Consider the following scenario:
MINERS: 10 miners are trapped in a flooding mine; they are either all in shaft A or all in shaft B. Given our information, each location is equally likely. We have just enough sandbags to block one shaft, saving all the miners, if they are in the blocked shaft, but killing them all if they are in the other. If we do nothing, the water will distribute between the two shafts, killing only the one miner positioned lowest. On the basis of these considerations, (1) seems true:
(1) We ought to block neither shaft.
While deliberating, though, we accept both
(2) If the miners are in A, we ought to block A
and
(3) If the miners are in B, we ought to block B.
We also accept
(4) Either the miners are in A or they are in B.
And (2)-(4) seems to entail
(5) Either we ought to block A or we ought to block B.
Paradox!
In a forthcoming paper, (“’If’s and ‘Ought’s,” JPhil), Niko Kolodny and John MacFarlane argue that the best way to resolve this paradox is to give up modus ponens. Instead, I’ll argue, acceptance of the contextualist semantics for modal expressions I advocate (in “A Flexibly Contextualist Account of Epistemic Modals” and “A Flexible, Contextualist Account of ‘Ought’”, http://www.unl.edu/philosop/people/faculty/dowell/dowell.shtml), together with a Kratzer-style semantics for the indicative conditional, allows for a resolution of the paradox without giving up on MP. This seems to me a clear advantage.
Before getting to my solution, I’ll first briefly summarize the semantic theory my solution presupposes (leaving out some whistles and bells that aren’t required at least for an initial statement of that solution) and the basic Kratzerian semantics for the indicative conditional. On the contextualist semantics I favor, modals may be semantically represented as quantifiers over possibilities. In most uses, including deontic ones, the domains for modal quantifiers are restricted, where the restriction, when not explicit in the linguistic material, is determined by context. The resulting worlds are then ranked, often by a standard that is contextually selected. My view of deontic modals differs from other views in that I recognize two different sorts of domain restrictors, circumstances and information. So, roughly speaking, “ought ɸ” is true whenever all of the best circumstantially alike worlds are ɸ-worlds or whenever all of the best worlds compatible with some body of information are ɸ-worlds.
On Kratzer’s account of indicative conditionals, the antecedent serves to restrict the domain of a covert modal in the consequent. VERY roughly, we can think of ‘if P, then Q” as saying “all the P-worlds are Q-worlds”. That’s only rough because the covert modal’s domain can be further restricted by context; the antecedent will typically not be the only restrictor. (So, we can get, for example: the P-worlds compatible with the circumstances are all Q-worlds”. Ex: “if the summer is unusually rainy, we’ll get mosquitoes” says, roughly, “all the worlds that are circumstantially alike with respect to the ‘local’ circumstances and in which it’s unusually rainy are worlds in which we get mosquitoes”.) Also, the antecedent may provide a clue as to whether circumstances or information provides the additional restriction. Finally, just as the use of an explicit restrictor phrase, such as ‘given what I know’ can make the restriction explicit, the appearance of a modal, such as ‘ought’, in the consequent can make the covert modal overt.
The solution to the paradox, then, lies is seeing that the domain in (1) must be restricted by information; after all, given the circumstances, the best worlds are all the worlds in which we save all the miners and we know that doing nothing won’t save all the miners. So, if the restriction is given by actual circumstances, (1) comes out false. So, when we hear it as true, it must be under an epistemic reading. (In other words, MINERS is a so-called Jackson case and so forces the informational reading. For a detailed contextualist treatment of Jackson cases, see my “A Flexible, Contextualist Account of ‘Ought’,” http://www.unl.edu/philosop/people/faculty/dowell/dowell.shtml.) So, (1) is true, roughly, just in case in all the best worlds compatible with our information, we do nothing.
What about (5)? (5) is compatible with (1) when each disjunct gets a circumstantial reading. It then is true just in case, either, given the circumstances, the best worlds are shaft A-blocking worlds or, given the circumstances, the best worlds shaft B-blocking worlds. But why should we think that the disjuncts have circumstantial modal bases?
There are at least two reasons to think they do. The first is: If they get informational restrictions, then (5) is false, since each disjunct is. But (5) seems true. When we hear (5) as true, it can’t be under an informational reading. Given that the empirically best supported semantic theory of modals recognizes only informational and circumstantial modal bases, the modal base for each disjunct must be circumstantial. The second reason is that the grounds for a modal claim may serve as a clue as to how it’s best understood. One such clue is (4) which is an exhaustive description of possible circumstances; either circumstances are such that all the miners are in A or they are such that they are all in B. If these disjuncts are to support MP with (2) and (3), the modals in the consequents must get circumstantial readings. Since the total grounds for (5) are (2)-(4), each disjunct must get a circumstantial reading. Since (1) gets an informational reading and (5) a circumstantial one, they are compatible. So, all is as it appears in the MINERS scenario; (1) and (5) are both true and (2)-(4) provide conclusive grounds for (5).
This is a really interesting line of thought, Janice!
I hope Jamie Dreier will weigh in on this, because K&M’s solution to the Miner puzzle reminds me of his treatment of the King Henry puzzle. I wonder if your way of handling the Miners can equally well handle King Henry.
Just curious, do K&M (or you) consider a reaction to the puzzle which singles out not MP, but rather the permissibility of practically reasoning in conformity with MP (in all contexts)? I ask this because in the epistemology literature a similar kind of puzzle arises for deductive closure principles in the face of formidable skeptical hypotheses. For example, where ‘O’ names any ordinary proposition we ordinarily take ourselves to know, and ‘H’ names a skeptical hypothesis that entails ~O, many people intuit that the following (Moorean) argument is not so good:
1. I know that O
2. If I know that O, then I know that ~H
C. So I know that ~H
The reaction to this sort of argument hasn’t been to reject MP, but to question whether knowledge is closed under known entailment — that is, whether it’s always permissible or effective to reason in conformity with deductively valid inference rules, such as MP. I figured an analogous move might be worth considering here.
Ugh, I’m a little too busy now to really weigh in. I’ll weigh in lightly.
Jan and I have discussed King Henry, and also the allied Lowlands Livestock paradox. I am happy with the idea that the ‘better than’ in the Henry problem is a modal, and (for these purposes) with the Kratzer semantics applied to the conditionals with the ‘better than’ apodoses. I do not agree that modus ponens can be thereby saved. I believe, rather, that the general approach provides one plausible explanation of why modus ponens fails.
Sorry this is so brief and rather enigmatic.
Hey Janice,
I’m glad you posted on this, I’ve been hoping to get a chance to read your paper for some time. I had two quick questions about your view.
On (2) and (3), these don’t allow strengthening on your view, right? (Otherwise, (2) entails “If the miners are in A but all the evidence indicates that they are in B, we ought to block A” and that’s not intuitively obvious.) I wonder if this is part of the reason why someone might think modus ponens fails here. (Maybe this has been resolved, but I thought that there was some debate about indicative conditionals and modus ponens.)
On (1), could you say more? You wrote, “So, (1) is true, roughly, just in case in all the best worlds compatible with our information, we do nothing.” I feel like I’m missing something. Doesn’t blocking neither shaft lead to a suboptimal outcomes in every world consistent with what we know?
Interesting post, Janice! Although I do have some quibbles.
“So, (1) is true, roughly, just in case in all the best worlds compatible with our information, we do nothing.”
Then I think (1) is false (perhaps what follows is what Clayton has in mind?). Worlds that are best in an unrestricted domain will remain best in a restricted domain of quantification (and the worlds that are best, sans phrase, are those in which we save everyone, as I think you agree?). (For the truly geeky: this is something like a corollary of Amartya Sen’s Principle Alpha, which says basically that the world ranking of chess players cannot rank an American #1 without ranking that person #1 in America.)
The source of the problem here is that the standard Kratzer treatment entails denial of “Serious Information Dependence” (in the sense of Kolodny and MacFarlane, p21). The only way around this for Kratzer, as far as I can tell, is to stipulate readings of modals on which the ranking itself depends on salient information. So I’m a bit puzzled by K&M’s focus on modus ponens: Kratzer’s semantics itself fails to validate modus ponens, but still generates the problematic entailments. (If you’re interested, I have a fairly technical discussion of these issues in “What we know and what to do”, linked on my webpage.)
You say that your account preserves modus ponens. But, so far as I know, Kratzer’s semantics for conditional “ought”s fails to validate modus ponens. Did you have some work-around in mind for this?
Best,
Nate
Look away for a second and the comments pile up! I think I can do two of these before I need to walk the dog…
John,
I have also been thinking about Jamie’s Henry case. I’m not sure what I want to say about it exactly, but I don’t think it can get the same treatment as MINERS. In the Henry case, the modals all seem to be best understood as circumstantial. Right now, my hypothesis is that there’s a change in ordering source from premises to conclusion, so it’s a case of equivocation.
I don’t know of anyone that suggests a solution to MINERS parallel to the solution to lottery paradoxes you mention. I’d have my doubts about whether such a strategy could work, if I’m understanding the move you have in mind. The scenario is stipulated in such a way that we’re certain of (4) (where that’s to be read as they’re either all in A or all in B). Once I’m certain of that and I’m certain that my sandbags can block exactly one shaft (also stipulated), why wouldn’t I be able to draw the inference from (2)-(4) to (5)? After all, the scenario leaves me certain of (2) and (3).
About the issue between Jamie and I…he may be right that MP can’t be saved. But I don’t think the MINERS case shows we have to give it up. And I’m pretty sure that the Henry case doesn’t either. And I think an account that didn’t force us to give it up is preferable to one that does, other things equal.
I do not see the paradox. I think there is a hidden epistemic condition in 2 and 3, namely “if we know that….” Because we do not know where the miners are, the ‘logic’ of the relationship between stated in 2, 3, and them being in A or B and 5 seems to be of little value in deciding what to do.
janice,
i share john’s worry. though i don’t think there is a hidden epistemic condition in 2 and 3, i think there is an objective and subjective reading of ‘ought’ that yields an equivocation.
if we read 1 subjectively, then 2 and 3 are false.
if we read 1 objectively, then 1 is false.
either way, if we read ‘ought’ the same throughout, there is a clearly false claim.
i know this is incredibly simple-minded, but i can’t see why this response isn’t successful.
Thanks for the comments, everybody!
Some general background adding some whistles and bells: In my “A Flexible Contextualist Account of ‘Ought'”, I argue that the basic, Kratzer framework, which only recognizes circumstantial modal bases for ‘ought’s, requires informational bases to handle Jackson cases. This is neutral between Kolodny/MacFarlane and myself. Also, to handle the worry that Nate and Clayton seem to share and Nate’s worry about serious information dependence, we need probabilistic information to play a role in determining which worlds are ‘best’. (Kratzer already allows *something* like this for comparative informational modals like “x is more likely than y”.) So, that’s what happens when ‘ought’s require informational modal bases. So, the standards selected in those contexts must be some sort of amalgam of probabilistic information and what we ordinarily think of as a standard, namely, a list of features we look for in a world in order to determine its rank relative to the other worlds, e.g. the number of lives saved or whether the agent in that world is acting on a maxim she can at the same time will become a universal law. (For details, see section 3.2 in the ‘Ought’s paper, esp. pp.17-19.) It sounds like Nate and I are on the same page here. Yay! I’ll try to have a look at your paper, Nate, before finishing the one in which I’ll be discussing the MINERS example and Jamie’s example (to be given at Mad Meta in Sept.) I’m not sure why you say what you do about Kratzer’s account and modus ponens. Is this discussed in your paper? We may have slightly different views of what it would take to validate modus ponens.
Clayton,
To my ear, adding “all the evidence indicates they are in B” forces an informational reading of the modal, or, at least it does when I hear it as false. (I’ve got a little Judy Thompson in my head saying, “well, if they’re IN A, you STILL ought to block A!”) Since I hear (2) as circumstantial, your sentence doesn’t follow from it. (And I wouldn’t have thought that’s to deny strenghening.)
John,
I’m not following you. Why do you think there is a hidden epistemic condition in (2) and (3)? Do you think there is no circumstantial reading? If so, why not?
Christian,
It sounds like the view you’re tempted by is close to the solution I’m pushing. On the view I defend in my “A Flexible Contextualist Account of ‘Ought'” (section 3.2), the so-called ‘subjective’ ‘ought’s correspond to those with informational modal bases, while the so-called ‘objective’ ones correspond to those with circumstantial bases. This has the important advantage of not positing amibiguity (so the ‘ought’ makes a single contribution to determining a proposition in all contexts in which is occurs). (Kolodny and MacFarlane have a really nice discussion of the advantages of contextualism and relativism over ambiguity theory in their “‘Ought’s: Between Subjective and Objective” paper. You need to ask them to get a copy, but they seem pretty happy to share it with those who do.)
janice,
This has the important advantage of not positing amibiguity (so the ‘ought’ makes a single contribution to determining a proposition in all contexts in which is occurs).
my initial reaction is to say that this is not an advantage, but a disadvantage. initially, ‘ought’ just strikes me as ambiguous. it strikes me as ambiguous in the same way that ‘wrong’ seems to be ambiguous. for suppose that the miners are in shaft A. there’s a sense in which it would be wrong to block off shaft B. that is, given that the miners are in shaft A, this will not prevent their deaths and one ought to prevent their deaths. however, in another sense, given what one has to go on, it would not be wrong to block off shaft A and B (and hence B) since this action would have the highest expected value.
i’m not sure whether this is consistent with thinking that ‘ought’ “makes a single contribution to determining a proposition in all contexts in which is occurs”. but i’m also unsure whether, were ‘ought’ to make such a contribution, this would be an advantage of a semantics of ‘ought’ that would entail it does make such a contribution.
anyway, i’ll check out the K & M paper. thanks for the reference.
Janice,
Thanks, that’s helpful. I’ll need to get into the guts of the paper.
Christian,
What is the subjective “ought”?
I know this question sounds silly and I’m not trying to be thick, but when people start distinguishing the subjective from objective “ought”, I quickly get lost. Two accounts of subjective “ought” that don’t seem to work:
1. You ought-subjective to A if you believe it is best-objective to A.
2. You ought-subjective to A if you believe you ought-objective to A.
These won’t work because you don’t believe you ought-objective to block neither shaft (unless you deny that (5) is true on its objective reading) and you don’t believe it is best-objective to block neither shaft.
clayton,
i’m not sure. but i was thinking of something in the neighborhood of:
one ought subjectively to do x iff the expected value of doing x is greater than the expected value of doing something else.
in the case above, the expected value of blocking neither shaft is greater than the expected value of doing something else. so, we ought subjectively to block neither shaft. so, 1 is true on the reading of ‘ought’ that is subjective.
i don’t have anything interesting to say about expected value. i’m simply thinking of a probability distribution over exclusive and exhaustive states together with an assignment of values to those states.
Christian,
Keep in mind that I am thinking of giving a semantics for ‘ought’ as it functions in English. That, other things equal, a semantics that posits less ambiguity is preferable to one that posits more is a well-established methodological principle among linguists and for good reason. (Think of the advantages a more compositional semantic theory has over a less compositional one and you’ll see why: an ambiguous languge is more complex and hence harder to learn than a simpler one. Yet you don’t have to be too clever or too old to master many natural languages, English among them. A theory that makes the language more simple is better able to explain why than a more complex one.)
And if your reason for positing the ambiguity is that you have the sense that there are both uses, that can be captured by recognizing the two different modal bases. So, again, giving a unified semantics is an advantage of a semantics for ‘ought’ in English. (If you still don’t agree, I suggest again reading the relevant section of the K and M paper…or taking it up with the linguists! 🙂 )
Clayton,
The way K&M cast the view, the subjectivist holds that “S ought to A” is true just in case A-ing is the best option open to S, given S’s information. That can’t be the full story since you’d still want to know when that body of information is determined. When S has to act, if she is to act at all? When “S ought to A” is uttered? We’d also need to settle on what makes something ‘best’, since there seem to be prudential and moral uses of ‘ought’. A contextualist can do this without positing further ambiguity (as I argue in 3.2), but I don’t see how a subjectivist can avoid that.
janice,
the linguistics is beyond me. i’ll check out the paper.
i was thinking that ‘ought’ is ambiguous though, not because i have a favored semantics for evaluative terms in English (or anything like that) that entails its ambiguity, but rather, because i think intuitively (1) and (5) are jointly true. so i guess i think of the semantics for a term as being constrained by our intuitions to apply the concept/s that correspond to it. And i think, intuitively, that i’m disposed to accept (1) and (5), and doubt that i’m special, is evidence that there are different concepts that correspond to ‘ought’ as it occurs in each.
and it just seems clear to me that there is some sense of ‘ought’ that is information sensitive, in some way or other, and another sense, that just isn’t. i suspect that there is bit more to ambiguity than that. but it sounds like you have a view that is compatible with much of this. i’ll check it out.
thanks.
I just had a question about the second argument for why (5) is circumstantial. Could you say a bit more about why this:
“If these disjuncts are to support MP with (2) and (3), the modals in the consequents must get circumstantial readings.”
is true. Why wouldn’t the disjuncts support MP even when the consequents were given an informational reading?
I guess my own view is that there’s only one objective ought which is possible evidence related. On that view, (2), (3), and (5) are false.
Janice
“I’m not following you. Why do you think there is a hidden epistemic condition in (2) and (3)? Do you think there is no circumstantial reading? If so, why not?”
I was probably trying to be overly clever here, but my intent was to distinguish between a reading of 2 and 3 that can be helpful in helping us decide what to do rather then simply reading 2 and 3 as stating the logical implications of the situation. The question that needs to be answered before we can decide what to do is, are the miners in A or B. So when I originally read 2 and 3 I simply added the phrase ‘we know’ to the if parts so I read it as ‘If we know that the miners are in A we ought to block A’, etc. It seems clear that if we know which shaft they are in then we ought to block that shaft, but because we do not know which shaft they are in we ought not to block any shaft.
If it were the case that no miners are in either shaft, 2 and 3 would still be stipulating a sound conditional relationship. Because of this, knowing the logical implications (the circumstantial reading?) of 2-5 is not relevant in how we make our decision to block A or B, or not block either when we know that there are miners in one of the shafts but do not know which one.
Also, and this is more of an informational request on my part, but I thought a paradox involved being logically committed to having to do two (or more) actions when only one can be done at a time. If this is so then 2-5 do not constitute a paradox.
Christian,
Thanks, it looks like your subjective ought is something Zimmerman’s prospective ought.
“It sounds like Nate and I are on the same page here.”
Yeah, pretty much. Although I *don’t* want to commit to calling my account a contextualist account (not that you’ve suggested that I’m committed this way — I just want to be clear).
Deontic modals, on my view, are doubly sensitive to information (in both their modal bases and the orderings on worlds with respect to which their domains of quantification are defined). But that information can come from context, some parameter of the index (e.g., an assessor), or even syntax. So, for instance, I think the “true” reading of (2) is gotten by giving it a semantics on which its “if”-clause shifts the information with respect to which the consequent is interpreted (i.e., by adding the information that the miners are in A), thereby allowing the ordering to rank worlds in which all miners are saved more highly than others. Basically, if I is the information with respect to which (2) is interpreted, then:
‘If the miners are in A, we ought to block A’ is true wrt I iff ‘we ought to block A’ is true wrt I-incremented-by-the-information-that-the-miners-are-in-A.
(Cf. Thony Gillies’ “Iffiness” semantics for indicatives.) The reason I mention this has to do with modus ponens. If the way that you predict a true reading of (2) is by allowing the “if”-clause to shift the information with respect to which the consequent is interpreted, that doesn’t settle the question of whether this sort of shifted interpretation is available *wrt the actually relevant information* (which does not, after all, contain the information about the miners’ location). And I tend to think that it’s not available, since I’ve got a strong intuition that (5), as interpreted wrt the actually relevant information, is false. Assuming the validity of disjunction introduction and elimination, that’s incompatible with the validity of modus ponens.
Thanks, Nate. That’s helpful. I need to read Thony’s paper, but I think I get the basic idea. I guess I think that, as for any context-sensitive term, whether we’ve got equivocation or a violation of MP depends upon how we translate the English into the formal system. So what you call violations of MP, I’m now inclined to call equivocation.
About your view of deontic modals: You don’t think there are any with circumstantial modal bases? I’d be interested to know why. What about “There ought to be less famine than there is”?
John,
I suppose we’re all free to mean what we like by “paradox”, but I had in mind a fairly standard view on which you have a paradox whenever you accept each claim in a set of (at least apparently) inconsistent claims. I imagine that’s why K&M call their puzzle a ‘paradox’.
Janice
I was looking at the proposed paradox as being in 2-5, not 1-5 as I should have. My mistake (and a stupid one at that)!
I look forward to reading your papers.
Hi, Jussi,
Sorry to be so slow responding; somehow I missed your post. Let me ask you a background question: pretheoretically, do any of (2), (3), or (5) seem true to you? Or does it seem to you that you can imagine (2) and (3) being offered and accepted in the course of a natural, deliberative conversation by those in the scenario? I haven’t taken a poll, but my guess is they’d all strike most people pretheoretically as true.
Hi Janice,
no worries at all. I’d still be interested to hear more about the second argument you give in the end.
Anyway, well, I don’t have much pretheoretical intuitions or at least no introspective access to which of my intuitions are not tainted by theory. Sorry about this. The question about natural conversations is interesting though, and I was just thinking about this. The following discussion seems natural to me.
Donald: “We ought to block neither one of the shafts”.
Mickey: “You’re simply wrong. We ought to either block A or we ought to block B. This is because, if the miners are in A, we ought to block B, and if the miners are in B, we ought to block A, and the miners are in either A or B. This is just a matter of logic”.
To me both utterances seem natural and my pretheoretical intuition is that, whatever else is going on, Mickey disagrees with Donald. For that to be the case, Mickey’s utterance has to negate the proposition put forward by Donald. This means that they must be making the utterance in the same context if we accept contextualism. This furthermore entails that there will be contexts in which the paradox arises even if we accept contextualism. So, all contextualism does is to get rid of is some paradoxes but not all. So, sometimes (1) and (5) might be uttered in different contexts but no reason has been given why they could not be uttered in the same context.
In principle, I’m sceptical about the idea that the words uttered in (1) and (5) could determine as such that uttering these words can only be done in different contexts. Other than that it’s hard to see what else could force us to think that they always must be uttered in different contexts. So, I think I’m still thinking that whatever advantages contextualism has getting rid of this paradox all they way is not amongst them.
I agree with John Alexander’s suggestion that (2) and (3) have implicit epistemic qualifications on them to make them true to be the most plausible approach, and am not sure exactly how the contextualist analysis is supposed to be better, or even different really.
It seems to me that Janice implicitly says that (1) has such a qualification, when she says that we can only find it true “under an epistemic reading”. I take this to just mean that what is really true is that “If we have evidence E about the miners, then we ought to block neither shaft–and we have evidence E, so we ought to block neither shaft” where E is the evidence described in MINERS. I would just add similar qualifications to (2) and (3), making them come out true under different evidence conditions. I think this is implicitly what they mean if read as moral judgments, and anyone who insists, like Mickey in Jussi’s post, that No, he really believes that we are morally obligated to block A just if the miners are in A, and we ought to do so now even if we have no evidence about what shaft the miners are in, is simply confusing two senses of “ought”–see below.
I think this is even clearer if we analyze moral ought-judgments, as I think we should, as second-order approvals of norms of responsiveness to evidence, to be evaluated by assessing the total harm and good done by agents following such response norms across all possible worlds in which they could do so. This forces us to clarify what we are really doing as intentional agents. For we never just “block A”–in every circumstance in which we do this, we block A in response to the evidence we are confronted with [one might say “I’ll always block A when I have the chance, whether there are even any miners around or any dangers”–but this is just a trivial form of a response norm, and is obviously one we must rationally disapprove of]. Blocking A in response to the actual fact of the miners being in A [apart from our evidence regarding the same] is simply not an action we can perform, and does not describe a response norm which we could approve or disapprove of.
This can take care of the King Henry example also. “Valuing today the knights’ non-arrival if we win/lose tomorrow” is not a response norm, for the action of valuing the knights’ arrival cannot be made in response to evidence of an event which has not yet occurred. The relevant response norm we can consider today is K: “value the knights’ non-arrival when anticipating an uncertain battle tomorrow”, and we have much reason to disapprove of such a norm. After the battle we may have reason to value part of the result of their non-arrival, but we will still have no reason to approve of anyone following K, including our past selves. This is because another likely part of the result of their non-arrival–losing the battle–is one we greatly disapprove of.
My view is also captured by the “subjective” side of the objective-subjective “ought” distinction. I don’t like that language so much because I think it misleading if we thought that the objective ought is still a moral one in any important sense. For we use “ought” ambiguously; sometimes in the moral sense of approving of a response norm and describing our obligations, other times simply to describe the most favorable possible outcome of some situation. The latter is not a moral judgment, IMO, and both apparent paradox and confusion result from treating it as one.
Janice
I have been looking at 1 – 5 to see if I can find the paradox between 1 and 5. On the surface there appears to be one, but consider the following:
We can restructure 2-5 as:
(4) Either the miners are in A or they are in B.
(2) If the miners are in A, we ought to block A
(3) If the miners are in A, we ought to block A
(5)Therefore either we ought to block A or we ought to block B
(1) is the result of this MP:
P1) If we do not know which shaft the miners are in, we ought to block neither shaft.
(P2) We do not know which shaft the miners are in
(1) Therefore, we ought to block neither shaft.
(1) is the result of what we know about the situation of the miners and is not epistemically dependent on our acceptance of (5. (1) does not state what we logically ought to do but what we epistemically know we ought to do whereas (5)is the logical, not epistemic, consequence of 2-4 and has no relevant epistemic status because we have not removed the ‘if’ conditional in 2 and 3. If we know that (1) is true then how are we motivated to consider 2 – 5 in making a decision that is already made?
What exactly am I missing?
Hi Janice,
A small point about the argument form as stated. It shares this form:
1′) I ought not kill my rich aunt
2′) If I kill her with a knife, I ought to do so with a sharp large knife
3′) If I kill her with a hatchet, I ought to do so with a sharp large hatchet
4′) either I will kill her with a knife or a hatchet
5′) So either i ought to kill my rich aunt with a sharp large knife or I ought to do so with a sharp large hatchet.
Many would deny this is valid, and classic make-the-best-of-it semantics for iffy “ought”s invalidates this form of argument, along with factual detachment. It is best to side step this issue by just adding unalterability qualifiers in front of each of your disjuncts in 4), since in the MINERS context, one of the disjuncts is unalterably true (unlike perhaps either of those in my 4′) even if one of them is true).
John,
There is a contradiction between (1) and (5) because you actually have removed the “if” clause of (2) and (3) by reaching (5), using proof by dilemma.
You contrast the “logical” consequence of (5) with the “epistemic” consequence (1). You seem to be implying that the “epistemic” consequence somehow is stronger than the merely “logical” one–and I actually agree with you here, and think that we have no obligation to respond to facts whose existence we are both unaware of and incapable of practically becoming aware of in the requisite time for a decision. The problem remains that if we had evidence supporting (2)-(4), then we would also have “epistemic” reason to accept (5). I take this as a reductio of the claim that we could ever have evidence supporting (2)-(4), and as a pragmatist I take that to be equivalent to saying that they can’t all be true. Since (4) is contingent, but obviously a claim which could be true, we must reject (2) or (3); since they’re formally equivalent, we should reject both.
Jussi,
I’m getting back to your original question. I gave two reasons for thinking (5) gets a circumstantial reading. You’ve asked me about the second of them. I think the first reason suffices to hear (5) as circumstantial (when one hears it as true). So, in a way, the second reason is gravy. But I accept the second reason too, so let me elaborate a bit.
Everyone should accept that inferences such as the following do not show that MP is invalid:
1. If his office hours are now, Sobel is in his office now.
2. His office hours are now.
3. So, Sobel is in his office now
where #3 is asserted long after the time # 1 & 2 are asserted, and where #1 and 2 are asserted at a time when sobel is in his office and #3 at a time when he is no longer there.
The reason no one should think that a counterexample to MP is because everyone should hold that validity is defined not for sentences, but for propositions, however those are understood. So, the use of context-sensitive expressions present some challenges to testing the formal validity of an argument; we need be careful to avoid equivocation.
On the contextualist semantics I defend elsewhere and presuppose here, modal expressions are context-sensitive. And on the widely-accepted (among linguists) Kratzerian semantics for conditionals, conditionals contain covert modals. If that’s right, then conditionals are context-sensitive expressions. So, to test whether an inference is an instance of MP, we need to make sure that the parameter values for the modal in the conditional premise are the same as those in the conclusion. If not, we’ve got equivocation the same way we do in the Sobel-inference, not a counterexample to MP.
(4) most straightforwardly looks like a statement about possible circumstances and we need it to get from (2) and (3) to (5). Working the inference backwards, suppose that what we get from MP and or-introduction on each of (2) and (3) were a disjunction of information- relative modals. This would be because we give each of (2) and (3) informational readings (for the same reason we need the time held constant in the Sobel inference). But if the antecedents of (2) and (3) are informational base restrictors and not circumstances restrictors, then we can’t rely on (4), a claim about circumstances, to conclude (5).
That was the long-winded story; the short story, again, is just that inferences containing modal premises and conclusions require sameness of domain and ordering source in order to provide tests for inference rules, for the same reasons we require sameness of semantic value for any other context-sensitive expression.
Scott,
On your first post: I gave an argument in my original post for why (1) must have an epistemic reading. Could you give an argument for why (2) and (3) must have an epistemic reading? I’m not seeing any. Also, for why distinguishing senses of ‘ought’ is a bad idea, see my reply to Christian.
John,
I’m not following you. Are you wondering why we would accept each of (1)-(5)? I think if you just imagine a concrete situation like MINERS in which you’re deliberating about what to do with the sandbags you have, you’ll find that asserting each of them would be pretty natural, though perhaps not in the order of (1)-(5).
Hi, Paul,
I’m afraid your comments are too compressed for me to follow. Are you alluding to the wide-scoping strategy for blocking the inference from 2′-4′ to 5′? If so, that’s not an issue I’m trying to side-step: I reject wide-scoping as an account of the English. So, as long as you accept that 2′-4′ may be stated in English, you’re going to have to say something else about the inference to 5′. (There are a couple of places to look for arguments against wide-scoping for the English indicative conditional; Kratzer 1991 is decisive on its own, but you can also see the original Kolodny/MacFarlane paper.)
Janice,
I think (2) and (3) must have an epistemic reading if they are taken to be talking about the same kind of ought as in (1), which presumably is the moral ought which would govern the action of a rational being; though in such a case they are false. Or, they can be given a non-epistemic reading, where “ought…” means “…will lead to the best outcome,” and then they turn out true, but no challenge to (1). Sorry if that wasn’t quite clear in my verbosity.
Arguably we should read assertions charitably as intended to be true in the context of utterance; but I guess I was taking the context of (2) and (3) as “part of an argument posing a problem for (1)”–suggesting the oughts are the same, rather than “people on the scene of the mine urging logical considerations,” which suggests different oughts. Though it doesn’t matter which reading we assume is intended if the paradox dissolves under either.
In your earlier comment, you say that the talk of “modal bases” can accommodate the ambiguity of “ought” which the subjective-objective distinction captures. You refer to an (unpublished?) K&M paper for a discussion of the advantages of doing so; but I guess I’m asking for a little hint of what these are prior to looking at that paper, something to make it worth my while to explore. You seem to be talking about {cicumstantial/informational} readings of “ought” instead of {objective/subjective} readings. The latter terminology is familiar to most of us; so if the former just is another way of saying the latter, but bringing in some complex modal semantics, I’m wondering what should motivate me to consider this way of speaking.
Here’s part of what concerns me; Clayton noted that your claim that “(1) is true, roughly, just in case in all the best worlds compatible with our information, we do nothing” seems just false; for we take (1) to be true, yet in the best worlds compatible with our information we do not do nothing. You reply that we can save the modal reading by redefining “best world” to mean, not what we would ordinarily take it to mean, but some function combining the goodness of the world with the probability of our attaining it given certain actions on our part (if I’m reading you right here). [Crucially, of course, this must be *subjective* probability; objective probability lands you into the same soup again.] But then the modal talk seems to take a very long way around to end up saying something that can be and has often been expressed in the less misleading language of “maximizing expected consequences.”
I’m still not getting it. We agree that (1) have an informational reading. Why does that mean (2) and (3) do as well?
Argh, I don’t see a response I thought I’d posted to Jussi’s Donald/Mickey case…
Jussi,
I don’t have the intuition that Donald and Mickey are disagreeing (though Mickey clearly thinks they are). Also, I don’t think the contextualist needs to appeal to a context-shift here, she only needs that the two uses of the modal pick up on different aspects of a context. To see how this can happen, consider an example involving “must”. “Must” has both a purely epistemic use and a deontic use which can get confused and lead to merely apparent disagreement:
You’re wondering where Sobel is. I say: “well, it’s his office hours. So, he must be in his office.” You say “no, he can’t be in his office. Reina told me she just saw him in the Union 5 minutes ago. He couldn’t have gotten back to his office by now.” I can say “Oh, I didn’t mean to say that, given what I know, he must be in his office. I meant that he’s supposed to be there.” When you say “no” you take yourself to be expressing disagreement with what I said. But you didn’t, since it can both be true that he can’t be in his office and he must (given that he’s promised) be in his office. We have two modals that can, in some context, be used to express disagreement, but that in this one don’t, since they pick up on different elements of a single context to determine compatible propositions. That’s what I think is happening in MINERS. Two uses of a modal expression, one picks up on information, the other on circumstances.
Hi Janice,
It’s not about wide scope. Sorry about the compression. Consider the classic semantics for iffy oughts in deontic logic (Hanson, Lewis…): ought q if p iff all the best accessible p worlds are q worlds. This does not support factual detachment:
FD:
ought q if p
p
so q
Given that p is true in the actual world, and all the best p worlds are q worlds it does not follow that q is true in the best worlds per se, since the best worlds may be ~q worlds. So in my kill my aunt example, even though the best worlds where I kill her with a knife are ones where I use a large sharp knife, it does not follow that the best worlds per se are ones where I kill her with a large sharp knife–in those I don’t kill her at all, and that’s not changed by the fact that in the actual world, rat that I am, I will kill her for the inheritance. (This sort of semantics, blocks using disjunctive dilemma as well on your 2-4 or my 2′-4′ for similar, really derivative, reasons.) But now suppose we add that it is *unalterable* for me that I will kill her with a knife? Then not only the actual world, but all accessible worlds are ones where I do so, and so we can detach on this semantics for iffy oughts and conclude that in the best accessible worlds per se, I kill her with a large sharp knife, and so that is what I ought to do. So valid detachment for this sort of semantics look like this:
ought q if p
*unalterable* that p
So ought q
(as well as this: ought q if p, ought p, so ought q).
In your MINER case, one of the two disjuncts is not only true, but unalterably true (say that the miners are in shaft A). So you can sidestep this issue for your point by adding the unalterability of each disjunct in the MINER example (but not in mine). (Kratzer was briefly on my diss committee at UMass, but then Cresswell instead when she left for a year; in her earlier work at least, I believe she did not endorse factual detachment for iffy oughts.)
Ah, thanks, Paul! I get it. This sounds like something Nate was wondering about and I’m going to say the same thing that I said to him (and also in response to Jussi): MP is a rule that applies to propositions, not sentences. When you’ve got a context-sensitive expression, care needs to be taken to make sure semantic values are consistent from premises to conclusion. So, you’re right: it doesn’t follow that the best thing to do simpliciter is kill your aunt with a large sharp knife. But, it does follow that the best worlds in which you kill her with a knife are worlds in which you kill her with a large, sharp one. In other words, it does follow so long as the domain of the modal in the conclusion is restricted in the same way as the one in the conditional premise. So, what you (and other smart folks) are calling failure of detachment, I am calling equivocation.
That said, I also think the factual premise can force such a domain restriction–I myself hear your example that way. So, detachment sounds fine to me in that case too. But for it to sound fine, you need to hear the parameter values in the conclusion set in the right way.
>I’m still not getting it. We agree that (1) have an informational reading. Why does that mean (2) and (3) do as well?
I assume this was in response to my last post; in any case it made me realize I still wasn’t quite clear on something. Charitable readings of ambiguous claims are under at least two constraining assumptions; that the claims are intended to be both true and relevant. So if someone asserted (2) and (3) in response to (1), or in an argument such as (1)-(5) purporting to describe a paradox, charity forces us to assume that the “ought” of 2&3 is intended to be–let us first try relevant–to the “ought” of (1); i.e., informational/epistemic. Of course then 2&3 are false. Or we could charitably assume 2&3 are intended as true, forcing the objective/ciurcumstantial reading, but then irrelevant to challenge (1) with. I should never have claimed that we should assume either reading simpliciter; rather, that either reading is possible in itself, and either has some charity behind it, given how 2&3 were presented in the various examples; being charitable in both ways simultaneously is impossible; and whichever one we assume, the paradox is dissolved without questioning modus ponens. In that I agree with Janice; I just don’t understand what advantage the modal language for explaining this is supposed to give us.
“I’m not following you. Are you wondering why we would accept each of (1)-(5)? I think if you just imagine a concrete situation like MINERS in which you’re deliberating about what to do with the sandbags you have, you’ll find that asserting each of them would be pretty natural, though perhaps not in the order of (1)-(5).”
Good morning Janice
I will try to be clearer. I do not think there is a paradox. Epistemically I am not committed to claiming that 1-5 all have equal epistemic value. Based on what we know regarding the miner’s location, once we accept 1, 2-5 become epistemically irrelevant. Their logical truth status has no bearing on what we decide to do. 2-5 can only be relevant in deciding what to do if we still think we can determine which shaft they are in. (This is why I think there is a hidden epistemic condition in 2-3.) Then they too have equal epistemic value because we still think we can locate them although we have not yet done so. We still do not have a paradox because we know that if we cannot determine which shaft they are in 1 is true and 2-5 are not relevant, and vice versa. If we lack knowledge as to their location, but still believe they can be located in time to make a difference then what we are left with is (6) Either we block neither shaft or we block A or B. This is not a paradox.
I wonder how my take on this paradox might fare.
My conclusion is that there are possible worlds in which all the miners are in A, but in which one ought not block A. Similarly for B, of course. So we should reject (2) and (3).
My reasoning is thus: What one ought to do is a matter of what principles one ought apply in cases of a particular type, and not merely a matter of the consequences of applying that principle in any particular case.
Consider what principle would have to be applied to block one of the shafts. It couldn’t be a principle based on one’s knowledge of where the miners are. It would have to be a haphazard principle akin to flipping a coin. Such a principle would only minimize harm in those worlds where the coin happened to come up the right way every time. I think we’d do less harm in more possible worlds by applying a different principle.
In sum, the paradox argument requires the false assumption that oughts can be unprincipled.
Also, wouldn’t a noncognitivist simply reject (2) and (3) on the grounds that “we should X” has no truth value?
Though I think we could construct a similar argument to the original paradox without offending the noncognitivist: While the subjects in the scenario say, “if the miners are in A, we should block A,” we can write that truth-functionally as, “if the miners are in A, we are justified in blocking A.” Noncognitivists don’t reject the possibility of justifying behavior, do they?
Hi,Jason,
Rejecting one of the apparently inconsistent claims is always an option when faced with a paradox. The problem with your strategy is that most folks think there is a reading of each of (2) and (3) that make them come out true. So, a solution that can capture that is always going to do better than one that simply rejects them as false.
As for the noncognitivism issue….a lot of so-called noncognitivists these days allow that sentences containing evaluative terms are truth-apt. And it’s good that they do, since, if they don’t, their claims fly in the face of a lot of empirical evidence to the contrary. (As a side-note, I’d be interested if someone could tell me of a so-called non-cognitivist that denies this (maybe early Gibbard?); if you read your Hare and Stevenson, you’ll see that each of their accounts requires that evaluative sentences are truth-apt.)
I think my proposed solution does allow for readings of (2) and (3) which are true: for example, they can mean that knowledge of which shaft the miners are in would be sufficient grounds for blocking that shaft. And I think this is generally how the sentences are being used. (And, on this reading, the sentences can be true without there being a paradox, as John Alexander has alread pointed out.)
Interestingly, this means the subjects in MINERS are using the sentences to mean “All worlds in which the miners are in shaft A are worlds in which we should block shaft A.” Only a rather sophisticated philosopher would use the sentences to mean something like that. And that philosophical reading is the one which I claim is plausibly false.
Anyway, thanks for the tips about noncognitivism.
Sorry, I meant that the subjects in MINERS are not using the sentences to mean “All worlds in which the miners are in shaft A are worlds in which we should block shaft A.”
Hi, Jason,
Well, if the linguists tell me that ordinary speakers of English do mean something like “the best miners-in-A worlds are blocking-shaft-A worlds,” as they do, I’m going to reject the claim that ‘only a rather sophisticated philosopher’ could mean something like that. Perhaps you are assuming that ordinary speakers should have introspective knowledge of the correct formal semantics for English? If so, you must think linguists are wasting a lot of time!
Also, if you give (2) and (3) informational readings, you’re either going to have to accept (5) which will be inconsistent with (1) and so not resolve the paradox or you will have to reject modus ponens. That’s the Kolodny and MacFarlane conclusion, but I’m thinking a solution that doesn’t have to give up MP, as mine does, is other things equal to be preferred to one that does.
I’m not sure about the linguists, but this just seems like common sense to me.
Imagine we are in the miners scenario, and we hear a person deliberating with (2) and (3). We say, “yes, if the miners are in A, we should block A. And if they’re in B, we should block B. Since they’ve got to be in one or the other, we should block one of them.” The deliberator will likely respond, “no, because we don’t know which one.”
At that point, we can say, “but that’s not what the linguists tell me you meant. You didn’t mean that your knowledge was required to justify the decision. You were just talking about miners-in-A and miners-in-B worlds.” Or, in more ordinary language, we could say, “but you didn’t say you had to know which shaft they were in. You just said they might be in one or the other of the shafts. Your knowledge had nothing to do with it.”
At that point, the deliberator might say, no, he meant that if he knew which shaft they were in, then he should close that shaft. Or, if he isn’t able to think so clearly about it, he might just get confused. He might say he’s no longer sure what he meant, exactly, but that he’s quite sure he doesn’t feel justified in blocking either of the shafts.
Maybe my intuition is leading me astray here, but I can’t imagine it playing out any other way.
Also, I think I can give (2) and (3) informational readings without leading to a paradox, and without rejecting modus ponens. I’m treating (2) and (3) as ordinary language statements and rejecting the suggested translation into propositional logic. We can’t apply modus ponens before we get the right translation, and once we get the right propositional expression, the paradox disappears.
Sorry, I meant I was rejecting the suggested translation into predicate logic, not propositional logic, of course.
Jason,
Do you mean that the deliberator is saying “no” to the premises, or to the conclusion? To (2) and (3), or to (5)?
I mean the deliberator is saying “no” to the conclusion, to (5). I’m disagreeing with Janice’s claim that (5) seems true.
I’ve put together a more comprehensive presentation of my argument here.
I see. But I’m still confused, Jason, because I thought your solution was to reject (2) and (3) as false. So why are you giving an example in which the speaker accepts (2) and (3) but rejects (5) as false?
(Just by the way, my personal inclination is to accept (2) and (3) and reject (5), so I am like the deliberator in your example.)
My solution requires first accepting that the sentences don’t always express the same propositions. Sometimes the same sentence can express a true proposition and sometimes a false one. When I argued that (2) and (3) were false, I was arguing against the truth of the propositions Janice introduced at the outset: propositions about all miners-in-shaft-A and miners-in-shaft-B worlds. I think those propositoins are plausibly false, and I do not think the deliberator is likely to believe those propositions anyway.
When the deliberator uses the sentences in (2) and (3), different propositions are expressed. E.g., the deliberator uses the sentence in (2) to express the proposition, “If we know the miners are in shaft A, we should shut shaft A.” This proposition is plausibly true, and under this reading of (2) and (3), there is no paradox. Also, this explains why the deliberator accepts (2) and (3) but rejects (5).
Just to be clear: Janice also thinks that the same sentence expresses different propositions on different occasions. (That’s why her solution is called “contextualist”.)
There is a big problem with saying that the non-epistemic readings of (2) and (3) are false. The problem is that when uninvolved observers know that the miners are in shaft A, they are pretty happy saying, “Those guys ought to block shaft A”, even when the deliberators don’t know which shaft the miners are in and the observers know they don’t. I think this means the consequents of (2) and (3) have to be true, when the antecedents are, whether the deliberators know the antecedents or not.
In my version (the King Henry paradox), the sentences are not about what ought to be done but what is better. (Of course, these may turn out to be the same thing.) This removes the distracting irrelevancy of whether the non-epistemic interpretations of (2) and (3) are true. I suppose everyone will agree that:
If the miners are in shaft A, it is better to block shaft A.
The non-epistemic interpretation of this is plainly true, isn’t it?
Hi Jamie,
I’m afraid I don’t see the problem you’re getting at. Sure, I can imagine somebody who knows the miners are in shaft A saying, “the deliberator ought to block shaft A.” Their statement does not depend on the deliberator knowing that the miners are in shaft A, but it does depend on their knowing the location of the miners. So I don’t see why this should cause a problem for my position.
Also, I don’t see what is gained by replacing “we ought to” with “it is better to.” The alternative phrasing is still vulnerable to my objection, which is that there are plausible miner-in-shaft-A worlds in which it is not better to block shaft A.
Oh, so you mean that the relevant knowledge is what the speaker knows? I thought you meant it was what the agent (what you’re calling the deliberator) knows. When you wrote,
what was your idea? When I am judging what you ought to do, is this a matter of which principles I can apply or a matter of which principles you can apply?
Can you tell me, by the way, which are the plausible miner-in-shaft-A worlds in which it is not better to block shaft A? Are those worlds in which blocking shaft A does save all the miners, but for some reason it is better not to save them?
We’re talking about different speakers making different statements here.
(I’m gonna shift to “agent,” because “deliberator” is a pain to type.)
When somebody says, “the agent ought to shut shaft A, because the miners are in shaft A, even though the agent doesn’t know that the miners are in shaft A,” they are making a moral judgment based on their knowledge. However, their judgment does not implicate the moral character of the agent. The agent might choose to shut shaft A for a bad reason–perhaps by flipping a coin. In that case, we might say their action had the better result, but we wouldn’t say they did the moral thing. They just got lucky.
So what do they mean, when they say the agent ought to shut shaft A? They mean that shutting shaft A will have the best result, and that they want the agent to shut shaft A, whether the agent has a good reason to do it or not. So far as I can tell, this doesn’t contradict my reading.
My idea is that forming a moral judgment is a matter of applying principles which range over cases, and not just a matter of doing something which has the best result in a particular case. So, when the onlooker judges what the agent should do, they are not making a statement about what would be the moral thing to do for the agent; they’re making a statement about what would be the moral thing to do for somebody in the agent’s position who had the requisite knowledge. Thus, while they think the agent should shut shaft A, they would not judge the agent immoral for doing nothing, since they know that the agent didn’t have the knowledge required to know which shaft to shut.
As for plausible miner-in-shaft-A worlds in which it is better not to block shaft A . . .
It is plausible there are worlds in which the miners in shaft A are saved only to then suffer an unimaginably horrific fate. In those worlds, it would’ve been better for the miners to die in the shaft.
We can also look at what is better for others, and not just what is better for the miners. Even in those worlds in which it is better for the miners to be saved, these might not always be the best worlds for the majority of people.
I indicated one reason for this in my initial post on the topic: In worlds in which we save the miners, we are doing so by chance, on the basis of a toss of the coin, and not according to a strong moral principle. Worlds in which we do that are plausibly worse than worlds in which we make such decisions on strong moral principles. (For example, there can be worlds in which we happen to save the miners in this scenario, but in which we kill lots of miners in other similar scenarios.)
Another reason is that there’s simply no reason to think that the overall value of a world can be judged by whether or not a few people die in it.
I don’t understand all the stuff about principles. I was only pointing out that you’d switched from using the agent’s knowledge to using the speaker’s knowledge as the epistemic basis. The latter is contextualist, the former isn’t. So you are now endorsing a view that in its epistemic part is like Janice’s.
I agree, of course, that there are some worlds in which the miners are saved that are better than some worlds in which they aren’t. But I’m certain that this is just a red herring. If the miners are in shaft A, then it is, in fact, better to block shaft A. This is not because every single world in which the miners are saved is better than every single world in which they aren’t. It’s because the result, the miners are saved, is better than the result, the miners die.
I don’t think I switched. What is relevant is the knowledge of the person whose moral character is in question. In the first set of statements–(2) and (3)–it is the agent’s knowledge which is relevant, because the agent’s moral character is on the line. In the second statement, the moral character of the speaker is on the line, while the moral character of the agent is not, so it is the speaker’s knowledge which is relevant.
As for the second part, it looks like an “ends justify the means” argument, which I don’t find attractive.
I should explain that last comment. It’s not that I’m against consequentialism per se. It’s just that it’s not clear what “because the result is better” is supposed to mean. If “the result is better” is a proposition, how is its truth determined?
The examples aren’t really about anyone’s moral character. They are examples about what someone ought to do. And my example is about what outcomes are better. Everyone (almost) agrees that these concepts are information-relative; the question is how they are, and of course what to say about the paradox(es).
As to the meaning and truth of “because the result is better”, I’m not sure I understand your question. Are you asking about the semantics for ‘better’? If so, that’s just what I’m interested in, and so I’m using the example to get a better grip. Or is what you’re asking more about the ‘because’ part? (The clause “because the result is better” cannot express a proposition, of course, because it is not a complete sentence, and a proposition must be a complete thought.) Or are you genuinely unsure whether it is better than the miners be saved than that they die?
If we take statements of what a person ought to do as statements about what somebody in that person’s situation should do, given the right knowledge, then I think that person’s moral character is on the line–unless that person does not have the required knowledge.
As for my question . . . I gave examples of worlds in which it is plausibly better for the miners to not be saved. You reject those examples–and the entire argumentative strategy–as a red herring. If you don’t accept that sort of argument against the truth of “it is better to save the miners,” then I wonder what sort of argument you might accept for that proposition. If you don’t have an argument for it, then wouldn’t you agree that it is at least plausibly false?
Okay, I probably just don’t understand the character idea.
I don’t know what argument I would accept for the proposition that it is better to save the miners than to let them die. Frankly, it seems much, much more obviously true than any theory about ordering of possible worlds. So, no, I do not agree that it is plausibly false.
Do you think that it’s better to let the miners die than to save them, or that the two are equally good — or do you think they are incommensurable? Each of those alternatives seems to me to be so outlandish that I can’t believe you accept any of them.
(By the way, I just heard from the miners, and they urged me to vote against you in the upcoming election for Director of Mining Safety Operations.)
Jamie,
I think it’s better to save the miners, if given the choice. More than that, I’d say that there is a moral imperative to save the miners. This is not because the world will be a better place if the miners live. It is because letting the miners die is unacceptable (unless saving them is demonstrably less acceptable). However, if you don’t know how to save them, then you are not responsible for letting them die (unless you should know how to save them, of course). That’s where the notions of character and principle come in: we judge a person’s moral character by whether or not they perform according to shared principles.
I don’t see what is gained by replacing “one ought to” with “it is better to.” My view is that, when we say “if the miners are in shaft A, we ought to shut shaft A” and “if the miners are in shaft A, it is better to shut shaft A,” we aren’t making such different claims. The latter may be weaker, but both phrases entail the same moral imperative: knowledge of which shaft the miners are in is sufficient to create a moral imperative to shut that shaft. If the miners are in shaft A, the agent has a moral imperative to shut shaft A–assuming that the agent knows the location of the miners. A moral imperative cannot extend to somebody who lacks the required knowledge. That’s what I meant when I said the agent’s moral character is not on the line. They can’t be judged for what they don’t know (in this case.)
In your counterexample, you imagined an onlooker who knows the location of the miners, and who says, “the agent ought to shut shaft A,” even though the onlooker knows that the agent doesn’t know the miners are in shaft A. In this case, the onlooker’s moral character might be judged. If the onlooker said the agent ought to shut the other shaft (thereby killing all of the miners), we would condemn their statement as immoral. By saying the agent should shut shaft A, they are acting morally, but they are not calling the agent’s moral character into question–they are not saying the agent has a moral imperative to shut shaft A.
The agent says, “if the miners are in shaft A, I ought to shut shaft A,” but when the agent utters the words, “I ought to shut shaft A,” they do not express the same proposition as when the onlooker says, “the agent ought to shut shaft A.”
Jason,
If the miners are in shaft A, it is better to block shaft A.
Okay, I feel better. Above you seemed to be saying that “It is better to save the miners” is “plausibly false”. But, you do think it is better to save the miners, so that’s good. Can I now get you to agree to this?
If you agree, then I think it’s pretty obvious why my ‘better’ formulation is better than the ‘ought’ formulation. If you will not agree, then I will give you an argument for it.
I’d rather leave the character business out of it. It may be an interesting question in its own right, but it brings many sticky issues in its wake. (For instance, lots of people are skeptical that there even is such a thing as character, and although I am not positively skeptical, I have no good answer to these skeptics.)
I agree with that statement, and I agree with both (2) and (3), under the reading I’ve given — but I don’t see these as substantively different statements.
Oh, okay. I thought you said you rejected (2) and (3).
I agree with you, then. Both (2) and (3) are true, and so is (4), but they do not entail (5).
Since (5) can be derived from (2), (3), and (4), using modus ponens plus a couple of disjunction rules, one of those rules must be invalid. My view is that it’s modus ponens that’s guilty.
Janice doesn’t agree. She thinks that (5) is true (and the argument is valid), but that (1) is also true. Contextualists get to say things like that. You are also a contextualist, I think, but you agree with me that the argument is invalid.
I agree that the argument is invalid, but I don’t agree that we any of our rules must be invalid. I’m actually torn between two perspectives here, though maybe these two views aren’t so different.
On the one hand, we can say that (2) and (3) are true, but their meaning is obfuscated by their grammatical form. If we are going to apply any rules of inference to (2) and (3), we need to first restate them in more explicit terms. We might restate them as “if the agent knows the miners are in shaft [A or B], then the agent has a moral imperative to shut shaft [A or B].” We must correct for the inherently misleading nature of the expressions, or else logic will lead us astray.
On the other hand, I’m tempted to just say that (2) and (3) are not true in a factual sense. I agree with them, as I said, but such agreement is not an agreement as to the truth of some matter of fact. It is agreement with a convention. My agreement is an act of enforcing a convention, and not affirming a truth. Since (2) and (3) are not true in the factual sense, the rules of inference don’t apply. However, on this reading, (2) and (3) do entail some facts, such as the fact that knowledge of where the miners are creates a moral imperative (for some set of individuals) to close that shaft.
Hi, Jason,
You and Jamie seem to have arrived at the place where we were earlier. First you said:
“I wonder how my take on this paradox might fare. My conclusion is that there are possible worlds in which all the miners are in A, but in which one ought not block A. Similarly for B, of course. So we should reject (2) and (3).”
To which I replied:
“Rejecting one of the apparently inconsistent claims is always an option when faced with a paradox. The problem with your strategy is that most folks think there is a reading of each of (2) and (3) that make them come out true. So, a solution that can capture that is always going to do better than one that simply rejects them as false.”
Then you said:
“I think my proposed solution does allow for readings of (2) and (3) which are true: for example, they can mean that knowledge of which shaft the miners are in would be sufficient grounds for blocking that shaft. And I think this is generally how the sentences are being used.”
To which I replied:
“if you give (2) and (3) informational readings, you’re either going to have to accept (5) which will be inconsistent with (1) and so not resolve the paradox or you will have to reject modus ponens. That’s the Kolodny and MacFarlane conclusion, but I’m thinking a solution that doesn’t have to give up MP, as mine does, is other things equal to be preferred to one that does.”
And then you said:
” I think I can give (2) and (3) informational readings without leading to a paradox, and without rejecting modus ponens.”
You’ve spun through these options again with Jamie and now say:
“I agree that the argument is invalid, but I don’t agree that we any of our rules must be invalid.”
and also:
“Since (2) and (3) are not true in the factual sense, the rules of inference don’t apply.”
To which I now say:
Are you now saying (2) and (3) are neither true nor false? Can you explain how this fits with what you said earlier when you said they were plausibly false and then you said they are true?
And:
If the rules of inference don’t apply to (2) and (3), what do you mean when you say that the argument is invalid?
Hi Janice,
As I said, I’m torn between two readings. Up until my last post, I’d only been arguing for one of them, though I don’t think they’re so different. I’ll try to clarify by answering your questions.
I think (2) and (3) state a moral imperative. I’m ambivalent about whether or not moral imperatives can be true or false. If they can’t, then (2) and (3) are neither true nor false. If they can, then they can be true or false, but their truth or falsity depends on how they are being used. If they are taken as statements about shut-shaft-A and shut-shaft-B worlds, then I think they are plausibly false. I don’t see a good reason to think most possible worlds are better if the miners are saved. If the truth or falsity of (2) and (3) depends on the outcome of the agent’s actions, then I think they’re plausibly false.
But, as I said, I don’t think this is how they are being used. I think they are being used to state (or enforce) a moral imperative, and the meaning of this statement is not about the outcome of the agent’s actions. It’s about what is not acceptable (by convention), given the right knowledge. It is not acceptable for the agent to leave shaft A open, if the agent knows the miners are in that shaft. This is a statement of fact, in so far as it is a fact about what some people find acceptable, but it is not a statement about what is better in any other sense.
The Miners Paradox requires that we treat (2) and (3) as statements of possible worlds, such that their truth or falsity depends on the truth or falsity of “the miners are in [A or B]” and, if that is true, then the truth of “we ought to shut shaft [A or B].” But I don’t think (2) and (3) can be read this way. So we cannot apply the rules of inference as the Miners Paradox requires. If (2) and (3) have truth conditions at all, it is not in the way their grammar suggests. So, if we can draw inferences about them, we have to rewrite them in a different way. I suggested “if the agent knows the miners are in [A or B], then the agent has a moral imperative to shut shaft [A or B].” This restatement avoids the paradoxical conclusion, and seems to fit with how the sentences are being used.
I just read through the comments which had been posted before I joined the discussion–I’d only skimmed them earlier. It looks like my view is quite similar to, and compatible with, Scott’s.
I see. You’ve changed your mind and now aren’t sure what to think. That’s fair enough.
Sorry if it seems that way, but I don’t think that’s a fair statement. I haven’t changed my mind, and I’m more or less confident in my account. I’m just ambivalent about whether or not moral imperatives are truth-apt, so I’m giving parallel accounts: one in case they are truth-apt, and one in case they aren’t. These parallel accounts don’t seem so different to me, so I’m not compelled to choose between them.
Sorry if I haven’t explained myself very clearly. I’ll try harder next time.