding out which set of information is the set she intends to restrict her modal. I’m thinking of news-sensitivity as a readiness to recognize new, relevant information as in the domain-restricting set. (aside: since you don’t like my reliance on dispositions, I’m assuming that’s not also how you understand news-sensitivity.) So, when Doctor replies to Consultant “You’re right; I ought to prescribe Z” she is best understood as manifesting that intention. What this means is that the Consultant’s information is one perfectly good sense ‘already in’ the domain-restricting set for Doctor’s original utterance, namely, in the same way that a student who is in my course at its end is in the domain for my utterance “every student will get a D or better” said before the end of the add/drop period. Which question Doctor is answering is partly a function of what information is in the domain-restricting set and since Consultant’s info is already in there, they’re answering the same question.
I offer this solution to *this case*. It is a natural and familiar kind of case of two people deliberating together (or one person deliberating and using another as a ‘sounding board’). It is most natural to assume that it is group information, not just the speaker’s information, that matters in these cases–that’s generally the point of the conversation. (I say more about when the relevant info is a body possessed by a group and how group membership gets selected in my epistemic modals paper.) Given all this, I’m not yet seeing a problem for the contextualist about cooperation that needs to be solved, at least not for this case.
Rats. Our realtor is here–off to look at houses. I’ll get to the second issue later today, I hope, but I’ve already said most of what I have to say about it in my last two posts.
J.
Ok. So I take myself at this point to have shown how, on my view, in the case in which Consultant does share his information, his “You (to Doctor) ought to prescribe Z” is a matter of addressing the same question Doctor takes herself to be addressing. That’s because in this case, the values for both parameters of the modals in Doctor’s and Consultant’s statements are the same.
The second case is the one in which Consultant is stipulated to refuse to share the information that Patient is taking W with Doctor and Doctor is stipulated to have no other way to acquire it.
Here, I’m mostly going to briefly repeat what I said in the first two of my last three posts about this case. I want to be a bit more clear about why I’m emphasizing that, on my view, the modals have two separate parameters. One is that keeping the parameters separate yields an elegant solution to puzzles about conflicts between preferences and circumstances (cases in which circumstances are such that not all of someone’s preferences can be simultaneously satisfied). In those cases, keeping the values for the parameters separate allows for clear preference rankings that we don’t get if we don’t keep them separate. (For details, see Kratzer’s “Modality”.)
Another advantage of keeping separate, though, is that doing so allows us to capture all of our intuitions in cases like the one I’m discussing. Doctor is in a Jackson-case situation and in those situations, most people hear the Doctor’s ‘ought’ as sensitive to information, not circumstances. When we add the non-information-sharing Consultant to that scenario, we’ve still got a Jackson-case. Since Consultant has different information than Doctor is able to have (by stipulation), it can’t be that our utterance of “Consultant ought to share his information” shares a domain with Doctor’s “I ought to prescribe either X or Y”. But since there are two parameters for modals, on my view, that does not mean that our modal utterance has nothing in common with Doctor’s. The values for the information parameter are different, but the value for the standard parameter is the same. That’s what makes it the ‘should’ of cooperation in this case and that seems to me to precisely fit what we pretheorectially want to say about it.
I don’t think that this is wholly different from what Steve and Gunnar want to say. Here’s where I now think we disagree. Steve and Gunnar think there is one parameter for our modal, an amalgam of information and standard, whereas I think there are two separate parameters and also that the parameter that sometimes selects information, selects circumstances. Steve and Gunner think there is a problem about how Doctor and Consultant can be seen to be addressing the same question in the original case. To solve that problem, they offer what seems to me the elegant solution of focussing on a practical interest that is shared between Doctor and Consultant. (I’m not sure whether they think this means that the Doctor and Consultant are both addressing a common question and, if so, what that common question is.) I, in contrast, think that with my flexible contextualist account, we can see how there is no problem in that case.
The case in which we say of the non-cooperative Consultant “he should share his information” is a different ‘should’ uttered in a different context. But there is a perfectly good sense in which it is the ‘should’ of cooperation, since the value for the standard parameters are the same. Picking up on a shared standard seems to me to echo in some ways what I take Steve and Gunnar to want to say about shared practical interests in the original case, though we differ on how to flesh out the core thought and on how widely to rely on it for explanatory purposes.
Thanks Janice. We’ve just about talked this topic to death, and I think we’re getting close to a full understanding of each other at this point. I still think you’re missing a bit of what is concerning me, however.
I agree with almost everything you say in the last two posts. In particular, I agree that news-sensitive contextualism enables us to understand how Doctor and Consultant are addressing the SAME question. However, in our paper at least, Gunnar and I are neutral on whether standards and information are two separate parameters or one, and I think you’ve not quite identified the locus of our concern. I’m worried that I’m just going to repeat myself, but here goes.
Let me emphasize that I think that news-sensitive contextualism DOES have a satisfactory strategy for showing how deliberation and advice both address the same question. My worry is simply that the aim of advice, and the explanation for why it is cooperative, cannot be JUST that it helps the agent get the right answer to the question she is asking. There has to be another part to the story.
You insist on ‘separating cases’, but think about it from the Consultant’s point of view. He knows that Doctor’s ‘ought’ is news-sensitive; this means that whether it is true depends on whether or not he provides his information (just as whether a student is in the domain of your claim depends on whether he decides to drop the course prior to its end). When he thinks about what to do, it has to be deliberatively open to him which of the two cases he inhabits. (You may want to object that he can’t CHANGE the domain for Doctor’s earlier judgment by his decision; I think the right way to think about this is that when he decides what to do, he in effect decides what the domain was from the start).
(By the way, I agree with you that if Doctor’s ‘ought’ is made in conversation with Consultant, then it may intend reference to his information whether he provides it or not. To neutralize this factor, imagine that Consultant is an ‘unexpected advisor’).
The curious thing about Consultant’s choice is that if he decides to do nothing, then Doctor already has the right answer to her question. Of course, if he decides to provide his information, then presumably Doctor will end up with the right answer to her question as well. But if we take the view that the reason why Consultant’s advice is cooperative is just that it assists Doctor get the right answer to her question, then there is a real puzzle why the cooperative thing to do is to provide the information, while it would be perverse for him not to provide it.
Let’s make the problem more acute, in a way I suggested previously. Suppose that in the absence of Consultant’s information, Doctor would already have the answer to her question right. But if Consultant chooses to make his information available, Doctor is likely to miscalculate and make the wrong choice. Suppose that instead of one alternative drug that is better than Z, there are ten different drugs all better than Z; if Doctor miscalculates she will give Patient one of these that isn’t the optimal one. So even if Doctor calculates wrongly, Patient will still be better off as a result.
Now, is it cooperative for Consultant to provide his information? By providing his information, it seems that he makes it LESS LIKELY that Doctor arrives at the correct answer to her question. But surely the cooperative thing to do is to provide the information, because that’s what’s best for Patient’s health.
So we think that the story about cooperation must have another part to it. Which is, we think, that what matters more fundamentally to both Doctor and Consultant than arriving at a true ought-judgment is (1) Curing the patient, and (2) avoiding his death.
Perhaps you have already addressed this problem above, and I’ve just missed it. You have given an account of cooperation that doesn’t say anything about providing an answer to the same question, and talks instead about the same standard. So perhaps we have after all arrived at the same view.
So what does any of this matter? As I’ve been saying, I think it matters because it undercuts the way you were motivating the need for a news-sensitive contextualism. But I’m happy to let the matter rest at this point.
Hi, Steve,
Well, maybe we can at least agree that I don’t see a problem for my view. Earlier, you seemed happy when I separated out the two cases and you then asked me to say what my view would say about a case in which it’s stipulated that Consultant does not provide his information. Having answered that in a way that you now say you agree with, you say there there is another objection that I am refusing to answer, but I’m just not seeing what it is. The Doctor at the time of her utterance has an intention. That intention’s content is given in the way that I have already characterized. One thing that does not give it’s content is the Doctor’s mental enumeration of each source of information. Compare again the case of domain restriction with quantifiers over individuals. Before the add/drop period is over, I *can’t* enumerate every student in the domain of “every student will get a D or better for the course”. But that doesn’t mean that there isn’t a domain for my quantifier. Depending on your view of determinism, it’s either settled and I don’t know what it is or it’s settled in the future. To settle the domain, we need both a speaker’s intention and the individuals in the world that ‘answer’ to it. A student’s decision to add or drop can, then, make a difference to what has to be true in order for what I’ve said to be true.
Consultant’s role in the cases in which the information is conveyed is a bit like the student’s, I suppose. It doesn’t matter here whether the Consultant’s information is expected or not. Let it be totally unexpected. Once conveyed, the Doctor can react to that information in a number of different ways. If she receives the information before writing the prescription, a natural thing for her to say is “oh. I guess I was wrong. I ought to prescribe Z”. If she does, on my view, she is manifesting a prior intention to include information like Consultant’s.
[Maybe (I’m not sure), it would help to remember that, on my view, it’s the speaker’s intention, given facts about the world, that together determine what the domain-restrictor is. But the speaker can be ignorant of or incorrect about the relevant facts. That’s how, even, when Consultant’s information is unexpected, the information may still be, as in the fuller case just described, already in the domain-restricting set. ]
Now, I see that you’ve invited me to look at things from the Consultant’s perspective. But that is just what I think I’ve done in my last couple of posts. I suppose I most explicitly addressed the question: How should we assess Consultant’s decision not to share the information?” since we had agreed to stipulate we were discussing a case in which the Consultant doesn’t share his information. So, let me explicitly address the issue that, as I best understand it, is the question you want answered: Should the Consultant share the information or not? What would be more cooperative? How can my view accommodate a ‘yes’ answer to the first question and a ‘sharing would be more cooperative’ to the second? Here I’m just going to say what I said about our assessment of Consultant when he doesn’t share his information. The question “should Consultant share his information with Doctor?” is a different question from the question “what should Doctor do”? When Consultant makes claims about what Doctor ought to do in conversation with Doctor, he is best understood as aiming to address the same question as Doctor. My view captures that in the way I’ve already discussed. But when Consultant asks himself “should I share the information or not?” he himself deliberating about a different practical question. So, no surprise, the domain for his quantifier may be different. What determines it? Consultant’s intentions at the time of his utterance. Let’s assume what I think best addresses your concern, that Consultant intends to be asking about what would be most cooperative. On my view, that having that intention means that the value selected for the standard parameter for the question is the same as the Doctor’s when she says “I ought to prescribe either X or Y”. If Consultant is reasonable, he will take that standard to be given by how Patient’s health is faring in each of the possibilities in the question’s domain. What restricts that domain? Well, here again the answer depends upon Consultant’s intentions. Depending up what his intentions are, the value for the second parameter may be either by some body of information that his intention selects, given the facts, or the circumstances themselves. Since Consultant is deliberating, let’s stipulate a natural case in which he intends to select a body of information. The next question is: which body of information is he best understood as intending? We’d need to fill out the case, but it seems typical for a speaker to intend a body that includes his own information. If so, then what we have is the following question: Among the worlds compatible with the body of information that at least includes Consultant’s, in which set of those worlds is Patient healthier, the worlds in which Consultant shares his information with Doctor, or the worlds in which he withholds it? It doesn’t take an MD to see that in the stipulated scenario, Patient does better in the worlds in which the information is shared–after all, he doesn’t die in nearly as many of them.
You seem to think that I am for some reason forced to assume that when Consultant is deliberating about whether to share his information with Doctor, he is putting to himself the very same practical question Doctor does when she asks herself which drug she ought to prescribe. You seem to think that if I can’t show how on my view those questions are the same, Consultant does not count, on my view, as cooperative when or if he shares the information. But my view *shouldn’t* hold that those two questions are the same for the simple reason that they are *different* questions and seeing Consultant’s act of sharing the information as cooperative does not require that they are the same.
Hi Janice,
One thing that we can perhaps agree on is that we are both very stubborn!
From the fact that I have already agreed with virtually everything you argue in your last post (except for your claims about what I think!) it’s clear that I haven’t done a very good job explaining myself!
All I have been trying to argue is that in order to understand how Consultant is being cooperative, we need more than simply an account on which he gives an answer to the same question that Doctor deliberates over. Your last paragraph almost gets to the crux of it, but my concern is not that you are FORCED to assume what you say above, but merely that you needed to say something further to explain how Consultant is being cooperative. Without saying something further, it seemed to me, you hadn’t yet given an account that fully meets the demands of contextualism’s opponents (like Kolodny & MacFarlane). What Gunnar and I try to do in our paper is to explain this extra bit of the story (although we also go on to suggest that the ‘extra bit’ of the story may actually be a sufficient defense of contextualism by itself, but put that aside).
However, it seems to me that you now HAVE provided that extra story, in your account of what makes Consultant’s advice cooperative. It seems to me that this account is effectively equivalent to our own. So I think we are now in agreement about how to defend contextualism.
Perhaps I can make one correction to your characterization of my concern, without upsetting this (proposed) truce? You suggest that I…
“seem to think that I am for some reason forced to assume that when Consultant is deliberating about whether to share his information with Doctor, he is putting to himself the very same practical question Doctor does when she asks herself which drug she ought to prescribe”
That’s not quite right. We can distinguish between (i) the question that Consultant asks himself, in deciding whether to intervene, and (ii) the question to which Consultant offers an answer, when he gives his advice. I do think that you’re forced to assume (X) that the question (ii) is the same question that Doctor asks herself–that’s just your view! But I have not tried to pin the identification of (i) with the Doctor’s question onto you. What I’ve been arguing is that (X) is not by itself enough to explain why Consultant’s advice is cooperative. And now you’ve agreed.
So…pax?
Oh, cool; we agree. An ideal outcome. Thanks again for all of your comments. I’ve found thinking this through with you very helpful.
A complete draft of my flexible contextualism about ‘ought’ paper is now up on my UNL webpage here:
http://www.unl.edu/philosop/people/faculty/dowell/FCOLastUndividedWeb.htm
Sorry about the formatting errors. Those are being worked on. Anyone who wants a copy in microsoft word, just send me an e-mail and I’m happy to send one along.
J.