In “Oughts, Options, and Actualism”
(Philosophical Review 1986),
Frank Jackson and Robert Pargetter defended the “actualist” view
that, for every act-type A, you ought to do A if and
only if your conduct would be (in the relevant way) better if
you did A than if you did not.
In my opinion, this is a deeply
objectionable view. It makes the truth about whether or not you ought
to do A dependent on the brute non-moral facts about what you
would do if you did A
(and about what you would do if you did not do A) –
even if these brute non-moral facts reflect only your utter
wickedness and depravity. In this sense, this “actualist” view
gives an agent’s wickedness the power to effect a radical
transformation in the obligations that the agent has.
For
example, imagine a wicked paedophile, who has just abducted a
10-year-old girl and imprisoned her in his secret cellar. Suppose that it is still possible – though unfortunately quite unlikely – that the paedophile will repent of his evil plans, and return the girl unharmed to her parents. Surely, if
anything is clear about this case, it is clear that it is not true
that the paedophile ought to rape the girl.
But (shockingly, as it seems to me) actualists like Jackson and
Pargetter may well disagree…
Suppose that it is also true in this
case that if the paedophile did not rape the girl, he would torture
her to death, whereas if he did rape her, he would not subject her to
any additional torture, and would not kill her. So, presumably, the
paedophile’s conduct would be at least somewhat better if he raped
her than if he didn’t. Hence actualists must say that the
paedophile ought to rape the
girl. This seems to me a reductio ad absurdum
of the actualist view.
The
alternative to this actualist view is, in my opinion, vastly more
plausible. Jackson’s appeal to counterfactual or subjunctive
conditionals does not yield acceptable results if it is applied to
thin and unspecific act-types (like ‘not raping the girl’); it
yields acceptable results only when applied to much thicker
or more detailed
act-types (like ‘not raping the girl, or harming her in any way, but returning her to her
parents immediately and turning oneself in to the police’).
In particular, I suggest, such
counterfactual conditionals only yield acceptable results when
applied to act-types that are evaluatively maximally specific,
in the following sense. A is a maximally specific act-type (in
relation to the situation of the agent at the relevant time) if and
only if none of the different possible ways in which the agent can do
A in that situation differs from any of the other ways in any evaluatively or
normatively significant respect. Then we can say that, out of these
maximally specific act-types, the agent ought to do an act-type that
is such that, if the agent did it, his conduct would be no worse than
if he did any of the other act-types.
Then I
would recommend extending this picture to the thinner and more unspecific
act-types in the following way: If in the circumstances, the agent’s
doing B is in the appropriate way entailed
by his doing A (or in other words, if his doing B is an essential
part of his doing A), then if the agent ought to do A, he also ought
to do B.
Jackson and Pargetter’s
well-known example of Professor Procrastinate seems to me to
be clearly a case of a “second-best” or conditional
‘ought’:
-
Given
that Professor Procrastinate is not going to write the review, he
should decline the invitation to write it.
This
is true, but only in the exactly the same way as other familiar
examples of the second-best conditional ‘ought’:
-
Given
that you’re not going to stop shooting up heroin, you ought at
least to shoot up with clean needles.
Admittedly,
the positive proposals that I have made here are controversial.
However, it should not be controversial, in my opinion, that Jackson’s and
Pargetter’s actualist view has consequences that are, at least prima facie, quite grotesquely implausible.
Ralph,
That would certainly be a shocking and prima facie implausible consequence.
Would you agree that the pedophile ought to rape-and-desist rather than torture-and-kill?
Consider this modification of J&P’s view:
I’m not sure if ‘would’ or ‘might’ works better here.
With the ‘[would/might] likely enough’ instead of the simple ‘would’, the fact that the pedophile could still simply return the young girl unharmed, combined with how much better this outcome is relative to the other outcomes (i.e. rape-and-desist and torture-and-kill), the view plausibly yields the result that the pedophile ought to return the girl unharmed. In other words, that outcome is enough better than the others to outweigh its unlikelihood.
Hi Ralph,
Excellent post! I completely agree that actualism has absurd implications. However, I don’t think that your example is the best one to show that. It works against the utilitarian actualist, but not against the deontological actualist. The deontological actualist can hold that it would be wrong for an agent to rape someone even so as to prevent oneself or another from killing that someone. If such a constraint is plausible, then the actualist can hold on to actualism while blocking your reductio. I do think, though, that there are examples in the neighborhood of yours that will work against any non-absolutist deontological actualist.
Also, I wonder whether you think that a fully informed and well-meaning advisor should advise the pedophile to rape the girl if she knows that the pedophile will take her advice and knows that the pedophile will kill the girl (which we’re assuming is much worse) if he doesn’t rape her.
Many actualists, including J&P, argue that if such an advisor should advise S to do x, then it follows that S ought (in some objective sense) to do x. I don’t like this argument, but I wonder where you think that it goes wrong.
John – can’t Ralph just modify the details of the case so that the unlikelihood swamps the benefits of the best outcome? It’s certainly possible to construct a case where the expected utility of the unspecific act type ‘not raping the girl’ is lower than the alternative, just because the depraved agent is overwhelmingly more likely to perform an act satisfying this description by torturing and killing her instead.
Ralph – your point about needing “maximally specific” act types seems spot-on. But I’m curious about why you frame this as concerning actualism in particular. I’m not super-familiar with this literature, but I thought that actualism vs. possibilism was a dispute about how your obligations at one point in time depend on what you will or could do at future times. Your problem of underspecified act types seems orthogonal to this debate, since it concerns options at a single moment in time. (I’m assuming that “freeing the girl” is an option that’s available to the paedophile right now, and not just in future.) Am I missing something?
Richard,
It’s not just about future actions; it’s also about simultaneous actions.
I would define actualism as the view that holds whether or not S ought to x depends on what S would simultaneously and subsequently do if S were to x, and not on what S could simultaneously and subsequently do if S were to x. Possibilism, by constrast, is the view that holds that Whether or not S ought to x depends on what S could simultaneously and subsequently do if S were to x, and not on what S would simultaneously and subsequently do if S were to x. These, of course, are not contradictories. Thus, there are people like Woodard who take an intermediate position.
Perhaps I should remind myself of the details of the paper, but the following consequence also seems intolerable. Suppose there are a range of actions which, if the agent S were to have libertarian-style free will, he could perform, and that would be better than the action A that S is actually going to perform. Then J&P would be committed to saying of each of those alternatives that S ought to do it. (S ought to hire candidate 1, and S ought to hire candidate 2, etc. So: it would be correct to say that S ought to hire candidate 6, even though candidates 1-5 are better?–i.e. so long as candidate 6 is better than candidate 7 who is actually going to be hired.) The actualist then can’t avail himself at least of the claim that S ought to do A iff that’s what an advisor should recommend (which is stronger than Doug’s proposal, admittedly).
I don’t think your treatment of Professor Procrastinate can be quite right. Isn’t it rather: Given that he is UNLIKELY to write the review, he ought to decline the invitation? This is something that standard deontic logic can’t handle. Even granted the unlikeliness of his writing the review, it’s still the case that in all the best possibilities he accepts the invitation. (A probabilistic semantics for ‘ought’ gives the right results here, however).
Richard,
I guess you could stipulate that. But what if returning the girl unharmed is infinitely better than the other options? To my mind, it does seem infinitely better. The only way to overcome that, then, would be to make returning the girl impossible, which changes the case.
Doug – thanks for the correction. (Though it seems an unnatural way to carve up the issues to me. I’m inclined towards a purely inter-temporal version of actualism, and the issue being discussed here does seem independent of that.)
John – I guess you could go that way, though it seems incredible. You would then seem committed to the view that it would also be prudent for other agents to gamble on attempting to save the girl from rape, even if their chance of success was negligible, and the outcome if they fail is that she would instead be tortured to death.
But I take it that Ralph did not mean for the example to have this structure. So fill in the details as necessary so that the case has the intended structure: option A is better than B is better than C, but due to the agent’s greater disposition to do C than A, the expected value of B is greater than the expected value of not-B. It seems wrong to say that the agent “ought” to do B, even though it’s [likely enough] better for them to do B than not. Instead, they ought to do A. So JP2 is false, if taken to apply unrestricted (i.e. to unspecific as well as maximally specific act types).
This may be a restatement of Steve’s point, at least in part, but I don’t see how J&P’s view is even coherent. Suppose you have three incompatible options, A, B and C, such that A>B>C. And suppose the following conditionals are true: if you didn’t do B, you would do C, and if you didn’t do A, you would do B. Then B is obligatory (because B>C) and so is A (because A>B). I guess people who like dilemmas won’t see this as incoherent, but this doesn’t seem like a dilemma at all.
Thanks for all those comments! Here are some quick thoughts in response:
1. Steve and Ben both highlight another odd consequence of J&P’s view — it clearly allows for incompatible courses of action all to be what you “ought” to do. In fact, however, J&P are perfectly aware of this. (Indeed, it’s the main issue discussed in the second half of their paper.) E.g., in their view, it is true that Professor Procrastinate ought not to say yes to writing the review, and also that he ought to say-yes-and-write-the-review. They argue that this doesn’t matter, for the following two reasons. First there is still a possible world in which the agent does everything that he ought to (albeit a possible world in which different counterfactuals are true of course); in Ben’s example, this would be the world in which you do A, and in Steve’s example, the world in which you hire candidate 1. Secondly, according to J&P, ought does not agglomerate over conjunction (i.e. ‘You ought to do A and ought to do B‘ does not entail ‘You ought to do A and B‘), so that it never follows that there is a logically impossible course of action that you ought to do.
I completely agree with what Ben and Steve say here, but since J&P believe they have a reply, I thought I’d try a simpler and cruder argument….
2. I basically agree with everything that Doug said in response to John and Richard. However, I think that my objection is already capable of covering deontological versions of actualism as well as consequentialist versions. To include deontological versions of actualism, we need to make it clear that actualism says that you ought to do A if and only if your conduct at the relevant time would be better if you did A than if you did not. (This would allow, in normal cases, that you should not kill one person today even to prevent yourself from killing five people next month.) As I was imagining my rather nasty example, the two options that the paedophile is most likely to take (raping the girl, or torturing and killing her) would each be carried out at the same time. So my objection works equally well against both consequentialist and deontological versions of actualism.
3. I completely agree with Doug, by the way, that appealing to intuitions about advisers is a bad way to explain the meaning of ‘ought’. Steve made an excellent point about this in the paper that he gave here in Oxford in June: rational advisers are aiming to achieve a whole range of objectives (typically including the goal of helping their advisees to act as well as possible); they are not just aiming to utter true ‘ought’-sentences!
4. I’ll try to respond to some of your other points soon…
Ralph,
Why can’t the deontologist hold that you can’t treat a person as a means — not even as a means to ensuring that your conduct at the relevant (present) time would be better than it would otherwise be? So you can’t rape a person at t1 even as a means to preventing yourself from killing that person at t1? This seems like something that certain deontologists (perhaps, Kantians — perhaps, Kamm) would say.
Actualists (at least certain varieties) hold that if the guy will either do less grave evil (raping the girl) or more (torturing and murdering her), he ought to do less. Ralph objects by calling the view grotesque. It would certainly be grotesque to deny, if anyone did, that in raping the girl the guy does a horrible wrong. But actualists do not deny that. In raping her he violates a strenuous moral requirement to refrain from raping and from murdering. Actualists join all humane opinion in saying that the violation is heinous.
I’m taking no stand on whether actualism is correct. I’m only responding to Ralph’s (only) objection, which is that it is a morally grotesque view. Here’s the actualist view, and I ask you whether it is (while possibly mistaken for other unnamed reasons) morally grotesque: If an agent will either do a lesser grave evil or a worse grave evil, then, while his act will be heinous either way, he ought to do the lesser one. What’s left of the objection from grotesqueness? There may be other objections, of course, but that’s different, and no others are offered by Ralph.
Ralph,
What did Doug say in response to me?
Dave,
Doesn’t your remark overlook Ralph’s point toward the end about conditional ought? This is true: Given that the guy will either rape-and-desist or torture-and-kill, he ought to rape-and-desist (i.e. he ought to rape-and-desist rather than torture-and-kill). But is it true that he ought to rape-and-desist? This seems implausible.
If we stipulate that those are the only two psychologically possible outcomes, then adding an ought-implies-can principle would generate the conclusion that he ought to rape-and-desist. But in the description of the original case, it’s still a genuine possibility that the guy will will return the girl unharmed — that is, it’s just barely true that he still can return her unharmed.
[I’ve written a brief follow-up post here.]
Hi John,
I’m not addressing or criticizing that claim of Ralph’s, just in order to focus on what the objection is first. The objection is that it is outrageous or grotesque (not just ultimately a philosophical mistake) to say that he ought to rape her.
Let’s grant that it’s jarring if we hear no more. But that would be as good an objection as saying the J&P jarringly say we ought to do evil. They say this because they say that is what we ought to do when we will either do evil or worse evil. Is that a monstrous thing to say? Well, no, because–and here’s the key point–they agree with the rest of us that it would be evil, and that we will be doing something we ought not to do: violating the requirement to refrain from lesser and greater evils.
Once the fuller story is told, the jarring ought claim seems to be one morally reasonable view. There may be ways to avoid it for a better view, but it is not morally obtuse or grotesque in any way. This may sound like faint praise by me, but that was the objection we were offered, and I’m inclined to think it fails, whatever other objections there may be to actualism. Jumping up and down about that “ought” might trade on connotations it would have in ordinary conversation, which wouldn’t be fair.
David,
You write: “[T]hey say that is what we ought to do when we will either do evil or worse evil. Is that a monstrous thing to say? …Once the fuller story is told, the jarring ought claim seems to be one morally reasonable view.”
If we’re to tell the full story, then we should point out that actualists say that we ought to do evil when we could do no evil but will either do some evil or some greater evil. This sounds pretty jarring to me. And it even seems monstrous when the evil that they’re saying that the agent ought to do is rape a young girl, when this is something that the agent could refrain from doing while also refraining from committing any other greater or lesser evil.
Thanks, Dave. I agree with you that if we’re left to decide between two terribly evil options, then we ought to choose the lesser of two evils, as bad or shocking as it may sound.
I presume you would agree, though, that if he can still return her unharmed, then he ought to return her unharmed. And that’s how Ralph’s case is set up … or at least that’s how Ralph purports to set it up in the 3rd paragraph. Re-reading it, though, Ralph does seem to take it back in the 5th paragraph, when he stipulates what would happen if the guy does, and if he doesn’t, rape. By the end of that paragraph, it doesn’t sound like the guy really can return her unharmed.
John
But you are agreeing that Given that P, ought Q and also that P. And you are denying that Q. The first sounds like a conditional, so you are flouting modus ponens.
Some of us do think that modus ponens is overrated, but Jackson is an old stick-in-the-mud who still holds out hope for it.
Hi Jamie,
Well, initially I was being misled by the possibility Ralph mentioned in paragraph 3. He said that the pedophile still can return the girl unharmed. With that option on the table, it’s not definitely true that P, and he obviously ought to return the girl unharmed.
But now I think that the dominant reading of the case has it that P is definitely true (given what Ralph says in paragraph 5). On that reading, it seems plausible (though still tragic) that he ought to rape-and-desist.
So I’m not flouting modus ponens — it’s just that different readings of the case lead to different takes on whether the antecedent, P, definitely is true.
Either way, though, I think that J&P have a response. If it’s not true, then they can argue that the expected utility of not-raping exceeds that of raping, and so he ought to not rape. If it is true, then the verdict that he ought to rape is tragic but plausible.
Doug,
You point out that the guy could return the girl unharmed. That is, he could not-rape-and-not-murder. And obviously it would be monstrous to deny that this is what he ought to do. And, to show how “monstrous” actualists are, you write, “actualists say that we ought to do evil [rape her] when we could do no evil but will either do some evil or some greater evil.” You say this, it seems to me, as if actualists take the monstrous view that non-rape-non-murder is NOT what he ought to do, but, instead, raping her is. (I say “as if,” because I don’t see how else to see how you think it makes actualism look monstrous.) But that impression is false. They agree that he ought to not-rape-and-not-murder. That is the strenuous requirement that they say he will violate if he rapes her. Can you still make the view look monstrous when this is made perfectly clear? I don’t see how.
Here’s the actualist view: If he rapes her he is committing a grave evil by violating as important a requirement as morality contains. Still, if he will either rape her, committing a grave evil, or murder her, committing an even worse evil, he ought to do the lesser evil. Either way, his act is heinous.
Just as a reminder: I’m asking what the objection from grotesqueness or monstrosity is, not taking on any other objections to actualism.
David,
I realize that actualists accept, as possibilists do, that he ought to not-rape-and-not-murder. And I never said (nor wrote as if) I thought otherwise. That is, I never said nor wrote as if I thought that actualists accept the monstrous view that it is not the case that he ought to not-rape-and-not-murder. What I find grotesquely implausible is the monstrous view that he ought to rape when he can refrain from rape, refrain from murder, and refrain from doing anything else evil. It’s this view, which actualism entails, that I find grotesquely implausible. If you don’t, then we just have very different intuitions.
John, one last point.
My understanding of the example is that, as you put it, the pedophile still can return the girl unharmed, but that as a matter of fact he will not. (In your last comment you seem to be treating these as contraries: you used to think the first was part of the example but now you think the second is…)
My point about modus ponens depends on the fact that the pedophile will not return the girl unharmed. That’s P, after all (or anyway it’s equivalent to P).
Thanks, Jamie. Yeah, that’s a bit of a slip.
I think one important thing is whether this “might” counterfactual is true: he might still return the girl. If it’s true that he might still return her, then I balk at the claim that it’s (definitely) true that he would/will either rape-and-desist or torture-and-kill.
Ralph,
I happen to agree with J&P that ‘ought’ doesn’t agglomerate over conjunction (it’s a consequence of a probabilistic semantics for ‘ought’: probably A and probably B doesn’t entail probably (A and B)). In the Prof. P case, for example, I think that it’s true both that PP ought to (accept the invitation and write the review), and that he ought to reject the invitation. But I’d deny that he ought to (accept the invitation and write the review and reject the invitation).
Further, like J&P, I think the explanation for this is (in part) that ‘ought’ is contrastive. Out of the options {accept the invitation and write the review, don’t accept the invitation and don’t write the review}, the former is best. Out of the options {accept the invitation, reject the invitation}, the latter is best–given that PP is unlikely to write the review.
So where do I disagree with J&P? In normal contexts (I think), the contrast class for ‘S ought to do A’ is all the things that S could do instead of A. I imagine everybody in this conversation except David would agree, and this is why it’s offensive to say that your agent ought to rape-and-desist. The alternative objection that I offered is thus that J&P are committed to its being possible for there to be multiple mutually exclusive relevant alternative actions that an agent ought to perform. This is objectionable even though ‘ought’ doesn’t agglomerate.
Steve thinks I wouldn’t agree about “normal context.” But why not? Grant that Steve is right, that there is something called a “normal context” in which “S ought to do A” means “…rather than anything else S could do.” That would explain why actualism can be made to look awkward. “He ought to rape her??” If you hear this and assume the normal context you will be misled, thinking that actualists deny that he ought to not harm the girl in any way (which they don’t).
That point about “normal context” would just mean that the question at hand presents an abnormal context: “granting that he ought neither to rape her nor murder her, ought he to rape her if he will otherwise murder her?” That carefully signals that this is not the normal context for “ought” statements. But the question remains.
So, the fact about normal context doesn’t show that the actualist’s view is false, just that it sounds awkward if the wrong context is being assumed…but we knew that.
J&P explain (p. 243ff) that the “overriding” ought (in our example, to not rape and not murder) is from one (larger) set of options. The ought that prescribes the rape is from the smaller set of options in light of the fact that the overriding ought will be violated. (And the two oughts are, therefore, consistent.) This shows that the crucial question does not occur in what Steve calls a normal context.
But this doesn’t support the charge that actualism implies anything offensive. (Steve seems to say it does, since he says this is where he “disagrees with J&P.) It shows only that one will cause offense if one does not clearly specify the applicable context. That’s why it’s jarring–because the normal context will often be erroneously assumed. “Joe ought to do evil” is potentially misleading if it’s stemming from the view that “Joe ought to do the lesser of two evils.” The latter is clearer about the applicable context. But that doesn’t make the former statement false.
Here’s what I think supports J&P. (It may not be sufficient support, but it provides an interesting challenge to those who reject their view.)
It’s hard to deny that
(1) If he’s not going to release her, then he ought to rape her.
At least this sounds correct to me.
But, furthermore, we’re just given that
(2) He’s not going to release her.
And it certainly appears to follow that
(3) He ought to rape her.
(Even modus ponens skeptics like me are not ready to give up such straightforward inferences as this one.)
It’s interesting how similar the logical structure here is to the “If you want… you should…” conditionals discussed in the comments of Uriah Kriegel’s What is a Categorical Imperative? posting here back in March 2006.
David: surely Actualism isn’t supposed to be a theory merely about an abnormal use of ‘ought’? Ralph introduces the view as holding: “for every act-type A, you ought to do A if and only if your conduct would be (in the relevant way) better if you did A than if you did not.” So I take it that Actualism is supposed to be a claim about the assertion conditions for ‘ought’ sentences.
Jamie: I think there’s an ambiguity here. As Ralph introduced Actualism, it’s the view that you ought to do A iff it’s better than what you would actually do otherwise. People seem to be assuming that this is equivalent to: you ought to do A iff it’s the best of the actions that are still psychologically possible for you. But it isn’t equivalent. Suppose that in fact there is only one action that is psychologically possible (e.g. killing the girl). Then it’s true that if the agent didn’t rape-and-release, then he would kill. And it’s true that if he didn’t just release, then he would kill. And it’s true that if he didn’t give her a paper-cut-and-release, then he would kill.
I think it is arguable that Jamie’s (1) is assertable given a special context where there are only two psychologically possible actions: rape-and-desist and torture-and-kill. But if Actualism is true (as defined by Ralph), (1) would be true under much stronger conditions, where torture-and-kill is the only actual psychological possibility. But in that context, I think we should reject (1) as offensive and/or false: for we could also say, ‘If he’s not going to release her [unharmed], then he ought to give her a paper-cut and release her’ (for example). If we’re not limiting ourselves to the psychological possibilities, why pick out the rape option as the best of the psychologically impossible alternatives?
Perhaps Actualism is really supposed to concern only ACTUAL psychological possibilities? But then no ‘ought’ sentence is true if determinism is correct. That seems rather a large assumption for a semantics of ‘ought’ to make.
Steve: You write, “surely Actualism isn’t supposed to be a theory merely about an abnormal use of ‘ought’?” If the charge is that it’s such an abnormal context that the ‘ought’ is simply not assertible in that way (or that it is a very odd and quirky way to speak?), then it has nothing to do with Ralph’s morally extreme example. You’d want to say, then, that “Prof. P. ought to decline the assignment” is not assertible because the required context is too abnormal. Intuitively, though, it’s quite natural to say that he ought to decline once you know the story (even though philosophers reasonably debate it’s truth). The appropriate context is easily set up. The context issue itself should be no more damning in Ralph’s example than in this one. There could be misunderstandings in either case, and they will be more jarring in more extreme examples, but I don’t see how any of this harms actualism.
Ah, I see, Steve. My view (and I think this is also P&J’s view) is that what matters to (the assertibility of) (1) is that we know the paper cut alternative isn’t going to happen. We don’t have to know that alternatives are psychologically impossible, only that they will not happen. And in Ralph’s example we know that if the evil doer doesn’t rape the girl he will kill her.
(I must say that the fact that the example is particularly gruesome seems to me to be very distracting in an unhelpful way.) But anyway we can eliminate the issue of unmentioned alternatives by setting up an artificially constrained case.
Suppose Dr. Evil captures you and locks you in a room where there are two buttons. If you press the red button, you’ll get a nasty electric shock (quite painful but no permanent damage) but the little girl will be released unharmed. If you press the blue button you’ll get a mild shock but the girl will be harmed and then released. If you press no button, Dr. Evil will release you unharmed but the girl will be killed.
I think (4) is true, and also that you ought to press the red button.
David,
My objection wasn’t motivated by Ralph’s particular example, but by the behaviour of ‘ought’ in general. What I meant to say is that Actualism can’t accommodate normal contexts at all.
But how are you understanding Actualism? Is it:
(A1) S ought to do A iff doing A is better than what S would otherwise do? (as Ralph introduces it)
Or is it
(A2) S ought to do A iff doing A is possible for S, and better than what S would otherwise do?
If your answer is (A1), then it seems that Actualism only gives the right results in a very limited context: where we’ve already narrowed the relevant alternatives somehow to {A, what S would otherwise do}. (Otherwise you get the result that it’s true of multiple mutually exclusive actions that S ought to do them, which doesn’t match the behaviour of ‘ought’ in any context). But normal use of ‘ought’ doesn’t work like that, so (A1) can’t accommodate it.
If your answer is (A2), then I grant that there is a fairly common kind of context where we use ‘ought’ like that. But you yourself need there to be contexts where we don’t use ‘ought’ like that, in order to avoid Ralph’s monstrosity objection. For on this view, in the case that the pedophile is incapable of doing anything less bad than raping, it would simply be false to claim that he ought not to rape.
The point is that Actualism (on either reading) is a BI-conditional, so it denies the truth/assertability of ‘S ought to do A’ in any other kind of context. You can’t simply point out that the contexts needed for Actualism to give the right results can be set up.
Jamie,
Thanks (I posted before I saw your post). I agree with you about your Dr. Evil case. But I think the challenge I just raised for David/Actualism stands. The fact that Actualism gives the right results in artificially constrained cases is not that probative.
I grant that if we restrict the alternatives we’re interested in to a certain set, then it’s appropriate to say that S ought to do whichever of that set is best. But that’s not exactly how Actualism has been characterized by anybody here, even with your clarification. Is Actualism really committed to the view that S ought to do A (out of all the remaining alternatives) just in case doing A is better than what S would actually do otherwise? Suppose we’ve restricted the possibilities to {A, B, C, D, E}, which are valued in that order, and that what S is in actual fact going to do is E. Then Actualism entails that S ought to do A, and he ought to do B, C, D.
What I think is correct is that what S ought to do is whichever of the relevant alternatives {A, B, C, D, E} is best. Period. What S would actually do if he didn’t take the best option is irrelevant. All the work here is being done in the restriction of the admitted possibilities, not in any counterfactual about what otherwise would happen. However, I think Actualism is easily conflated with this view. Typically the relevant alternatives are alternatives that we can’t rule out as impossible. But an alternative can be possible without being what would actually happen if not something else.
But perhaps I’m just misunderstanding what Actualism means by ‘what would actually happen’?
Steve,
I’m a little confused about what Actualism officially says. This is because (1) I think some issues have been raised that hadn’t occurred to J&P and so they aren’t distinguishing between certain versions of the general view, and (2) I frankly do not understand the notion of what someone “would actually do”. (What is the ‘actually’ doing there? I am assuming it isn’t pleonastic.)
The main claim of J&P, I think, is that it is what the person would do if she did A, and what she would do otherwise, that matter. This answers two different questions, which they call the evaluation question and the selection question. As far as I can tell, it’s the selection issue that is in play in this discussion thread.
I think the name ‘Actualism’ is unhelpful and misleading. And if I were trying to make the main point that Jackson and Pargetter are making, I think I would frame things with ‘will’s instead of ‘would’s (though now that I think about it, there probably isn’t any significant difference in the implications if we switch — I just find it easier to understand).
I might agree with your view, but I’m not sure because I don’t see how to select the “relevant alternatives” in a context. Suppose we take your abstract example, in which the alternatives for S are {A, B, C, D, E}. (I thought artificially constraining an example was a good way to do this, but since you think not it’s fine with me if we can just stipulate that these are the relevant alternatives.) And let’s suppose that they are ranked in that order, with A the best and E the worst.
Now, what do you think of this conditional?
I think it is true. I imagine you either think it is true or think we need more information about the context; if the latter, what information do we need?
Jamie,
Thanks for the explanation. I should stop being lazy and go and read J&P.
I’m happy with (5)–although I think it can be correct to say ‘(Even) if S doesn’t do A, he ought to do A’, which is a point that I think I recall Jackson making somewhere. (But that’s just because the antecedent clause can function in different ways, so set it aside).
I’ve been understanding the Actualist idea as appealing to what’s metaphysically possible for the world that is metaphysically closest to the world in which the ought-judgment is made in which S doesn’t do A. Whereas I take relevant alternatives to be determined pragmatically by our interests in certain epistemic possibilities (or, in the case of hypothetical scenarios, imagined possibilities). So I’ve been worried by the mismatch, which seemed to open up a possibility on Actualism that it would be correct to say that S ought to do D. (Since E is what would actually happen if S didn’t do D, even though the relevant alternatives we can’t rule out epistemically/ hypothetically include A-C). But if as you suggest I’ve been misled by the label, then maybe I don’t have any remaining objections to actualism.
Steve – you ask, “Is Actualism really committed to the view that S ought to do A (out of all the remaining alternatives) just in case doing A is better than what S would actually do otherwise?”
Just to clarify, this is definitely not what Actualists claim. Here’s a key passage from J+P that I quoted in my follow-up post (linked above):
“There are two matters that need to be borne in mind when considering what ought to be done in terms of the option approach. One is the evaluation of options. We have urged the plausibility of evaluating options in terms of what agents would do were they to adopt them. That is Actualism. The other matter is how to select the right set of options. That depends on exactly which question you want the answer to. If you want the answer for some action as to whether an agent ought to do it, look at the set consisting of the action and what the agent would do instead; if you want the answer as to what an agent ought to do at or during some time, look at all the maximally relevantly specific actions possible at or during that time.”
Your question — what S ought to do out of all the alternatives — is the question Jackson calls “what an agent ought to do at a time”, and the answer is the sensible one of whatever’s best (option A, in your case).
It’s worth separating out Actualism as an normative thesis (which really only comes into view when we consider intertemporal cases like Prof. Procrastinate) from J+P’s semantic proposals regarding the selection question. As it happens, they would say that there’s a true reading of “S ought to do D” in your case, since they’d understand this as raising the question of whether D is the thing to do between the options of D and what the agent would otherwise do. But I don’t think much really hangs on the semantic dispute — in particular, Actualism as a normative thesis doesn’t.
Hi Ralph,
I too was moved to respond to the actualists. It’s possible that you’ll find my “Defending a Possibilist Insight in Consequentialist Thought” (Philosophical Studies 142 (2009): 183-195) interesting. I argue there that standard accounts of subjunctive conditionals fail in many consequentialist contexts. I then propose a possibilist-modified version of the standard account that might resolve the conflict. It’s available on my website: http://web.nmsu.edu/~jvessel/Philosophy/index.html.
J-P
Steve,
I’m having a hard time with this.
Is there an example in which we know that E is what would happen if S didn’t do D, even though we can’t rule out A-C? This seems impossible, though maybe there is some recherche example?
Or, maybe you mean an example in which the speaker doesn’t know that E is what would happen, but we the evaluators of the speaker’s utterance do know?
Also, I have a follow-up question. Since you agree that (5) is true, and obviously its antecedent is true, are you willing to infer the consequent?
Richard,
Thanks. It looks like I’d still want to join Ralph in resisting what J&P say about whether an agent ought to do some action, for the reason that the comparison set at least sometimes should include not only what the agent would do, but also what he could do (perhaps merely in the sense of physical capacity). So take the question, ‘Ought S rape-and-desist?’ Surely that’s a question of the first kind, not the second. But while I grant that we can get J&P’s answer, I also think there’s nothing wrong with answering: ‘No, of course not. Because he ought rather to release her unharmed, although we know that isn’t going to happen.’ As you/Ralph present J&P’s view, they’re committed to denying that.
Jamie,
So I’m puzzled too about what J&P mean by ‘what S would do’. If they merely mean to contrast with all the alternative possibilities not ruled out by the context, then that’s my view too. But that’s not what the subjunctive suggests to me, and further, it suggests to me that J&P have in mind that there is one particular thing that S would do if he didn’t do A–although perhaps this is mistaken. If it’s right, however, it looks like the subjunctive has to be understood as picking out something like a favored determinate world. And that’s a view that I think is definitely wrong.
I’ll agree there aren’t any cases where we know that E would happen if not D, but can’t epistemically rule out A-C. But (1) we might be considering the alternatives hypothetically. E.g.: ‘Ok, we know that we’re actually going to hire E, perversely. But which of the candidates ought we to hire?’ Won’t the Actualist answer(s) be: We ought to hire A; we ought to hire B, we ought to hire C, we ought to hire D? (Whereas I think the only correct answer is clearly that we ought to hire A). Or rather, given Richard’s correction, the answers to all of: ‘Ought we to hire A?’ through ‘Ought we to hire D?’ will be “Yes”. (2) Even if we’re not in a position to say that S ought to hire D (because we don’t know that S is actually going to hire E), on Actualism wouldn’t it still be true?
Your follow-up question: I’d deny modus ponens holds for modal conditionals as a general rule. I follow Kratzer and Lewis in thinking that the antecedents of these conditionals function by restricting the modal base. Whether you can then detach the consequent depends on whether the condition in the antecedent is now restricting the modal base contextually or not. (Or another way to put it: you can’t actually detach the consequent, because probability/modal conditionals don’t detach. But you can assert the consequent without explicit restrictor, so long as the restrictor is now contextually implicit.) I have a paper coming out on this stuff in Synthese, ‘What Ought Probably Means (and Why You Can’t Detach It)’.
Steve,
Re what you say to Richard: If someone asks whether the guy ought to rape the girl, J&P can certainly agree that you will want to answer in a way that makes clear the two different contexts that might be intended–what the comparison set is taken to be. It would be obtuse to say, simply, “yes,” and the view doesn’t force them to. The reason is that it could be so easily misunderstood as denying that he ought to release her unharmed. However, if it is a philosophical question, about an act, the rape, whether it ought to be performed or not, their view is that the correct answer is “yes.”
As to what you said to me earlier, I’m not sure why you say actualism can’t handle normal contexts, which means those where the comparison set is all actions available to the agent. Actualism says that the answer in that case is that he ought to release her unharmed. Sometimes when someone asks, ostensibly, about a particular act they have the normal context in mind, whereas it might seem that the actualist is forced to treat it as the narrower context. But they have a plausible answer for each context: He ought to release her unharmed. He ought to rape her if otherwise he would kill her. But, as I say, I’m not sure I caught your point about a problem with normal contexts.
Much of the resistance here to actualism seems to me to amount to showing that it will sometimes be easily misunderstood as saying something offensive. That, of course, wouldn’t show that it does say anything offensive, which is the question. “He ought to rape her” risks serious misunderstanding and offense. But when the view is made clear, as in “Since, otherwise, he will kill her he ought to rape her, although harming her either way is heinous,” then it is patently not a morally defective view (whatever other pro’s and con’s it might have). It’s a technical view.
Now I’ve read J&P. I remain convinced that they’re wrong. I’ll try to explain, for anybody who is still following. (Sorry for the long post).
As Richard pointed out, J&P say that Actualism is a view about evaluation of options. That bit seems pretty plausible. But (as Jamie observed) it’s only Actualism about selection of options that has been at issue here from the start. So that’s what I’ll use ‘Actualism’ to mean.
J&P defend Actualism against Possibilism. But there’s another alternative: call it Contextualism. I think (1) Contextualism is correct, (2) Contextualism is incompatible with Actualism, and (3) many of the defences being offered here for Actualism really only support Contextualism.
Contextualism is the view that (1) ‘S ought to do A’ always means that S ought to do A given restriction to some set of worlds W determined by context, and (2) that S ought to do A given W iff A is the best of the relevant alternatives in W. According to Contextualism, sometimes W is restricted by certain information about what is actually going to or not going to happen, as on Actualism, and sometimes it is not, as on Possibilism.
Actualism is incompatible with this view because it claims that ‘S ought to do A’ is true if AND ONLY IF doing A is better than what S would otherwise do (p. 247: ‘UNEQUIVOCAL[LY]…For each A, it ought to be done by an agent just if what he would do if he did A is better than what he would do if he did not…’) Further, Actualism claims that this ‘S ought to do A’ is not elliptical for any ‘ought, given…’, but is rather what S ought ‘fullstop’ to do (p. 238). It follows that Actualism is committed to its being simply and unequivocally FALSE (in Ralph’s example) that S ought not (fullstop) to rape. That belies David’s claim that Actualism allows readings on which it is true and readings on which it is false. On Contextualism, however, it is true that S ought not to rape given contexts where W is not limited to worlds where S either rapes or does what he would otherwise do. Since David thinks there are such contexts, he’s really defending Contextualism, not Actualism.
David suggests that Actualism claims simply that S ought to rape, GIVEN that he is either going to rape or kill. But according to J&P, that isn’t Actualism. J&P say that both Actualism and Possibilism can AGREE that [mutatis mutandis] S ought to rape, GIVEN that he is either going to rape or kill, and that S ought not to rape, GIVEN all the things he could possibly do. Actualism, they claim, is not a view about what S ought to do GIVEN THIS OR THAT, but a view about what S ought to do fullstop (p. 238). This gives new life to Ralph’s monstrosity objection: J&P are committed to it being unequivocally true that S ought to rape. (Although given how they would have us interpret it, it isn’t really monstrous so much as incompatible with ordinary use of ‘ought’).
J&P recognize, of course, that it’s bizarre to deny that S ought not to rape. Hence the distinction between (i) whether S ought to do A, and (ii) what S ought to do (at a time), where (ii) is open to merely possible alternatives. But I don’t think this distinction is stable. Surely if S ought to do A, then A is (at least part of) what S ought to do, and vice versa. So we find J&P claiming (m.m.) that S ought to rape, but denying that S ought to rape ‘right now’. But then WHEN ought S to rape? Not at any other time, so apparently S ought to rape at no particular time?? This seems incoherent.
Finally: notice that J&P’s arguments for Actualism consist in showing that Possibilism doesn’t issue in good advice. (If S is going to kill the girl if he doesn’t rape, then you shouldn’t advise him not to rape). But all that shows is that when we’re making ‘ought’ claims to guide people, we should restrict them to sets of worlds that we think still have a chance of being the case. It doesn’t show that in other kinds of contexts (e.g. in post facto criticism of others’ behaviour) we shouldn’t include merely possible options. But questions of the form, ‘Ought S to do A?’ are just as legitimate in these contexts as questions of the form, ‘What ought S to do (at this time)?’ So Actualism is false. It conflates two things: (1) when it’s true that S ought to do A, and (2) when it’s good advice to tell S that he ought to do A.
Anybody still reading…?
I think Jamie put the case nicely. S has 5 possibilities, {A, B, C, D, E}, which are ranked in that order, and we are asked about
(5) If S doesn’t do A, he ought to do B.
Jamie thinks this is true, and it’s certainly intuitive. But now with MP, and so long as S won’t do A, S ought to do B, full stop. We can put the “grotesque and monstrous” objection by saying that, on this scheme, provided S is a perverse enough person who won’t do A, B, C, or D, we can conclude that S ought to do E, which could be arbitrarily bad. In other words, if you are a sufficiently evil person, it can be such that you ought to be evil. Moreover, since you can decide what you will and won’t do, you can make it the case that you ought to do evil simply by deciding not to do anything better. Something seems to have gone wrong there.
The problem, according to me, is with the detaching MP. We can detach the consequent of (5) with “S ought not do A” but not with “S won’t do A.” I don’t think anything monstrous follows from that. Any semantics where this conditional supports detachment is going to lead to the grotesque and monstrous outcome.
Right, Heath, that’s one way to go.
The thing is, it is very hard to understand what a conditional is supposed to mean if it doesn’t support modus ponens ever.
Steve, I’d say the same for your view. There is no other conditional (even according to Kratzer, right?) that doesn’t support modus ponens at all. Conditionals are inference tickets, as a wise man once said. They are tickets to infer the consequent from the antecedent.
Can you say this, seriously, without feeling incoherent:
What about this:
What about this:
Those sound straightforwardly incoherent to me.
Jamie,
Yes, modus ponens is generally fine. Having given it more thought, I think your case for Actualism fails not because modus ponens fails (I grant that we can get the reading that conforms to Actualism), but rather because there is a legitimate reading on which we can reject your (5). That’s what you’re challenging me to do, and I’m happy to bite the bullet. I’m sure you’ll agree it can be appropriate to assert the following [how do you do those indents?]:
“He won’t do A. But still, he ought to.”
The following seems fine too:
“Even if he won’t do A, he ought to do A, not B.” (e.g.: Even if he won’t release her, he ought to release her, not to rape her).
Granted it’s a little strange to deny (5), but there’s a pragmatic explanation for that. As is well known, asserting ‘If X then Y’ implicates that if not X then possibly not Y/ that Y depends on X in some way–an implicature canceled by ‘even’. This implicature makes it natural to read ‘If S won’t do A’ as restricting the set of options for the ‘ought’, as Actualism needs.
So to your challenges: it’s generally bad to use the same word in successive sentences with different parameters/senses, if there’s no cue that there is a shift in sense. But if you include such a cue, then we can say such things coherently. For example,
“If he won’t release her (and we know he won’t), then he ought to rape her. But he REALLY ought not to rape her, of course.”
So back at you: can you seriously disagree with that?
To emphasize a point I’ve made a number of times: J&P/Actualism claims not simply that there is an Actualist reading of ‘S ought to do A’ claims (which I grant), but also that this is the ONLY reading of ‘S ought to do A’ claims. That’s what I’m rejecting.
Steve,
Yes, I agree that those two (“He won’t… But still he ought” and “Even if he won’t, he ought to do A, not B”) are perfectly fine. The second, at least, is fine according to J&P too, right?
I also agree that your Really conditional
is completely acceptable.
So you now want to say, not that (5) is false, but that it is ambiguous with both a false reading and a true one. But you also say now that conditionals generally support modus ponens. So you should agree that there is a reading of the monstrous conclusion that is true. Have I got it right now?
(I don’t understand the point of your implicature story, which I initially thought was supposed to help us understand why (5) is hard to deny even though false. Adding an implicature should make it easier to deny, not harder.)
To make the nice indents, use the blockquote tag.
I’m finding both alternatives highly problematic, and want to suggest a third possibility, or a set of them. Moral obligations are neither functions of what will actually happen, nor would could happen, but what our current evidence suggests will happen, as a result of our various choices. For consequentialists like myself, this is the view called subjective consequentialism, as opposed to objective consequentialism; I think a parallel distinction could be made in deontology but I don’t know if anyone’s systematically discussed this.
Now if we try to reconstruct the example in these terms, the prerequisite for making it remotely plausible that the kidnapper ought to rape the girl is that he has strong evidence that if he doesn’t, then he will kill her, but if he does, he will release her afterwards. It might appear that he could have such evidence, say if in past kidnappings he has done one or the other, and never anything else. But really it takes more than that: he would need evidence that he will come under some compulsion to kill her if he fails to rape her, and that it is not possible for him to simply release her unharmed.
Now I can easily imagine someone telling himself this, as a way of trying to excuse his behavior. It is much harder to imagine someone actually having serious evidence–as opposed to a self-serving belief–that this is the case. He would need evidence that in fact he will lose control of his agency, his power to choose, coming under the grip of unstoppable compulsions, as a result of some choice that he current can make freely.
I think that some of the conundrums encountered here come from too-easy reliance on stipulations of the hypothesis, like “we know that if he doesn’t A, he will B…”. But how can we know this? More to the point, how can the kidnapper know this, or have reliable evidence supporting it? I reject ideal observer theories in part because they lead to counter-intuitive results like the ones being considered here; they leave out the question of whether the agent has reasons to do A or B based on the situation he finds himself in, and attend to the (relevantly irrelevant) question of what an ideal observer has reasons to advise him to do.
The only way I can easily imagine the agent having reliable evidence for the results is something like the Dr. Evil example: the agent is assured by some powerful external force which he cannot control that certain results will come from his choices, and certain other alternatives will become unavailable as a result of some choices. Even then, I think there’s something misleading about saying:
>If you aren’t going to press the red button, you ought to press the blue one.
I would again ask the antecedent to be recast as: “if you have strong evidence that you won’t press the red button” in order to make the consequent plausible. But what could such evidence consist in? If the red button is fenced off or taken away, perhaps. But if it is within your power of choice to press the red button, then it seems to me that you cannot possess strong evidence that you will not press it; there is nothing preventing you from doing so. That’s not to say that you will press it; you might not (say, on a sudden and unpredictable impulse). But without an antecedent specifying evidence that you will not do so prior to making the choice, I don’t find the conditional compelling. If such compelling evidence exists, then I would say yes, you ought to press the blue button, seeing to it that the girl is harmed but not killed, etc. Then this result seems tragic but not paradoxical.
Jamie,
Right, not false but ambiguous. As to whether the Actualist can say those same things–it’s a bit murky because of J&P’s dubious distinction between whether S ought to do A and whether A is what S ought to do.
But they are adamant that in the scenario, there is only one answer to the question, ‘Ought S to rape?’, and that is ‘unequivocally’ YES. Anyone who says ‘It’s false that S ought to rape’ has said something false. That’s Actualism about selection of options, and it’s what I’m rejecting.
Here’s the implicature story in more detail. If you say ‘If X then Y’, then there’s an expectation that you’re not in a position to make the simpler and stronger claim, ‘Y’. It suggests that you think Y depends on X. But in what way could the truth of ‘S ought to rape’ depend on the truth of ‘S isn’t going to release’? The only prima facie plausible answer is that the speaker intends/the antecedent functions to limit the option-set. Hence “If S isn’t going to do A, then he ought to do A” is unacceptable, unlike “Even if S isn’t going to do A, he ought to do A”.
Scott,
Unlike Actualism, Contextualism allows for subjective ‘ought’s, i.e. those constrained by what is known/believed. But surely not all true ‘ought’ claims are subjective. Otherwise it wouldn’t make much sense to say things like ‘I just don’t know enough to be able to tell what I ought to do’, or to ask for advice from better informed people. There is some good stuff on this. Niko Kolodny & John MacFarlane have a very provocative paper (in progress). Judy Thomson has written quite a bit on it, championing a more objective position.
Steve,
I’m still confused about the point of your implicature story.
I agree that “If S isn’t going to do A, then he ought to do A” is unacceptable. Maybe your implicature story is the correct explanation.
But you wrote this:
So I am expecting a pragmatic, implicature story that explains why (5) is hard to deny. To remind you,
—
You know, this example really is a lot like the Gentle Murder paradox.
Steve, I haven’t had time to look closely at your idea of contextualism. So I just want respond to two things falsely (I think) attributed to me in the post on Sept. 16 where you introduced that idea. I can’t (as far as I know) easily search all my posts, but I doubt I said either thing, and in any case they wouldn’t reflect my view. Plus a few other brief comments.
You write “Actualism is committed to its being simply and unequivocally FALSE (in Ralph’s example) that S ought not (fullstop) to rape. That belies David’s claim that Actualism allows readings on which it is true and readings on which it is false.” I hope I didn’t say that, and doubt that I did. I said that Actualism is compatible with the rape’s being a heinous wrong, since Actualism holds that he ought to neither rape nor murder the girl.
And you write, “David suggests that Actualism claims simply that S ought to rape, GIVEN that he is either going to rape or kill.” You don’t quote me, and again, I hope I never said that since I’m aware that this is supposed to be common ground with Possibilism. I do sometimes formulate it this way: “Since he will otherwise murder her he should rape her.” That pointedly goes beyond what is common ground, though, since it is committed to “he should rape her,” unlike “Given…”
And you write, “This gives new life to Ralph’s monstrosity objection: J&P are committed to it being unequivocally true that S ought to rape. (Although given how they would have us interpret it, it isn’t really monstrous so much as incompatible with ordinary use of ‘ought’).” Whether or not I agree that it is incompatible with ordinary use (more on that below), you are rejecting Ralph’s view that it is “really” a morally monstrous view. I’m not sure how that “gives new life” to it.
Finally, on “ordinary uses:” It does not show that a term is being misused just to show that it might well be misunderstood because of erroneous assumptions. I contend that such misunderstandings can be dispelled within ordinary usage by the Actualist as follows: “Since he will otherwise murder her, he ought to rape her instead. So, yes, he ought to rape her.” And if someone says, “you think it’s ok to rape her???” they say, “No, of course not. If he rapes her he will heinously be violating the requirement to neither rape nor murder her.” The fact that a conversation might be required is really no objection. If the opponent says, “But surely you must admit that he ought not to rape her,” the Actualist should say, “Even if otherwise he would kill her?? I do not grant it.” At that point, I believe the Actualist’s non-monstrous view is quite clear and all consistent.
Nothing here is incompatible with ordinary uses of “ought,” so far as I can see. It would certainly not be ordinary to just say “he ought to rape her” without explanation. But that can’t be how we determine proper (or “normal” in the normative sense in question) and improper uses of a word. The explanation one goes on to give, after all, might show that the original statement is true, with no novel way of using of the term, even if was initially shocking. Many statements are predictably misunderstood. That doesn’t show they’re not true on the “ordinary use” of the terms. An example outside of the Actualist/Possibilist context: “All men are not created equal.” Monstrous? False on ordinary usage? “Men are indeed equal but since men weren’t created, all men are not created equal.” Oh, ok then. Nothing monstrous there after all. And yet the original jarring statement is not rescinded, nor has language been abused.
Yikes! A Brown tagteam!
Jamie,
Sorry, more explanation needed. The point of contrasting (1) ‘If S isn’t going to do A, then he ought to do A’ (unacceptable) with (2) ‘Even if S isn’t going to do A, he ought to do A’ (acceptable) was to show that there is something about the antecedent of (1)that restricts acceptable readings of the ‘ought’ to options that exclude S’s doing A–which therefore makes it hard to deny (5) ‘If S isn’t going to do A, then he ought to do B’. However, it can’t just be the hypothesized fact that S isn’t going to do A, because then (2) would be unacceptable as well, which it isn’t. So what’s the difference between ‘If…’ and ‘Even if…’? I think it’s just that ‘If…then…’ implicates ‘If not…then possibly not…’, while ‘Even if…’ blocks that implicature. If that’s right, then the reason why (1) is unacceptable and (5) is hard to resist is that ‘If S isn’t going to do A, then S ought to do B’ implicates that ‘S ought to do B’ is somehow made true by it’s being the case that S isn’t going to do A. And the only plausible sense in which that is true is if we are restricting the options to exclude A. So the natural reading of an utterance of (5) is one on which the speaker intends the antecedent to function by restricting the relevant options, in which case (5) is true. To read it any other way it is so bizarrely false that it’s hard to get that reading.
Maybe this story isn’t right, but hopefully it’s clearer now?
David,
I apologize if I’ve mischaracterized your position! I’ll try to respond soon, when I get a chance.
Hm, it’s hard to follow.
I get particularly muddled when I try to get from the part that comes before “If that’s right” to the part that comes after.
It doesn’t seem quite right to me that “If A, B” implicates “If not A, possibly not B”. Rather, it seems to me that “Even if A, B” implicates “B”. But these are too far off, so maybe I’m getting confused and your way is more accurate.
But now how does it follow that “If A, B” implicates that B is made true by A? (I assume you meant the implicature to be general, and not specific to (5).) I am pretty sure there is no such general implicature; still, maybe I’m missing something and you’ve given an argument that there is.
Hm, now I’m thinking you didn’t mean the implicature to be that A makes B true, but something looser or broader than that (maybe you just meant Adams’ rule?).
I think I’m going to ask David Christensen to take over for me now, because I’m getting a little tired.
Jamie,
I hope it’s the hour that’s making you tired, and not my stubbornness!
I’m not sure what Adams’ rule is, at least by this name, but it was loose of me to say ‘makes’. Perhaps better: ‘If A, then B’ implicates that the speaker’s confidence that B is contingent on A.
David,
(1) On the ‘monstrosity’ objection: I concede that the view isn’t monstrous, since (as you and Richard have pointed out) it’s a technical view–i.e. given how we are to interpret these ‘ought’s, there’s nothing monstrous about them. But I think something like the objection still stands. That is: Actualism is committed to making claims that SOUND monstrous, given how ‘ought’ is actually used–which means that Actualism is not a correct theory of ‘ought’. (I think you’ve suggested above that you’re only interested here in whether the monstrosity objection will stick. Maybe we’re talking past each other, because I’m interested rather in whether Actualism could be correct).
(2) I completely agree that “It does not show that a term is being misused just to show that it might well be misunderstood because of erroneous assumptions”, and I agree that ‘S ought to rape’ can be acceptable once the right context is set up. My objection is rather that Actualism claims that an acceptable use of ‘ought’ is actually unacceptable. You get to the crux of it here:
(X) “If the opponent says, “But surely you must admit that he ought not to rape her,” the Actualist should say, “Even if otherwise he would kill her?? I do not grant it.””
I think that this is indeed what the Actualist is committed to, and that here he shuts off a legitimate option: of saying, ‘Yes, I agree that he ought not (really!) to rape her…but out of killing and raping, he ought to rape’. What I can’t figure out is how this is consistent with what you said above:
(Y) “Grant that Steve is right, that there is something called a “normal context” in which “S ought to do A” means “…rather than anything else S could do.” That would explain why actualism can be made to look awkward.”
What’s been bothering me is that (it seems to me) if ‘S ought to do A’ sometimes means that, then Actualism doesn’t cover all such claims. But J&P call it ‘unequivocal’ that “For each A, it ought to be done by an agent just if what he would do if he did A is better than what he would do if he did not…’ (p. 247). Can you explain how (X) and (Y) are consistent? (Perhaps what you meant in (Y) is that this ‘normal context’ is mistaken for but doesn’t count as a case of ought-fullstop, where ought-fullstop is always Actualist?)
Steve, I think a purely subjective consequentialism can account for claims like ‘I just don’t know enough to be able to tell what I ought to do’, or the need to ask for advice from better informed people. The second is straightforwardly justified by the fact that we often have (subjectively available) evidence that the likelihood of our current evidence, or our action-guiding beliefs, which may or may not properly correspond to the evidence, is wrong, and could lead to harm if acted on, while the cost of gathering more information (like seeking advice from better-placed persons) is relatively low. And the fact that we have such subjective evidence is an objective fact about us. This is why I relativize our obligations to evidence, not to beliefs; the latter could be made up as we go along, but the former is not entirely up to us. And there’s nothing paradoxical about having evidence that one’s current evidence about P is inadequate for committing to some action; we have precisely such evidence in many cases, like when I think I was told that I need to pick up somebody coming in from Atlanta at 3, but it could have been 4, and I know I could potentially save myself a lot of trouble by going online to look at the flights.
I take the first statement to be a more roundabout way of saying roughly the same thing, though of course you could also that in cases where there is no time/opportunity to gather more information. In that case, I take it to mean that your confidence that you will not cause harm is not high and you hence have grounds for being uncomfortable with some imminent choice. My take on that is that being hesitant and uncomfortable with your decision is often productive and morally good, for it can motivate you to search for more information which might help; but if you really can’t get it, and act on uncertain information, you are actually less blameworthy if a “mistake” is made that the phrase suggests.
Thanks for the Thomson reference, I’ll look up what she says on this.
Scott,
I agree that a subjective view can be flexible. But I think that what you say here isn’t really compatible with the view. If ‘I ought to do A’ means something like, ‘A is the best thing to do given my evidence’, then how can ‘I don’t know whether A is what I ought to do’ mean ‘I don’t know whether A is what is really best’?
Note too that if ‘I ought to do A’ means ‘A is the best thing to do given my evidence’, then the right thing to say in your airport example would not be ‘I don’t know enough to be able to tell what I ought to do’ but rather ‘I ought to seek more information’. (To be clear, I’m not denying that the second statement is ok, only that subjectivism can account for the former, which is also ok).
Steve, I wouldn’t take “I don’t know whether A is what I ought to do” quite literally, or attempt to preserve its apparent conceptual structure, which I think is just slightly confused, though in practice people often enough derive correct implications from it, namely that one should think harder or gather information (if there’s time for this). You are correct about the second two claims; I didn’t say that the airport situation gave me insufficient info to commit to any action, but to “some”, by which I meant some specific action, i.e. leaving at time X. I do have enough information to commit to information-gathering action.
However you may also be making the point that “I don’t know what to do” doesn’t always mean “I ought to gather more information.” This is true; sometimes when one must choose between X, Y, etc. it is unclear whether it would be more useful to gather more information, think harder about the information you have, or just commit to X or Y etc. and hope for the best. I don’t want to rule out ambiguity, but I think that to the extent that one’s evidence about what is best is truly ambiguous (we’re not just pretending it is out of self-serving convenience or laziness), then the ambiguously-matched options really are approximately equally right, and we ought, then, to choose between them. One of them may later turn out to have been truly leading to the best consequences, but given what we knew at the time, there was no truth of the matter as to which one option out of those evenly matched in subjective evidence of their goodness was the one we ought to have chosen.