Non-Consequentialism and Strength of Obligation

Walter Sinnott-Armstrong has given us a new argument for
consequentialism (“How strong is this obligation?  An argument for consequentialism from
concomitant variation,” Analysis 69
(2009): 438-442).  The datum: other things kept equal, the
obligation to keep a promise to take a friend to the airport is stronger than
the obligation to keep a promise to meet a friend for lunch—overriding the
latter obligation takes less than overriding the former.  The best account of the source of this
relative strength is that (in normal circumstances) the consequences of
breaking the first promise are more harmful than the results of breaking the
second.  If we generalize, it appears
that the relative strength of all obligations depends on the level of harmful consequences
that would result from violating the obligation.  Thus consequentialism seems to best account
for strength of obligation.

Sinnott-Armstrong [WSA] recognizes that this argument
“assumes that (i) the strength of the moral obligation does not explain the degree
of harm…, (ii) no third factor can explain the strength, the harm, and their
correlation…, and (iii) the correlation is not accidental…” (440).  I think that assumption (ii) is not
well-founded.  More basically, we should
question WSA’s ultimate assessment, that in order to respond to his argument,
“deontologists need to explain why some moral obligations are stronger than
others without invoking the harmful consequences of violating those moral
obligations” (442).  There is another
response available for deontologists, one that does invoke harm, and it
undermines assumption (ii).  

Consider that it used to be considered difficult for
consequentialism to account for some commonsense duties, like the duty to keep
promises.  WSA’s reply to this objection
is to maintain that the remote non-optimal consequences of promise-breaking
mean that most promises must be kept, and that one is not is obligated to keep
one’s promise if breaking it would have optimal consequences.  This won’t persuade people, like me, who
think that there are cases in which it is wrong to break a promise even if
breaking it would have optimal consequences. 
But consequentialists have another reply to the objection that they
can’t account for the obligation to keep promises: they can ‘consequentialize.’

Start with a different illustration.  It is now relatively well-established that
consequentialists can, contrary to former opinion, account for rights, by
working rights into their rankings of states of affairs.  Consequentialism requires that agents produce
the best state of affairs, but that requirement is in its bare essentials neutral
with respect to what counts as the best state of affairs.  So if we intuitively think that people have
rights, consequentialism can remain intuitive by saying that states of affairs
in which rights are preserved and respected are better than states of affairs
in which this is not the case.  And this
is just one instance where the consequentialist can formulate a substantive theory
of goodness to make consequentialism align with intuitive morality.  It can also, to return to the earlier
example, say that states of affairs in which promises are kept rank higher
(other things equal) than those in which promises are broken.  So here is how Doug Portmore (2007,
39) puts the general formula for consequentializing: “Take whatever considerations that the
non-consequentialist theory holds to be relevant to determining the deontic
status of an action and insist that those considerations are relevant to
determining the proper ranking of outcomes.” 
Following Portmore, we can apply this recipe across the board and get
the Deontic Equivalence Thesis: any
two remotely plausible ethical principles can generate identical sets of
deontic verdicts, verdicts about which actions are required, optional, or
forbidden.

Of course, this thesis maintains that what’s fair for
consequentialism is fair for non-consequentialism as well: if consequentialists
can hijack the non-consequentialist’s deontic verdicts by manipulating rankings
of outcomes, so non-consequentialists can hijack the consequentialist’s deontic
verdicts by manipulating the content of deontological fundamental
principles.  For instance, Portmore
(59-60) points out that Kantians can take whatever factors the consequentialist
thinks are relevant to the ranking of states of affairs and say that those
factors are also relevant to determining whether actions treat humanity as an
end-in-itself.  A Kantian who stays up
late at night being tempted by utilitarianism could even say that the only way
to treat humanity as an end-in-itself is to maximize utility. 

The very presence of this recipe for hijacking other
theories’ verdicts means that it is false that deontologists must not invoke
considerations of harm in their accounts of why one duty is stronger than
another.  They can, and should, be happy
to assert that the reason that your duty to keep your promise to take your
friend to the airport is stronger than your duty to keep your promise to meet a
friend for lunch is that failure to keep the first promise would be more
harmful than failure to keep the second. 
After all, deontologists can say, e.g., that we have a right not to be harmed, or that harming
people fails to treat them as ends-in-themselves.  More than just being able to say such a
thing, deontologists do say such
things.  This is one reason why it
wouldn’t suffice to reply that real deontological
theories cannot deontologize (or that real
consequentialist theories cannot consequentialize).  Paradigm examples of deontologists (Kant,
Ross) have prioritized harm-avoidance
as a key duty. 

Since deontology can import considerations of harm, it is
false that “deontologists need to explain why some moral obligations are
stronger than others without invoking the harmful consequences of violating
those moral obligations.”  They can and
do appeal directly to harmful consequences. 
And, therefore, for all that WSA has demonstrated, it might still be
that some third thing does explain
the correlation between harm and strength of obligation.  For instance, if Kantianizers are right, then
treating people as ends-in-themselves might be that third thing.  That is, both a breaking a promise itself and harming a person by breaking a
promise fail to treat the promisee as an end-in-herself, so when the agent must,
say, cause one of two harms by breaking one of two promises, it is obligatory
to avoid causing the greater harm, as that constitutes a more significant
violation of the foundational principle to treat people as ends-in-themselves.  In this way, the strength-of-obligation
argument, in its appeal to harm, does not marshal considerations that are incompatible
with deontology.  Since harm is relevant
to such deontological right-makers as treating persons as ends-in-themselves,
deontologists can let considerations of harm determine strength of obligation,
just as much as consequentialists can let the respecting of rights factor into
the rankings of outcomes. 

(Note that this way of answering WSA also shows that
deontology can manifest the same theoretical unity that he thinks is a virtue
of consequentialism: just as his consequentialism appeals to one factor (harm)
to ultimately explain both the existence and the strength of obligation, so
some kinds of deontology can appeal to one factor, such as treatment of persons
as ends-in-themselves, to ultimately explain both the existence and strength of
obligation.  Sometimes, of course, these
ultimate explanations will depend on intermediate links, such as harm for
deontology and, in one of WSA’s examples, broken trust for consequentialism.)

14 Replies to “Non-Consequentialism and Strength of Obligation

  1. I agree that (ii) seems to be the weak point but I wonder if there’s a simpler way to incorporate harm to deontology in a way that would match WSA’s data. One way to think about this is to think that promises are made and accepted within a social practice. This practice is constituted by a certain rules. This rules specify what sort of excuses and exceptions there are to the obligation to keep the promise. The promisors and people promised to who have to accept the promise tacitly assume that the promise is governed by these rules. When I promise to take someone to the airport, I understand that it’s not wrong not to do whatever happens just like the person I make the promise understands this.
    Now, it seems like the rules we have in the promising practice are sensitive to how harmful the consequences of breaking the promise would be. But, they are not only sensitive to that.
    And, we can think of different kind of justifications for the rules. One might be rule-consequentialist, other might be based on the idea that it is our practice, other might be contractualist, yet another might be Kantian, and so on.
    In any case, on this view, it is not the harmful consequences per se that explain the difference in the obligations, but rather the antecedent rules of the practice in which we make promises. So, I cannot really see the force of the argument.

  2. Jussi,
    I’m not sure that I’m seeing how you’ve incorporated harm in a simpler way. It sounds to me like you’re saying that harm is not always part of the explanation of strength of obligation, and that rules, and non-consequentialist rationales for the rules, determine strength of obligation in a sometimes harm-independent way. If I were to extrapolate from WSA’s argument, I suspect he might say that considerations of harm provide the most unified explanation of all relevant considerations, including what rules we have, what exceptions we allow to those rules, etc. (I encourage anyone interested to just take a look at his paper for the details of this argument.) You’re right that I’m granting him that premise, to focus on a different point: even if harm alone proximately determines strength of obligation, it’s not the case that only consequentialism can make sense of that, since harm might function as an intermediary between strength of obligation and some non-consequentialist factor (be it treating persons as ends, immunity from reasonable rejection, or whatever). As to the premise you target, it might be vulnerable in other ways, too, since it relies on induction–that harm will always determine (immediately or intermediately) strength of obligation.

  3. Josh,
    thanks. I have to check the paper. WSA’s potential response seems shaky to me.
    Here’s a worry though about your response. The consequentialising project – creating a consequentialist equivalent of a non-consequentialist theory works only if consequentialism is empty, if it doesn’t provide any normative friction so to say.
    This means that going the other way round is only going to work if the other, deontological theories are normatively empty. That is, it requires that causing the greater harm can always necessarily be what constitutes not treating someone as an end in themselves.
    However, with my Kantian hat on, I don’t think the Kantian should accept this. I think they should hold on to the idea that the categorical imperative (or any other equivalent theory) is a substantial view that has substantial normative results independently of which action causes the greater harm. And, if you think that, then it would be an incredible co-incidence that the Kantian view would be necesarily coextensive with the consequentialist view.
    So, think of a case in which I can by harming someone’s leg avoid causing someone else a permanent brain-damage. The consequentialist of the kind who compares mere harms would need to say that this is the right thing to do. Of course, the traditional Kantians would say that this constitutes using someone as mere means. So, anyone starting the Kantianizing project would need to argue that the CI does not have such substantial consequences and that it can be stretched to cover the consequentialist result in this case. This would require arguing that Kantians have been mistaken about the consequences and substantiality of their own view.
    I think this would be a loss for the Kantians, just as I think the consequentializing project is a problem for consequentialism.

  4. I need to read the article, but it’s a weird argument. The examples I use to explain Ross to my students have roughly the structure of the cases that WSA thinks motivate consequentialism. It may be that considerations having to do with fidelity rule out various options so that we have two to consider and there’s a p.f. duty of fidelity to go to the airport and to go have lunch. The considerations having to do with benefits and harms then ‘break the tie’. In other contexts where two options would bring about equal amounts of harm/benefit to different parties, considerations having to do with fidelity could break the tie. On its face, it seems better when Ross runs the argument than WSA but I don’t have a copy of the Analysis article so I’m probably missing a lot.

  5. Clayton,
    The part of WSA’s article that addresses your concern (which, as your post reveals, I didn’t much discuss in my original post) has to do with parsimony and unity. He writes, “Why would one factor [in your example, fidelity] determine whether any moral obligation at all exists, while a completely separate factor…[here, harm/benefit] determines how strong that moral obligation is?” I’m not sure that a Rossian will be very impressed by the principle that a theory with one factor explaining both strength and existence of obligation is better than one with separate factors for each, but that’s WSA’s claim.
    Jussi,
    I don’t agree that consequentializing makes consequentialism empty or that deontologizing makes deontology empty. Each still offers a distinctive theory of what makes right acts right and wrong acts wrong, even if they have identical deontic verdicts (verdicts as to which acts are right and wrong). Regarding your example of harming and Kantianism, note that it doesn’t actually imply that traditional Kantianism isn’t substantively committed in a way that excludes that kind of Kantianizing. Traditional Kantianism includes the foundational principle “Treat others as ends in themselves,” plus some substantive “middle principles” (as some call them) about what constitutes treating others as ends in themselves, just as traditional consequentialism has, say, welfarist theories of value that make it incompatible with some consequentializing projects. So the consequentializing and deontologizing projects should just be understood as locating verdictive flexibility in the most fundamental principles of each, which can be, but need not be, combined with substantive “middle principles” that prevent the kinds of losses you worry about and that are congruent with conventional understandings of each. In this way, Kantians have not been mistaken about the substantiveness of their own views, since their views usually include both the fundamental principle and certain substantive commitments as to what it is to treat a person as an end in herself. Note that Kantianism isn’t on this view necessarily coextensive with consequentialism; it is coextensive only to the extent that we import certain middle principles into each of the two theories.

  6. Hi Josh –
    Great analysis. Makes me want to pick up Analysis. A few thoughts:
    1. Your response depends on the thought that acts can vary in how much or how well they treat people as ends in themselves. It sounds funny to my ear to speak of this as a matter of degree. Either an act treats me as an end in itself or it doesn’t. Might Kantians explain strength of obligation better by appealing to the different extents to which acts affect rational autonomy?
    2. I’m not so confident that deontologists can ‘deontologize’ any non-consequentialist consideration as easily as consequentialists can ‘consequentialize’ any deontological consideration. The example you mention, rights violations or affirmations, can be consequentialized fairly easily by reconceptualizing them as value-bearing states of affairs. But it’s less clear to me that some consequentialist considerations can be so easily deontologized. Suppose that there is an obligation to ø which is highly dependent on the consideration that ø-ing will provide pleasure to large numbers of people. Let’s call the strength of this obligation S. Whether that consideration can be deontologized may depend on the kind of deontologist you are, but how, for instance, would a Kantian show that the obligation to ø is S-strong because ø-ing fails to treat people as ends in themselves to degree S? Denying people pleasure does not obviously fail to treat people as ends in themselves at all.
    3. Kantians and deontologists should answer criticisms like WSA’s, but in some respects, I’d like deontologism to have some normative friction, as you put it in response to Jussi. WSA and other consequentialists think it advantageous that their views can consequentialize non-consequentialist considerations, but I’ve wondered whether, rather than being an advantage, this doesn’t just show that “consequentialism” is trivial. In the universe of possible rightmakers, consequentialism throws an all-comers party, refusing to tell us which of the candidate rightmakers really are rightmakers. In contrast, I’d like deontologism to insist that it has the right account of rightmakers and that consequentialism is sometimes flat out incorrect about certain considerations being rightmakers at all. I’m having trouble coming up with examples at the moment, but aren’t there cases where deontologists will insist that certain states of affairs consequentialists think are relevant to rightmaking are actually irrelevant? So irrespective of the effectiveness of your reply to WSA, I kind of don’t want it to work.

  7. I have to read Walter’s argument but on the face of the claim that the strength of deontic constraints always varies with harm seems to me to be wrong. In my paper “How is the Strength of a Right Determined?: Assessing the Harm View,” American Philosophical Quarterly, October 1995, I argue that harm can’t be the only factor on which a right’s strength depends. I think that there are clear cases of acts that are seriously wrong that cause no harm. Indeed, they might even benefit the right’s bearer and still be wrong.
    @Michael: There are lots of cases where consequentialist right makers seem beside the point. Police take a nude picture of a woman who is arrested and show it around other police stations. The deontic status of that act in no way depends on how many cops see the photo or on how happy it makes them in aggregate. In this case the happiness the picture causes seems beside the point. It’s not getting less and less wrong as it brings about more and more happiness.

  8. Michael, your #1 is a good point–maybe that’s a better way to go.
    I’d want to resist your 2 and 3, though. I agree that on certain (traditional) substantive conceptions of treating a person as an end in herself, denying people pleasure might not count as a failure to so treat a person. But we can specify it differently, such that it is a failure to treat a person as an end in herself. And, as I said in reply to Jussi, I think that this doesn’t make it trivial. Consequentialism does tell us what its right-maker is: good consequences. It just leaves it open how to specify that. I think that there are plenty of cases where good consequences are irrelevant to determining rightness (I’m a Kantian, after all), but I can’t get there by pointing to counterintuitive deontic verdicts of consequentialism, since I think that consequentializing makes sense. Instead I think we have to appeal to intuition elsewhere (another story for another day, no doubt).
    Sam, this ties in with your reply to Michael. Classical utilitarianism seems to have counterintuitive right-makers, as in the case you give, but that’s not to indict consequentialism more broadly. The consequentializer will just say outcomes in which naked pictures of people are broadcast without their consent rank lower than outcomes without such phenomena. So, the right-maker posited by consequentialism per se is does not appear to be beside the point, for whatever case we throw at it.
    However, your APQ paper seems to be right on point. I must confess that this paper slipped under my radar, but the premise that you apparently take up and that I’m granting here to WSA — that strength of obligation covaries with level of harm — is at the center of WSA’s argument.

  9. Josh,
    I think there is a problem with your response. Imagine that instead of consequentialism, we had WSA* who argued for orangeism. On this view, only touching anything orange is wrong.
    Now, you come along and say to WSA* that Kantianism has verdictive flexibility – it needn’t be combined with the traditional middle-principles. So, you can say that touching anything orange is wrong because it is failing to treat humanity in others and in yourself as an end in itself and that makes it wrong. At this point, the obvious question is how could *that* be wrong treating others as mere means? We ask this question because the phrases treating others as ends and as mere means have some substantial meaning.
    And, that’s exactly my feeling when I hear the claim that causing the greater harm would constitute using someone as mere means. How could that be given what the phrase ‘using someone as mere means’ means? So, there must be some plausibility to the idea that avoiding harmful consequences is what it is to use someone as mere means that fits what the phrase means.
    Note that the situation with consequentialising is similar. When in this project we give a consequentialist version of some other view we are going to give an account of what is good. That account better be a plausible view of goodness. But, goodness seems to be more flexible in this sense. It carries less restrictive prior meaning than phrases like using people as ends in themselves and as mere means.

  10. Hi Josh,
    Interesting post. I have only two brief comments. First, it seems that one crucial premise in Sinnott-Armstrong’s argument is the following: “If we generalize, it appears that the relative strength of all obligations depends on the level of harmful consequences that would result from violating the obligation.” But why does Sinnott-Armstrong think that we should generalize? It seems the relative strength of my obligation to aid a person depends not just on the harmful consequences that would result from my failure to aid him or her but also on my social proximity to that person. Even when the potential harm is the same, I have a more stringent obligation to save a friend than to save a stranger, and an even more stringent obligation to save my daughter.
    Second, you might want to contact Ben Sachs. He has a paper that makes a similar point about how consequentializing is a double-edged sword. He doesn’t use this point to respond to Sinnott-Armstrong or any argument of the sort, but it’s something you might want to cite if you’re thinking of developing this into a paper responding to Sinnott-Armstrong.

  11. My response is in the same neighborhood as Doug’s. Everybody should allow that consequences make some moral difference; the question is whether they make all the difference. It follows that everyone should allow that consequences will make a difference, as economists say, at the margin: of two otherwise similar cases, a difference in consequences will make a moral difference. But it just doesn’t follow that consequences make all the moral difference.
    (A young woman with two otherwise similar romantic interests finds herself more attracted to the more handsome man. Furthermore, upon reflection, she thinks that the handsomer a hypothetical man was, the more she would be attracted to him. It doesn’t follow that her romantic attraction is exhaustively explained by physical attraction.)

  12. Doug and Heath, WSA doesn’t give us much of an argument for generalizing that the consequences, and harmful consequences in particular, are what always matter. In fact, he doesn’t even really spell out that he is generalizing. What he does do, though, is go through a few cases and explain how he traces the morally relevant features to harm. In particular, he explains that he thinks that friendship affects the strength of the obligation to keep the promise insofar (and only insofar, presumably) as it affects the consequences of breaking the promise (p. 440), and although this goes unsaid, his theory requires him to more specifically restrict the impact of friendship to how it affects the harms and benefits of breaking the promise.
    I agree with the two of you that other factors besides harm and benefit seem to affect rightness and wrongness, but WSA’s argument suggests that, for each case we present to show that those other factors matter, he will give a story about how they matter only insofar as they affect harm and benefit (and how that can preserve most of our intuitive judgments). So we’re left with a mere standoff of intuition, which is very real but dialectically weak, and this is why (I think) it is important to observe that even if we give him his premise about harm, it still isn’t sufficient to constitute an argument for consequentialism.
    (Doug, I have no plans for an in-print paper replying to WSA — this post is it! But thanks for the tip on Sachs’ paper, which sounds interesting.)
    Jussi, it sounds in your latest post like you want to say mostly that the principle to treat persons as ends in themselves has some substance, enough to rule out deontic equivalence with orangeism. (Does using orangey views as illustrations in our posts require us to tip our hats to Mark Schroeder?!) I agree with that. I also think that consequentialism is not plausibly construed as capable of being deontically equivalent with orangeism. The Deontic Equivalence Thesis says only that remotely plausible ethical theories can be made deontically equivalent. So, you’re right that the principle, treat people as ends, is not entirely without substance — I wholeheartedly agree with that and never meant to deny it. What I do want to deny is that it has enough substance to be incompatible with any
    plausible deontic intuitions a consequentialist might have, and, mutatis mutandis, the same goes for consequentialism and plausible intuitions that a deontologist might have. In fact, the history of Kantianism makes transparent that there is at least some verdictive flexibility in the principle–e.g., many contemporary Kantians strenuously disagree with Kant himself about what constitutes treating oneself as an end in oneself. More generally, the mere meaning of “treat people as ends in themselves” surely doesn’t determine every deontic verdict by itself. Middle principles are essential to get applications from the fundamental principle, because the meaning of the fundamental principle is so thin. (Put in the lingo of some other conversations, the concepts involved are consistent with a wide variety of conceptions of treating people as ends in themselves.)
    So, given that there is some verdictive flexibility in both views, the only question, for the purposes of evaluating whether deontologizing has greater or fewer dialectical resources than consequentializing, is whether one is more flexible–or less semantically constrained, to use the language of your last post–than the other. I’m not seeing a compelling argument for thinking that the Kantian principle has more “restrictive prior meaning” than the consequentialist principle, and specifically I’m not sharing the intuition you apparently have that causing someone harm doesn’t treat them as means (at least in normal cases). But it would be interesting to learn of any other such arguments.

  13. Here’s a case from Parfit’s brilliant chapter on treating people as mere means:
    “Attempted Murder, when Brown attacks me with a knife, trying to kill me, I save myself by kicking Brown in a way that predictably breaks his leg.
    Though I am harming Brown as a means of stopping him from killing me, I am not treating Brown as a means.”
    That sounds right to me.

  14. I agree, Jussi, that although kicking Brown in this way treats him as a means, it doesn’t treat him as a mere means. (But that’s not a “normal” case, right? In any case…) Don’t we come to this judgment, though, because you and I (and Parfit) are working with standard conceptions of treating people as mere means? What would you say to the person who says that you are treating Brown as a mere means, but that whatever wrong is in this treatment is overridden by a more pressing obligation to stop you from being treated as mere means? Such a person might have a non-standard conception, but they aren’t literally contradicting themselves, are they? I don’t think that they would actually be abusing the language of “mere means” to say this. So it might sound strange to traditional Kantians to go with a reformist Kantianism, but that’s to be expected, just like the reformist consequentializer will say some things that are strange to the traditional consequentialist. The question is whether in saying those strange things, the reformer is contradicting themselves or no longer remaining a Kantian or consequentialist (or whatever they claim to be).

Comments are closed.