With this post we are starting a new feature at PEA Soup: Author replies to book reviews published in Ethics. Our inaugural discussion is between Chrisoula Andreou (Utah) and Justin Snedegar (St. Andrews). Chrisoula reviews Justin’s new book, Contrastive Reasons (OUP, 2017) here. Justin Snedegar’s reply follows below.
Thanks first of all to the Daves for the opportunity to continue the discussion here. And thanks most of all to Chrisoula for her excellent review of my book. Her questions and objections have given me the chance to think harder about some central issues that didn’t receive all the attention they deserved in the book. In particular, she’s made clear that there are important questions about the nature of the objectives the promotion of which I appeal to in my contrastive analyses of reasons. I used a desire to remain neutral between competing views of these objectives as an excuse for not discussing them much, but this neutrality was about relatively substantive questions about whether the objectives were desires, values, etc. Chrisoula’s objections show that there are questions about structural or formal features of objectives and of the promotion relation which are crucial for my theory, and indeed for many theories of reasons.
Chrisoula does a very nice job summarizing the book in her review, so I will get straight to discussion of her objections here.
For reference, and to see right away why questions about the nature of objectives and their promotion will be crucial for me, here are my analyses of reasons for and reasons against:
For: r is a reason for A out of set of alternatives Q iff there’s some objective O such that r explains why A-ing promotes O better than any other alternative in Q.
Against: r is a reason against A out of Q iff there’s some O such that r explains why A-ing promotes O less well than some other alternative in Q.
Contrastive promotion?
Chrisoula first objects to my assumption that whether and how well an action promotes an objective is not contrastive, and so does not depend on the relevant set of alternatives. In fact, she suggests that my central example from Chapter 2 can only be made to work if I explain the reasons involved using an objective the promotion of which does seem to be contrastive in this way. In the example, I need to get to campus, but am out of shape and don’t want to wear myself out getting there. So the following two claims seem true:
D/B: That campus is twenty miles away is a reason to drive rather than bike.
B/R: That campus is twenty miles away is a reason to bike rather than run.
I claim that the objective that explains both of these (contrastive) reasons is my desire not to wear myself out. Chrisoula points out that if I’m sufficiently out of shape, then biking twenty miles will wear me out (even if running would wear me out even more), and so biking to campus would not promote this objective at all. If so, then B/R is not true, or at least the reason is not explained by my desire not to wear myself out.
She suggests a fix, which is to hold that the objective in this case is actually a desire to minimize wear. Then both D/B and B/R might be true, but this is precisely because promoting the minimization of wear is a contrastive matter: minimizing wear means wearing myself out less than the alternatives. Whether an action promotes this objective will obviously depend on what the alternatives are (e.g., biking promotes it relative to {bike, run} but not relative to {bike, drive}).
In the book, I insist on a non-contrastive promotion relation in order to have a non-contrastive foundation for reasons that allows me to explain various constraints on how reasons can vary with the alternatives. To take an example that will be important later, if r is a reason to A rather than B and a reason to B rather than C, both explained by objective O, then it seems that r must also be a reason to A rather than C, explained by O. If contrastivism allows unconstrained variation of reasons across sets of alternatives, then it will allow for violations of this and other plausible constraints. With the assumption that the promotion relation gives us a non-contrastive ranking of how well different actions promote a given objective, we can provide these constraints. For r to be a reason to A rather than B and to B rather than C, both explained by O, it must be (given my promotion-based analysis of reasons) that A-ing better promotes O than B-ing, and that B-ing better promotes O than C-ing. Thus, given a non-contrastive and transitive (transitivity will be important later) ranking, A-ing will better promote O than C-ing. If the promotion of some objectives is contrastive, though, there is no guarantee that this will be true: whether and how well A promotes O may differ depending on whether the comparison is with B or with C.
In fact, though, Chrisoula’s objection has shown me that I can allow for some objectives to be such that whether an action promotes them is contrastive in this way. What is actually most important for me is that there is a non-contrastive ranking of alternatives to appeal to. Objectives like minimizing wear or maximizing happiness should be understood in terms of a ranking of alternatives, in terms of how well they promote a more basic, in some sense, objective, like avoiding the pain associated with being worn out or making people happy. (I don’t need to claim that there is one or even a small number of “basic” objectives. All I claim is that the objective of, e.g., maximizing happiness is somehow derivative on the more basic objective of making people happy.)
What I have to insist on is that how the alternatives in a given set are ranked, and so which actions promote the objectives involving minimization or maximization, is lifted from the underlying ranking of all the alternatives given by this kind of more basic objective. If all of this can be made to work, then I can allow for the promotion of some objectives to be contrastive and still explain the constraints on how reasons can vary. To see this, suppose I take Chrisoula’s suggestion and say that the objective that explains the reasons in D/B and B/R is minimizing wear. That campus is twenty miles away explains why driving better promotes this objective than biking and why biking better promotes it than running (and hence D/B and B/R are true). But given what I’ve said about how to understand this kind of objective, it will follow that driving better promotes the objective than running, and so it will be true that the fact that campus is twenty miles away is a reason to drive rather than run, just as we’d hope.
Transitivity
So I was wrong to insist that whether and how well an action promotes an objective is always non-contrastive; this will (unsurprisingly, I now see) depend on the nature of the objective in question. Chrisoula’s second objection also concerns what I took to be a minimal assumption about the promotion relation, namely that it gives us a transitive ranking, such that if A better promotes O than B, and B better promotes O than C, then A better promotes O than C. As I said above, this assumption allows me to explain why considerations that are reasons for A rather than B and for B rather than C, both explained by O, are also reasons for A rather than C, explained by O.
In Chapter 5, I aim to show how we can insist on this kind of “transitivity of reasons” while allowing (if we want!) for the intransitivity of the ‘more reason than’ relation. To do this, we just need to adopt the thesis that Chrisoula names VARY: the strength of the reason for some action depends on the importance of the objectives that provide the reasons, which can vary with the comparisons being made. So even if we guarantee that r is a reason for A rather than C, explained by O, the importance of O may be quite different relative to {A, C} than it is relative to {A, B}. Thus, the weight of the reason to A rather than B might be quite different than the weight of the reason to A rather than C. This would allow us to maintain that any considerations that are reasons for A rather than B and for B rather than C are also reasons for A rather than C, while accepting the possibility that there might be more reason for A than for B, more reason for B than for C, but not more reason for A than for C.
Chrisoula argues that this is actually unstable. How well a given action promotes some objectives—call them multidimensional objectives—seems to depend on multiple factors, which may themselves serve as objectives. Thus, if the importance of these objectives—the dimensions of the multidimensional objective—can vary with the set of alternatives, then it may be that how well an action promotes the multidimensional objective will likewise vary with the set of alternatives, due to this variation in the importance of the various dimensions. Given For and Against, above, this would undermine my ability to explain constraints like the transitivity of reasons.
As Andreou notes, the easiest response would be to just reject VARY—after all, I am not at all convinced that ‘more reason than’ is intransitive. Nevertheless, I think it is worth seeing how far we can get in trying to allow for the possibility, and isolating the kinds of theses we’d need to endorse to do so.
The example that Chrisoula appeals to is one in which my objective is to have an appealing head of hair, and I have the option of taking some number of pills that affect the number and distribution of hairs on my head. As I take more pills to get more hairs on my head, the distribution of those hairs changes, leaving an increasingly thinning patch on the crown of my head. The claim is that in a comparison of n and n-1 pills, the distribution of hairs does not make a very important contribution to the overall appealingness of my hair. But for comparisions between, say, zero pills and ninety pills, it makes a more important contribution to the overall appealingness. Thus, the fact that I will grow more hair is a reason to take one pill rather than zero, …, and ninety rather than eighty-nine, explained by my objective of having an appealing head of hair. But it is not a reason to take ninety rather than zero, since taking ninety will not lead to a more appealing head of hair, due to the importance of the distribution.
An initial thing to note about this example is that it isn’t clear that it is actually an example of a failure of transitivity of reasons. Chrisoula cites the fact ‘that I will grow more hair’. But the fact is actually that I will grow more hair by taking n pills than by taking n-1 pills, for the relevant value of n. So we actually have a bunch of different facts serving as reasons, for the possible values of n; the transitivity of reasons is meant to explain why when one and the same fact is a reason to A rather than B and to B rather than C, it is also a reason to A rather than C.
So assume the fact is something like that I want an appealing head of hair, or that my class reunion is coming up (which is why I want the appealing head of hair). This fact is then a reason to take n rather than n-1 pills for any n, but not a reason to take, say, ninety rather than zero. Once we do this, the force of the example just comes from the plausibility that ‘more appealing head of hair than’ is intransitive (due to its multidimensionality). If we accept this, then we will likely just think that contrastive reasons may be intransitive in the way I rule out in Chapter 4, since objectives like having an appealing head of hair will give us intransitive promotion rankings. So I would have been wrong to rule this out.
But this does not yet point to an instability in allowing for the importance of an objective to vary in a way that generates intransitivity in ‘more reason than’ without allowing for the intransitivity of reasons. To get to that, we would need to argue that allowing for the importance of a reason-providing objective to vary with the comparison (that is, accepting VARY) forces us to allow for the factors that contribute to a multidimensional objective to make contributions of varying degrees, depending on the comparison (that is, to accept VARY*). Chrisoula has given us a plausible example of the latter, but assuming VARY is not part of the argument for this. It is nevertheless tempting to think that these two things will go hand-in-hand, since the same structure shows up: in the ‘more reason than’ case, we have individual pro tanto reasons contributing to an overall ‘more reason than’ verdict. In the case of a multidimensional objective, O, we have individual factors contributing to an overall O verdict or score. But though structurally similar, these involve different kinds of relations. So there is no direct argument from VARY to VARY*.
Thanks to all involved!
Justin:
Regarding your discussion of promotion, I agree that if you can support the constraints you’re looking to support without a fully non-contrastive promotion relation (as it seems you can), then adjusting your view in the way you describe is the best response to the first concern I raise.
Regarding your discussion of transitivity, I agree that VARY does not force us to accept VARY*, but I do think that once one accepts VARY, it is odd to refrain from accepting VARY* without some explanation as to why, since the argument for accepting the former is so closely related to the argument for accepting the latter. Relatedly, without some such explanation, I don’t think you can comfortably allow for VARY while suggesting that your assumption that the promotion relation is transitive is “minimal.”
Thanks, Chrisoula.
About intransitivity: I agree that my assumption that the promotion relation is transitive is not so minimal, given arguments like the one you give that some objectives, like ‘having an appealing head of hair’, generate intransitive rankings of how well different actions promote them. (For what it’s worth, much of the discussion of promotion (at least at the time I was writing this) was put in terms of probability-raising. I wanted to not assume that view, but as you can see I imported important structural assumptions.) So if we’re convinced by these arguments, then that means that the theory should not deliver the constraint that reasons are transitive in the way I insisted on. I believe that I could drop this constraint but maintain the others, though I’ll need to think through that. So ultimately, your comments have made me think I should probably either give up on VARY or else allow for intransitivity of reasons.
But I do think that there’s an important difference between the view that how important an objective is to promote (or respect) may depend on the specific alternatives (VARY), and the view that when it comes to determining how well an action promotes an objective, different factors will matter more or less, depending on the specific alternatives (VARY*). The first is clearly a normative question, while the second is at least arguably a question about the nature of the concepts involved in the objective (e.g., appealingness as a head of hair).
So for example you might have a view according to which different values are lexically ranked, such that lower-ranked ones can only serve as tie-breakers, or maybe not be used at all, in choices involving higher-ranked ones, but in choices where no higher-ranked values are present, they may be very important to promote. To borrow and amend the kind of case that Caspar Hare and others have recently used, you might think that returning lost property is very important when no higher-ranked values are at stake. But the fact that Ron’s lost iPod is on the same island where Annie is stranded is just not important when lives are at stake, and so can’t provide reasons for you to go to that island rather than the other island where Debbie is stranded. I’m not sure exactly what kind of explanation there is for this fact, according to this sort of view, but it will presumably be some kind of ethical story about appropriate responses, including what kinds of things it’s appropriate to think about in deliberation, when people’s lives being in danger. Generalizing, then, one motivation for VARY might be a view about what kinds of response and deliberation is appropriate when certain sorts of values are at stake.
If something in that ballpark is your motivation for accepting VARY, then there’s no obvious extension to VARY*, which doesn’t involve appropriate responses to any values, but rather the nature of what I called multidimensional concepts like appealingness as a head of hair.
All of this goes beyond anything I said in the book. And none of this undermines the compellingness of the case you give for VARY* in the review. But it might be one way to explain how VARY could be true that does not put much pressure on us to accept VARY*.
Hi Justin!
This is a bit wavy gravy, but here goes.
I take it that you think there’s an objective parameter in the reason relation. Why? Because objectives provide a clean rationale for structuring the sets of alternatives relative to which claims about reasons are true, on your account. So, in the name of making things concrete, rather than a Scanlonian R, with parameters for a individual (x), proposition (p), circumstance (c) and action (a), you advocate something more like R where O takes objectives as values, like the objective of not wearing oneself out before going to work.
I’m wondering how attached you are to thinking of objectives as an additional relatum in the reason relation. Here’s why I care. I think talking about the relata in the reason relation is misleading in an important sense. The value for p in R plays a much different role than the values for x, c, and a. In particular, the reason defined by that relation, on standard assumptions, is numerically identical with the value for p. It isn’t numerically identical with the values for x or c or a. For example, the reason for me to buy Shyam a beer at the Pacific APA is that we haven’t seen each other in a while. Here’s the relevant relation, again using Scanlon’s view as a model:
R.
Note that you can pick out the reason only by naming the p value; not by naming any of the other values:
1: The reason for me to buy Shyam a beer is that we haven’t seen each other in a while.
2: #The reason for me to buy Shyam a beer is me.
3: #The reason for me to buy Shyam a beer is at the APA.
4: ?The reason for me to buy Shyam a beer is to buy Shyam a beer.
The explanation of 1-4 is obvious: the reason is numerically identical with the value of p; not with the values of any of the other parameters.
But here’s an interesting fact about objective-talk: it patterns with 1 not with 2-4. Suppose that the objective of fraternizing with my academic sibling structures the relevant alternatives to buying him a beer. We can pick out that objective in the style of 1:
5: The reason for me to buy Shyam a beer is to fraternize with my academic sibling.
There are two things that could be going on: either ‘to fraternize with my academic sibling’, in addition to picking out an objective, also picks out a proposition that is a second reason to buy Shyam a beer OR, and here’s my preferred view, objectives, just like propositions — but unlike the values of individuals, circumstances and actions — are parts of reasons. Just as 1 behaves differently from 2-4 because 1 picks out (one of) the thing(s) that bears the property of being a reason, 5 behaves differently from 2-4 because 5 picks out one of the things that bears the property of being a reason.
I think 1-5 suggest a different picture of the reason relation than the one you advocate. On a Scanlonian deformation of your view, your reason relation is R whereas I think the linguistic data favours thinking of it in the following way: R<x,,c,a> where takes up a single argument space in virtue of being the complex object that bears the property of being a reason.
I think this tweak gets you what you want in claims about the (intransitive) logic of reasons, but is closer to the genuine ontology of reasons.
I see that the comment boxes don’t take text with right-angle brackets very well, having deleted many important characters in my post. I’ll try to rewrite it using different characters and I encourage whoever manages this page to delete the earlier post and this one.
(repost from before, where ‘]’ is used rather than the corresponding right-angle bracket for ordered tuples — fingers crossed this works)
Hi Justin!
This is a bit wavy gravy, but here goes.
I take it that you think there’s an objective parameter in the reason relation. Why? Because objectives provide a clean rationale for structuring the sets of alternatives relative to which claims about reasons are true, on your account. So, in the name of making things concrete, rather than a Scanlonian R[x,p,c,a], with parameters for a individual (x), proposition (p), circumstance (c) and action (a), you advocate something more like R[x,p,c,a, …, O] where O takes objectives as values, like the objective of not wearing oneself out before going to work.
I’m wondering how attached you are to thinking of objectives as an additional relatum in the reason relation. Here’s why I care. I think talking about the relata in the reason relation is misleading in an important sense. The value for p in R[x,p,c,a] plays a much different role than the values for x, c, and a. In particular, the reason defined by that relation, on standard assumptions, is numerically identical with the value for p. It isn’t numerically identical with the values for x or c or a. For example, the reason for me to buy Shyam a beer at the Pacific APA is that we haven’t seen each other in a while. Here’s the relevant relation, again using Scanlon’s view as a model:
R[me, we haven’t seen each other in a while, at the Pacific APA, to buy Shyam a beer]
Note that you can pick out the reason only by naming the p value; not by naming any of the other values:
1: The reason for me to buy Shyam a beer is that we haven’t seen each other in a while.
2: #The reason for me to buy Shyam a beer is me.
3: #The reason for me to buy Shyam a beer is at the APA.
4: ?The reason for me to buy Shyam a beer is to buy Shyam a beer.
The explanation of 1-4 is obvious: the reason is numerically identical with the value of p; not with the values of any of the other parameters.
But here’s an interesting fact about objective-talk: it patterns with 1 not with 2-4. Suppose that the objective of fraternizing with my academic sibling structures the relevant alternatives to buying him a beer. We can pick out that objective in the style of 1:
5: The reason for me to buy Shyam a beer is to fraternize with my academic sibling.
There are two things that could be going on: either ‘to fraternize with my academic sibling’, in addition to picking out an objective, also picks out a proposition that is a second reason to buy Shyam a beer OR, and here’s my preferred view, objectives, just like propositions — but unlike the values of individuals, circumstances and actions — are parts of reasons. Just as 1 behaves differently from 2-4 because 1 picks out (one of) the thing(s) that bears the property of being a reason, 5 behaves differently from 2-4 because 5 picks out one of the things that bears the property of being a reason.
I think 1-5 suggest a different picture of the reason relation than the one you advocate. On a Scanlonian deformation of your view, your reason relation is R[x, p, c, a, …, O] whereas I think the linguistic data favours thinking of it in the following way: R[x, [p,O], c, a] where [p,O] takes up a single argument space in virtue of being the complex object that bears the property of being a reason.
I think this tweak gets you what you want in claims about the (intransitive) logic of reasons, but is closer to the genuine ontology of reasons.
Hi Justin,
I see what you mean about the possibility of different motivations for accepting (or at least being sympathetic to) VARY and how they may or may not impact the case for VARY*.
Since you wondered about my motivation: It is not the one you consider (though I’ll be thinking more about it now that you raise it); rather it is related to the idea, associated with “spectrum cases,” that “the relative significance of [objectives or] factors relevant for comparing alternatives merely differing in degree, may differ from the relative significance of those [objectives or] factors for comparing alternatives differing in kind” (Temkin, “A Continuum Argument for Intransitivity,” p. 194); and this motivation is relevant to VARY* too. But, as you illustrate, other motivations may not be.
Thanks, Nathan. I am definitely open to the kind of view you advocate, according to which the objective is in some sense part of the reason. Besides the interesting point you mention, I think this would help to make sense of some other features of reasons. In particular, it would let us draw distinctions bewteen reasons that overlap in the p parameter, and so (at least if reasons are facts/propositions) are the “same” reason, but nevertheless should be distinguished.
I’m not sure, though, that this is the only way to make sense of your observation or this fact about overlap. For overlap, at least, it may be enough just to treat O as part of the grounds, or maybe the background conditions, for the reason. We would then at least be able to point to that to draw the kinds of distinctions we want to draw. I wonder if this kind of view would be enough to explain why we can cite what seems to be the same reason using the objective, rather than the proposition? I suspect that thinking about how citing explanations (or maybe, to follow Dan Fogal, causes) works will be relevant here: are there contexts in which we can felicitously cite what is best thought of as the grounds of an explanation for p as the explanation for p? Of course there will probably be parallel debates about whether these “grounds” of the explanation are really part of the explanation.
Thanks, Chrisoula. In the book, the argument I rehearsed to motivate VARY was a spectrum argument (from Parfit’s discussion of the repugnant conclusion), so you were totally right to point out that similar looking arguments support VARY*.
At this point I feel like I should make clear to those who haven’t read the book that I am not personally very tempted to think that ‘more reason than’ is intransitive. The way this comes up in the book, in Chapter 5, is the following: given that a popular diagnosis (esp. from Temkin) for why ‘better than’ or ‘more reason than’ is intransitive is that some values or reasons are essentially comparative. Contrastivism looks like a view on which reasons are essentially comparative, and so might look like a view that proponents of intransitivity would like. But in Chapter 4, I develop the view so that it rules out the intransitivity of reasons (i.e., the view says that if r is a reason for A rather than B, and a reason for B rather than C, then it is also a reason for A rather than C). Proponents of intransitivity might then be disappointed. So one aim of Chapter 5 is to show that they need not be too disappointed, since the kind of intransitivity they care about is different from the kind I rule out, and with some optional amendments (namely, what Chrisoula calls VARY), we can allow for ‘more reason than’ to be intransitive. Chrisoula has correctly pointed out, though, that given the kinds of arguments proponents of intransitivity tend to give, using multidimensional objectives, even the transitivity of contrastive reasons is called into question.