Practical Conditionals

Practical conditionals are a problem.  We all use conditionals like, “If you want a great steak, you ought to go to Manny’s Steak House.”  But suppose I do want a great steak; does it follow that I ought to go to Manny’s?  No—maybe my doctor has told me to lay off the red meat.  Then is the conditional false?  That doesn’t seem right either, if Manny’s really is the best place for a great steak.

We can put a sharper point on this problem.  Suppose your best friend’s wife is very attractive, attractive enough to put ideas in your head.  (Readers of other persuasions will have to generate their own example.)  But suppose also that you think sleeping with one’s best friend’s wife is morally repugnant.  Consider the two claims, both plausible in their own way:

(A) If you want to seduce your best friend’s wife, you ought to spend a lot of time alone with her.

(B) If you want to seduce your best friend’s wife, you ought not spend a lot of time alone with her.*

Add the premise that you want to seduce your best friend’s wife, and we have two parallel modus ponens arguments, one of which concludes that you ought, and the other that you ought not, spend a lot of time alone with her.

Clearly, both these arguments can’t be sound while there is no equivocation.  So here are the options as I see them:

  1. It’s impossible for both conditionals to be true.  This just seems wrong; there can be cases where, on some reading of each conditional, both are true.
  2. The arguments have different structures, so one is valid and the other is not.  This is probably the standard view:  the claim is that (A) is Ought( you want to seduce à you spend time alone) while (B) is You want to seduce à Ought not( you spend time alone).  Jamie Dreier has a paper attempting to debunk this theory.  In any case, I’ll ignore it for now.
  3. There’s an equivocation somewhere.  Since the contents of the desires and oughts don’t matter, nor the person to whom the conditional is addressed (you, me, or Jones), the only candidates here are

3a.  an equivocation on ‘ought’.  Maybe (A) contains a subjective ought and (B) contains an objective ought.  One problem here is that the consequents of (A) and (B) can be replaced by imperatives:  “…spend time alone!”  “…don’t spend time alone!” and while both conditional imperatives seem reasonable in their different ways, the ‘ought’ is not around anymore to blame the conflict on.

3b.  an equivocation on ‘want’.  This is the possibility that attracts me and which, so far as I know, is unexplored by anyone else.  I’ll be interested to see what others think.

Here’s my main reason for being attracted to the idea of an equivocation on ‘want’.  Many conditionals, including (A) and (B) I think, encapsulate arguments or lines of reasoning. From “If P, then Q” we can usually get, “P, therefore Q.”  But if we try this on (A) and (B), we see that the lines of reasoning being employed are quite different.  (A) encapsulates an instrumentalist inference:  Here is an end you have, so you ought to take the most effective means to that end.  The inference behind (B) is not instrumentalist.  And what I would argue is that the starting points of these two inferences are different.

Instrumentalist inferences reason from desires or intentions to more desires or intentions.  The premise(-state) is of the same type as the conclusion(-state).  Whereas the inference encapsulated in (B) is reasoning from a belief about one’s desires.  If I can invent some formalism, the difference between the two lines of reasoning is

(A-argument)

[Des]    I seduce my best friend’s wife

[Des]    I spend time alone with her.

(B-argument)

[Bel]     I want to seduce my best friend’s wife

[Des]    I don’t spend time alone with her.

So my claim is that the antecedent “you want to seduce your best friend’s wife” in (B) above is exactly what it seems, a description of a desire.  But in (A), it’s the expression of a desire.  That’s why the modus ponens argument using (B) goes through, but the one using (A) does not.

(Q: Er, where did the ‘ought’ go?  Shouldn’t the conclusions of these inferences be, “[Bel] I ought (/not) spend time alone with her”?  A:  Um, yes.  Short version of my semantics of ‘ought’:  those [Bel] conclusions will be true just in case the [Des] conclusions are correct, in a sense of ‘correct’ to be specified.  Work with me here, and concentrate on the premises!)

Here’s another argument for equivocation on ‘want’:  we sometimes use ‘want’ this way in other contexts.  "If you’re going to Kenosha, you want to take the highway."  This is obviously not saying that facts about your desires follow from facts about your destination.  The encapsulated inference is

[Des]    I go to Kenosha

[Des]    I take the highway,

not anything concluding in

[Bel]     I want to take the highway.

So again, “you want to take the highway” in the consequent of the conditional is expressing a desire, not describing one.

* I was first turned on to this kind of example by Michael Huemer, in a much earlier thread.  His example was so revolting that I couldn’t concentrate on it for any length of time, so I’ve changed it in this post.

19 Replies to “Practical Conditionals

  1. I apologize for the lack of white space in this post. Any of the other posters who can figure out how to get around Typepad’s habit of eating my line breaks are encouraged to share their wisdom.

  2. Heath, have you tried putting in the html break macros? (It’s
    .)
    I think the “want” really does mean something different in the Kenosha conditional; that shouldn’t be controversial. Mark Liberman says here that it’s the “need” or “requirement” kind of “want”.
    The other stuff actually fits pretty well with what I think, too.

  3. Oooh, that looked right when I PREVIEWed. Oh well. The ‘break’ macro is “br” in angle brackets.

  4. Heath,
    Looks like you’ve illustrated that strengthening antecedents is invalid for these sorts of conditionals. So,
    1. If you want to sleep with your friend’s wife then you ought to spend more time with her.
    does not entail,
    1′. If you want to sleep with your friend’s wife and you’re a decent person, then you ought to spend more time with her.
    For lots of conditionals strengthening antecdents is valid, but pretty clearly (1) does not entail (1′), so it is not valid here. Looks to me like your conditional (B) is just an implicit version (1”) and your (A) is a version of (1). Those are consistent.
    1”. If you want to sleep with your friend’s wife and you’re a decent person, then you ought not to spend more time with her.

  5. Heath,

    Interesting post. I’ve long thought that people are making some kind of mistake with ‘want’ in these kinds of cases, but I’ve never thought of it in terms of practical conditionals.

    I’m not sure, though, that your Bel/Des account of the equivocation does the trick. Wouldn’t you often end up committed to both consequents (e.g., when you both desire to seduce your friend’s wife and believe that you desire to do so)? Also, wouldn’t (B) be true even if you desired to seduce her but didn’t believe that you did?

    At any rate, I think there’s something a bit less technical going on here. I think the problem is that English has no easy way to distinguish between what we might call “temptation-desires” (T-desires) and “all-things-considered-desires” (ATC-desires). If I’m tempted to seduce my best friend’s wife because I’m really attracted to her, but all things considered, I’d prefer not to seduce her (perhaps because I find it morally repugnant), then I ought not to spend time alone with her.

    This shows the equivocation:

    (A) If I T-desire to seduce her, then I ought to spend time alone with her.
    (B) If I T-desire to seduce her, then I ought not to spend time alone with her.

    (A*) If I ATC-desire to seduce her, then I ought to spend time alone with her.
    (B*) If I ATC-desire to seduce her, then I ought not to spend time alone with her.

    (A)&(B) is inconsistent, and (A*)&(B*) is inconsistent, but (B)&(A*) is not. So, the problem disappears once your recognize that ‘want’ is being used in two different senses.

  6. Heath,
    I’ve been thinking a lot about this stuff myself recently, so I really appreciate your post. I’m in the camp that thinks there’s an ambiguity in the ‘ought’. A few thoughts:
    On your proposal, I have two objections. (1) This seems a puzzling sense of ‘expressing’ a desire. If I say, ‘if YOU want to E, then you ought to M’, would you say that I’m expressing YOUR desire? If so, I want to know more, because this is neither expressing a desire in the sense of saying that you have it, or in the expressivists’ sense. (2) ‘Want’ is commonly used in these constructions, but there are a lot of alternatives. ‘If you INTEND to E…’, ‘If you’d LIKE to E…’, ‘If you’re GOING to E’. Do you think all these terms are similarly ambiguous? (Finally, what about, ‘IN ORDER TO E…’, which I think is the most perspicuous formulation?)
    Jamie’s proposal, if I remember it right, is that the practical conditionals are really biscuit conditionals; that is, the ‘if you want…’ antecedent addresses the relevance conditions rather than the truth conditions of the ‘ought’ claim. Perhaps that’s what you mean too — Jamie seems to think you’re saying much the same thing. I actually agree that we’re dealing with biscuit conditionals here (I think both the antecedent and the ‘ought’ are ambiguous). But if that’s right, it seems a bit misleading to say that there’s an ambiguity in the ‘want’. It’s the function of the clause, not the meaning of any of the terms, that is different in the two cases.
    David,
    I don’t think the distinction between T- and ATC-desires is going to solve the problem. For there seems a sense in which it is true that if you want to seduce your best friend’s wife, then you ought to spend time alone with her. But I think there’s also a clear sense in which it is false that if you ATC-desire to seduce your best friend’s wife, then you ought to spend time alone with her. It’s morally wrong, and what you ATC desire is irrelevant to that.
    My own crazy view is that the ‘ought’ doesn’t detach from the antecedent because we’re dealing with probability conditionals, which don’t detach. So:
    (1) If you want to seduce your best friend’s wife, then you ought to spend time alone with her,
    says approximately, ‘Your seducing your best friend’s wife is most likely given that you spend time alone with her, than that you don’t.’ Whereas,
    (2) If you want to seduce your best friend’s wife, then you ought not to spend time alone with her,
    says approximately, ‘If you want to seduce your best friend’s wife, then [Your conforming with your moral obligations] is most likely given that you don’t spend time alone with her’
    I argue for this in a paper in progress, What Ought Probably Means, and Why You Can’t Detach It

  7. Clarification on my last post: I shouldn’t have suggested that you have to say that ‘If you’re GOING to…’ and ‘In ORDER to…’ are ambiguous, since they seem incompatible with the moral reading of the conditional. I guess I was thinking that it’s a bit of a stretch to say that they express desires. Another one that would have to be ambiguous is ‘If you’re TRYING to…’

  8. For the record, I do not say that practical conditionals are (like) biscuit conditionals. I say they are like faith-based conditionals (“If you believe the dispatcher, the plumbers will be here in an hour”) and conditionals of focus (“If we’re only talking about taste, you should order the fried mozzarella sticks”).

  9. Heath, you say:
    Clearly, both these arguments can’t be sound while there is no equivocation.
    Couldn’t both conclusions be true without equivocation? It isn’t obvious to me that ought not entails that it’s not the case that you ought. Suppose you think that the oughts express Rossian prima facie duties. One can have conflicting prima facie duties if Ross is correct, and it doesn’t seem apt to describe this as a case of equivocation. Or, what might amount to the same thing, suppose that to say you ought is to say that there is a reason with some weight to do the action in question. Couldn’t the conclusions of both lines of reasoning be correct? And again, I see no equivocation.
    (I’m going offline once again for a few days some time tomorrow, so if you respond and I don’t, don’t take it as a snub. . . Why is it that this blog always gets most interesting when I go offline?)

  10. It seems to me that there could be ambiguities involving both “ought” and “want” as these terms feature in (A) and (B).
    If you truly all-things-considered desire to seduce your best friend’s spouse, then perhaps that is what you prudentially ought to do (here, I follow David Morrow’s disambiguation of “want,” which I find plausible).
    However, even if you all-things-considered desire to seduce your best friend’s spouse, that does not mean that you morally ought to do so (this addresses the problem that Steve Finlay raised).
    If you desire to some small degree to seduce your best friend’s spouse, nothing follows about what you prudentially ought to do. (We’d first have to know about your other desires, their respective strengths, and so forth.)
    If you desire to some degree to seduce your best friend’s spouse, then you morally ought to keep away from him / her. Perhaps, you prudentially ought to do the same.
    So far, I have left unresolved how prudence, morality, and other norms might determine what a person just-plain-ought to do. But that is a terribly difficult problem, I think. After we have answered it (if we ever do answer it!), we will be able to say some more things about the truth values of (A) and (B) if “ought” indicates the just-plain-ought.
    Also, we will still need some retort to the problem mentioned by Heath White, viz., that the “oughts” in both original sentences can be replaced by imperatives.
    I would start by saying that, when (A) and (B) are modified so that they use imperatives rather than “oughts,” it’s not clear to me whether they are inconsistent or not. (I realize that some views about the logic of imperatives would say that they are inconsistent.)
    Next, I would note that, conversationally, imperatives can be used to make various “ought” claims. Just as I might indicate that a car is ugly by saying, “That’s a real beauty,” another speaker might (given a certain conversational context) indicate that Jane morally ought to help an elderly person cross the street by saying, “Jane, help that person cross the street!” So perhaps to the extent that the statements still sound inconsistent, we are “hearing” them as pragmatic indicators of truth-apt “ought” claims (prudential, moral, or otherwise).

  11. Maybe I’m missing something but how could the antecedent be an expression of desire if it’s not about me at all and if I lack the desire. Lot of the antecedents are of the form ‘if you want’, ‘if one wants’, and so on. How could I express my desires by talking about you? And, I could accept (A) and (B) in a sense without having a best friend who has a wife or without wanting to seduce her. I would have thought that expression requires the thing to be expressed.

  12. Thanks for the comments, folks. Here are the replies I am most able to make:
    When Jamie says that practical conditionals are (like) faith-based conditionals (“If you believe [what] the dispatcher [said is true], the plumbers will be here in an hour”) that is pretty much what I’m trying to get at with the—-perhaps inadequate—-expressing/describing desires distinction. A faith-based conditional has the description of a belief in its antecedent (“you believe what the dispatcher said is true”), but what it means is the expression of that belief (“what the dispatcher said is true”). Obviously, you or anyone else may or may not believe what the dispatcher said is true. Likewise, a practical conditional’s antecedent describes a desire (“you want to seduce your best friend’s spouse”) but what it means is the expression of that desire. The desire in question is not one that you or anyone else need have. My view is that we use this locution because there’s no very good way to express desires in English. I called this an ambiguity in ‘want’, which puts the difference down to semantics. But maybe it’s really pragmatics. Does anyone know how to tell the difference?
    Probably the best way to say what the practical conditional wants to say, as Steve points out, is “In order to seduce your best friend’s spouse, you ought…”. But I guess my main problem with his proposal is that it seems to me that there is such a thing as what you just plain ought to do—-at least we ask what we ought to do all the time—-which is not a claim about probabilities. I think he will need a principle like, “You just plain ought to do whatever is most likely to achieve your ends” and I don’t see how that use of ‘ought’ is probabilistic.
    Mark vR suggests that if we are dealing with prima-facie oughts, there need be no equivocation. True. I guess I was assuming there is such a thing as an ATC ought, and that these kinds of problems arise for that too.

  13. Mark notes that ought not p may not contradict ought p; Jason says that the imperative versions of (A) and (B) may not be contradictory. I don’t think we have to settle those difficult issues, though, because it is pretty clear that in any case modus ponens fails. That is, we don’t need the intermediate conclusion, “if modus ponens worked in both arguments then a contradiction would follow.” We can just notice that from (A) and its antecedent we are not willing to infer the consequent, whether that consequent is an ought or an imperative.

  14. This is a very interesting discussion. I’ll have to think a bit more about the ideas that have been proposed by Heath, Jamie, Mark, Steve, Jason, and Jussi. Still, I’m inclined to think that the simplest view of these matters is the following:
    1. In (A) we have an end-relative or purpose-relative ‘ought’, and the antecedent of the conditional functions to indicate the end or purpose to which this occurrence of ‘ought’ is relative. (So this conditional is a little bit like the “biscuit conditionals” or “faith-based conditionals” that the rest of you have been talking about.)
    2. In (B) we have an ordinary indicative conditional containing an all-things-considered ‘ought’ in the consequent. Hence there is no problem here with detaching the consequent to arrive at an ATC conclusion about what one ought to do.
    So I think we have both an equivocation on ‘ought’ and a difference in logical form. (We need a difference in logical form I think to explain why the consequent of (B) detaches while the consequent of (A) doesn’t.) However, like many others who have posted on this thread, I am sceptical about the idea that there is any ambiguity in ‘want’.
    (BTW, I talk a bit about all these different kind of ‘ought’ in Chap. 5 of my recently published book The Nature of Normativity.)

  15. I was initially puzzled by practical conditionals in trying to figure out how instrumental reasoning works. When we reason instrumentally, it sure seems like there is something like this going on:
    (Minor) I want X
    (Major) If I want X, I ought to Y
    (Conc) I ought to Y.
    If that isn’t instrumental reasoning (and I’ll agree that this isn’t quite it), then something pretty analogous is. And in the case of reasoning from ends to necessary means, it’s as “deductive” as can be. But if anything like this goes on, note the following requirements:
    – at least sometimes, there can be an ATC ought in the conclusion
    – therefore, at least sometimes, there can be an ATC ought in the major premise
    – in these cases, the oughts detach
    – in these cases, the sense of ‘want’ in minor and major is the same
    – in these cases, the major can be paraphrased by “in order to X I ought to Y”.
    My main point here is that any view about what’s going on in practical conditionals needs to be part of a view about what’s going on in instrumental reasoning. So,
    Mike, if my (A) were your (1), MP would go through in that case, but it doesn’t.
    David, it seems to me like a perfectly reasonable (even decent) person could say “even if you ATC-desire to seduce her, you ought not.” But your view has no room for this.
    Steve, on your view, the oughts of instrumental reasoning should never detach. But it seems that sometimes they do. Can you construe (Conc) differently from (Major)?
    Mark vR, there have to be some cases of ATC oughts in the conclusion
    Jason, Ralph, since there have to be some cases of ATC oughts in the conclusion, they have to be there in the major premise as well. It can’t always be a qualified ought.

  16. Jamie,
    Sorry for the misrepresentation!
    Heath,
    On ATC ‘oughts’, my view is complicated, but basically I favor a kind of expressivism about those. (Their semantics are probabilistic, but they represent an expressivistic use). I won’t drag this thread into those complexities.
    On practical reasoning, you’re spot-on. I do deny that there is a difference between Conc and Major, and hence that there is any inference here at all. (I argued this recently in an APA commentary on one of Robert Audi’s recent books. Something similar was suggested by G. H. von Wright, if I remember [my memory has already been called into question on this thread]). One reason for thinking this is that most plausibly our desires are ‘backgrounded’, rather than the contents of our thoughts, when we engage in ‘instrumental reasoning’. But how can a desire function as a premise if it isn’t even present to consciousness?
    More detail: I think that the ‘conclusion’, ‘I ought to Y’, is elliptical for ‘In order to X, I ought to Y’. My psychological states don’t enter into it at all. But if I in fact desire X, this desire operates in the background, and motivates me towards Y. This is my basic model of motivation by normativity.
    Ralph, I can’t wait to read your book!

  17. I take the point, made by both Jamie and Heath, that we need an account of valid practical reasoning – and an explanation of the failure of modus ponens in cases like the present one.
    But I also think that the validity of the argument (whether or not it is a genuinely practical argument) will turn on questions regarding the interpretation of “ought,” “ought not,” and “want” in the major premises – and the interpretation of “want” that is employed in the minor premise, You want to seduce your best friend’s wife.
    The more I think about it, this seems less a matter of an ambiguity among different senses of “want” than it is a matter of missing information. How much do you want to do this? What else do you want to do? We would need to know the answers to these questions (and probably other questions, too) in order to know whether the argument formed by using (A) alongside the minor premise is valid.
    Any plausible view about non-moral normative reasons will have to recognize that (a) the relative strength(s) of a person’s desire(s), and (b) the compossibility of the objects of that person’s various desires both also matter when it comes to what the person really ought to do.
    To know for certain that we have a genuinely practical argument in this case (i.e., one where the conclusion concerns what one just-plain-ought to do) we also need to know enough about the demands of morality to know whether moral considerations are present in the circumstances – and how these moral considerations weigh against prudential and/or merely personal reasons for action in those circumstances.
    I’m sure that there are deeper issues of deontic logic and practical reason in the vicinity, here, but I think that Heath’s example raises the problems that I have mentioned, too.

  18. Jamie (and really everyone),
    I agree that the phenomenon needs to be explained so that nothing follows directly from what I suggested, nor does it show (nor was it meant to show) that there’s nothing to talk about here. I was just cautioning against moving quickly to thinking that we need ambiguity in ‘want’ or ‘ought’ to explain the phenomena since there were models where the inferences would go through and that I think we have reason to accept anyway.
    It is possible that the reluctance to detach in the first argument could be explained pragmatically. If the oughts in that argument are prima facie and we still don’t want to detach, that’s a reason to think there is pragmatic bar to detachment rather than something to do with the semantic values of the terms.

  19. Hope I’m not too late to the discussion, having been offline for a bit. Lots of interesting ideas here, but I’m not sure things are all that complicated. Heath, against the equivocation on ‘ought’ diagnosis, you say: “the consequents of (A) and (B) can be replaced by imperatives: “…spend time alone!” “…don’t spend time alone!” and while both conditional imperatives seem reasonable in their different ways, the ‘ought’ is not around anymore to blame the conflict on.”
    This argument doesn’t seem compelling to me.
    Suppose we read both your original (A) and (B) as wide-scope ought claims about ambiguous oughts, e.g.:
    (A) OughtInst (If you want to seduce, then spend time alone)
    (B) OughtMoral (If you want to seduce, then don’t spend time alone)
    Now, the analogous imperatives look like this:
    If you want to seduce, then spend time alone!
    If you want to seduce, then don’t spend time alone!
    Why is (merely practical) conflict between the two imperatives problematic? Well, you say the problem is that they both seem “reasonable” in their different ways. But that is just to say, it seems as though the following assertions cannot both be true:
    Reasonable(If you want to seduce, then spend time alone!)
    and
    Reasonable(If you want to seduce, then don’t spend time alone!)
    The obvious reply is that what goes for “ought” surely goes for similar normative terms, like “reasonable” as well: it’s not surprising that there would be an analogous equivocation here which we can represent as, e.g.:
    ReasonableInst(If you want to seduce, then spend time alone!)
    and
    ReasonableMoral(If you want to seduce, then don’t spend time alone!)
    (A sort of side note: it doesn’t seem to me as though detachment is permissible in either of the two arguments. It may be true that you morally ought, if you’re going to murder someone, to murder them without painfully torturing them first. But we can’t detach here! For similar reasons, it’s not clear to me that detachment is permitted on a natural understanding of your premise (B).)

Comments are closed.