(Cross-posted from Hands and Cities)
This post describes what I see as a basic but powerful objection to treating certain deontological distinctions as justified via some intrinsically important moral difference they reflect or respond to. In brief, the objection is that the distinctions in question are not, in the right way, important to the potential victims of harm. I focus on the doing vs. allowing distinction to illustrate the point, then briefly discuss how I think the objection does and does not generalize to other features of non-consequentialist ethics.
I think of this objection as “basic” in the sense that it seems to me quite natural and immediate, and a number of people have indicated to me that something closely related is core to their own skepticism towards deontological justifications of the relevant type. My own thinking on the topic stems mostly from many hours of conversation with Ketan Ramakrishnan, whose work (for example, the last section of his 2016 paper “Treating People as Tools,” as well as some unpublished work) treats closely related worries in depth. I expect that the type of objection I describe is discussed elsewhere in the philosophical literature as well, perhaps extensively, though other examples of in-depth treatment aren’t, at the moment, coming to mind (this isn’t a literature I’ve gone particularly deep on).
I. Doing and allowing
Many philosophers aspire to validate some sort of moral distinction between doing and allowing harm. Here’s an example.
- Case 1: Maria is driving to the hospital on a country road with no other traffic, and she needs to get there soon, or she will die. On the side of the road, she sees Wilfred, who is hurt, and who will die if he doesn’t get help. If Maria stops to help Wilfred, though, she won’t make it to the hospital on time, and will die herself.
- Case 2: Again, Maria is driving to the hospital, and needs to get there soon on pain of death. And again, she encounters Wilfred. This time, though, Wilfred is healthy, but immovably stuck in the middle of a narrow bridge, such that the only way for Maria to get to the hospital on time is to run over Wilfred’s body, killing him.
In both cases, Maria must choose: her life, or Wilfred’s. But the cases seem different along some important dimension. In particular, Maria’s running over Wilfred in Case 2 seems intuitively impermissible; whereas in Case 1, it seems plausibly permissible for Maria to save herself, rather than Wilfred, or to do something like flip a coin. And similarly, we might think it permissible for a bystander to intervene to stop Maria from running over Wilfred in Case 2; but not to force her to help him in Case 1.
(This sort of case is discussed in Kamm’s Intricate Ethics, p. 131. I think this particular contrast may have been first suggested to me by Ketan Ramakrishnan.)
These cases in particular might seem fanciful and unrealistic, but distinctions in this vicinity actually come up a lot — indeed, a disturbing amount — in everyday life. Donation choices are an obvious example: we think very differently about failing to make a donation that would save someone’s life then we would about actively killing them. And doing/allowing distinctions in extreme ethical situations seem continuous, at least to some extent, with distinctions between “problems I create” vs. “problems that I have an opportunity to help with” in more mundane ethical situations: e.g., you have a strong responsibility to clean up messes you make, and a weaker responsibility to clean up the messes of others, etc.
We can loosely distinguish between at least two ways of attempting to justify (as opposed to debunk) distinctions of this kind.
- Pragmatic justifications: These sorts of justifications appeal to the costs, benefits, and pragmatic feasibility of adopting and upholding different practices, using different cognitive heuristics and decision-procedures in the midst of uncertainty and fallibility, dividing up responsibility in different ways, placing certain kinds of demands/expectations on ourselves and each other, allowing certain forms of coordination, verifying and enforcing certain types of norm-compliance, supporting different types of trust and insight into each other’s character and dispositions, etc. A paradigm example here would be something in the vicinity of rule-consequentialism.
- Intrinsic justifications: These attempt to identify a moral principle that assigns intrinsic significance to the distinction in question — where this principle is not itself rooted in some further pragmatic justification. A paradigm example here would be something in the vicinity of Rossian pluralism — a view to which I tend to view much of contemporary non-consequentialist ethics as a successor. A basic (if somewhat uninformative) version of such a view might say something like: you have a stronger pro tanto claim/right that people not actively harm you than that they prevent harm from befalling you, and this is just a brute feature of moral universe, which admits no deeper explanation.
My focus here is on objecting to intrinsic justifications. I’ll note, though, that the lines between pragmatic and intrinsic justifications can be blurry, and some approaches may not fit neatly into either bucket (for example: I tend to think of “veil-of-ignorance” type theories as having more of a pragmatic than an intrinsic flavor, but it’s not super clear how they should be categorized). That said, it should be possible to evaluate whether the objection I want raise applies to a theory, regardless of how it’s categorized.
II. What do we want from an intrinsic justification?
One can object to intrinsic justifications of doing vs. allowing in a variety of ways. One salient possibility is to try to undermine our sense that there are criteria available for distinguishing between doing and allowing that fit our intuitive verdicts about cases (this sort of objection applies to pragmatic justifications to some extent too, but intrinsic justifications seem, naively, more committed the substantiveness of the distinction). What exactly makes one harm a “doing,” and another an “allowing”? Suppose, for example, that your financial advisor has mistakenly concluded that you wish to donate money to save the life of someone in the developing world, and is on his way to make the donation on your behalf when you call to correct him (see McMahan (1993)). Is this a doing, an allowing, or something else? Did you kill the person in this case, even though your failing to donate would normally have been a “letting die”? (If so, be careful about setting up automated donation payments — cancelling them later will higher-stakes than setting them up in the first place. And be careful, as well, weighing in on grant applications that would’ve been approved without your input).
Or suppose that an actor in a play fails to show up for the performance she’s supposed to star in; instead, she just sits at home (see Foot (1978)). Did she ruin the performance, or did she merely “allow it to be ruined”?
The philosophical literature features a proliferation of further distinctions that attempt to capture the subtleties that examples and counter-examples of this type make salient (see the SEP article on Doing and Allowing for more examples). Foot, for example, distinguishes between initiating, sustaining, enabling, and forbearing to prevent some harmful sequence; McMahan, between the relevance of factors like whether a given type of aid or protection was terminated or not, self-sustaining or not, already operative or not. As the complexity of the conceptual landscape grows, one begins to wonder about the viability of the exercise as a whole. And one might suspect, more broadly, based on a certain kind of direct reflection on the nature of action, that ultimately “doings” vs. “allowings” will not end up a particularly salient or “jointy” joint in nature, to which intrinsic justifications can readily attach (though exactly how much not being a deep joint in nature undermines assigning a distinction moral significance is a further question).
This isn’t the type of objection I want to focus on here, though, so for simplicity, let’s assume that some criteria for distinguishing between doing and allowing have been identified, and we have a principle before us that fits our intuitive judgments about cases fairly well. (Alternatively, we can assume that we’ve decided we’re OK with leaving the concepts of “doing” and “allowing” unanalyzed, in the same way we might leave the concepts of “cat” and “dog” unanalyzed in thinking about principles like “cats are more important than dogs.”)
As a number of philosophers have emphasized, possessing a principle that fits the cases well isn’t enough: we also want the principle in question to look morally attractive, resonant, and sensible on its own terms. Thus, Kamm (2007) — a philosopher generally known for her commitment to doing justice to fine-grained intuitions about cases — writes that once you have identified a principle that fits the cases, the next step is to:
“… consider the principle on its own, to see if it expresses some plausible value or conception of the person or relations between persons. This is necessary to justify it as a correct principle, one that has normative weight, not merely one that makes all of the case judgments cohere” (p. 5).
As an example: Ramakrishnan (2016, p. 161) suggests that this step fails for moral principles on which someone’s distance from you in physical space has brute and intrinsic moral significance for how much you should help them, even if such principles fit our intuitive verdicts about cases fairly well (see Kamm (2007), Chapter 11, for discussion). That is, it seems to Ramakrishnan directly implausible to posit as a basic moral fact that physical distance matters in this way; if it does, one is inclined to say, there must be some deeper explanation of this fact — a demand for deeper explanation that Ramakrishnan notes he is less inclined to make of something like “pleasure is intrinsically good.” And he worries that his own account of claims against being used for the purposes of others is subject to a similar objection, since he suspects that this account will ultimately assign brute moral significance to facts about the causal structure of choice scenarios — facts that Ramakrishnan finds lacking in inherent significance when he steps back and reflects.
My central worry about intrinsic justifications of doing/allowing distinctions is that the type of abstract moral resonance Kamm and Ramakrishnan are looking for here isn’t going to be forthcoming, even if we can get all the cases right. Here’s why I’m skeptical.
III. Shouldn’t the distinction matter to the victim?
Let’s return to Wilfred in the cases above. And let’s posit that Wilfred himself does not care at all about whether his death is “done” or “allowed”; whether someone kills him or merely lets him die. Thus, for example, when Wilfred considers a future in which he is crushed by an avalanche while a bystander who could’ve saved him looks on, vs. a future in which he is crushed by an avalanche that a bystander actively caused, he feels no preference between the first vs. the second; he would give up nothing he values in order to die in the first way rather than the second way. Rather, his objection to dying is always and only that he dies; that he is torn away from life, family, love, beauty; that for the rest of eternity, he will not exist. He cares about what he loses, not about how it gets lost. It’s the avalanche that really matters to Wilfred, not the bystander.
And now suppose that Maria, approaching Wilfred from a distance on the road, cannot make out whether she is about to encounter Case 1 or Case 2. As someone committed to behaving morally, though, she knows (assuming the relevant distinction between doing and allowing holds) that her own life is at stake in the difference. If she is in Case 2, she will be obligated to die, rather than to run over Wilfred. If she is in Case 1, she will have the option to live, or at least to flip a coin. Clearly, something very important is at stake in the difference between these two cases — something important enough to compel Maria to give up her own life.
But now it’s natural to ask: important to whom? Not, we have hypothesized, to Wilfred. Wilfred doesn’t care about whether Maria’s actions with respect to him are doings or allowings; he does not take himself to have any stronger interest in avoiding death by the former vs. the latter. His interest, he thinks, is in avoiding death, period.
Perhaps, we might say, Wilfred is mistaken about this. It should matter to him whether he is killed vs. allowed to die, just as it should matter to a depressed (but soon to be cured) person whether they fall out a window, even if it doesn’t seem like it to them at the time. But this doesn’t seem right to me. That is, it doesn’t seem true to me that Wilfred has reason, and especially not strong reason, to care about whether the causal sequence leading up to his death is a doing or an allowing; to hope for one vs. the other; to invest resources steering the future in one direction vs. the other. And it doesn’t seem like other people, concerned about Wilfred’s interests, have reasons in this vein either. Someone spending money trying to make sure any avalanche that crushes Wilfred is allowed rather than caused doesn’t seem to me like they’re doing Wilfred any paternalistic favors; give the money to his kids instead.
Why, then, must Maria die in the one case, but not the other? Intuitively, one would’ve thought that the explanation would be about some interest or justified concern of Wilfred’s at stake in the second case, but not the first. It is Wilfred, after all, who is getting run over, or not, in Case 2; Wilfred who, the doing/allowing distinction assumes, could blame Maria is she runs him over, but not if she merely lets him to die; Wilfred on whose behalf Maria must sacrifice her own life. But this makes it seem very strange that Maria’s killing in Case 2 — the killing she must die to avoid performing — would be no worse for Wilfred than her letting him die in Case 1 — a letting die she need not die to avoid. If it’s no worse for Wilfred, why is the difference worth dying for?
(Note that it’s possible to say that even though being killed is no worse for Wilfred than being allowed to die, nevertheless Wilfred has reason to care about whether he is killed vs. let die — e.g., he has reason to care about something irrelevant to his interests, or at least his self-interest. For example, we might think that Wilfred has reason to care about whether he is wronged, even if the wrong is not worse for him — after all, he blames Maria for killing him but not letting him die, so this seems like a justified difference in concern. But attempting to ground the moral difference between killing and letting die in Wilfred’s justified concern about not being wronged, where one assumes killing is wrong but letting die is not, smacks strongly of circularity. Our question, after all, is about what justified concerns of Wilfred’s could ground the moral difference in question; justified concerns that only exist granted that this difference exists seem poor candidates for such an explanatory role.)
Remember that we’re setting aside “pragmatic” explanations here — e.g., explanations that appeal to the benefits, in general, of different social practices, cognitive heuristics, divisions of responsibility, etc — whether for people in Wilfred’s type of predicament or Maria’s. Instead, we’re looking for something intrinsically important about the difference between Case 1 and Case 2.
IV. Maybe it matters to the chooser?
Perhaps this difference is intrinsically important to Maria, in the sense of responding to or protecting some important interest or justified concern on her part? Again, I’m skeptical: it seems hard to tell a plausible and non-pragmatic story about what the interest in question could be. The most salient feature of the doing/allowing distinction, after all, is that it makes it especially hard to justify killing Wilfred — so hard as to compel Maria to die — and it’s hard to see why this justificatory burden would be somehow rooted, ultimately, in Maria’s interests or concerns (and whatever these concerns are, it’s hard to see Maria valuing them over her life as a whole). And more broadly, it really just seems like the story about why Maria can’t kill Wilfred should be, as it were, Wilfred-focused.
(If we think of the “default” moral scenario, absent the doing-allowing distinction, as one in which both lives are treated as equally important, such that Maria, in both cases, should do something like flip a coin, then one option for appealing to Maria’s interests is to imagine a kind of “package deal,” in which Maria gets an agent-relative permission to save herself in Case 1, in return for granting a constraint on killing Wilfred in Case 2. But this, to me, starts to take on something more like a “pragmatic” flavor; and it’s not clear why we would think of the permission and the constraint as packaged together anyway.)
V. Something else?
A final option for intrinsic justifications is to just posit some intrinsically significant difference between Case 1 and Case 2, but where this significance is not grounded in any interest or justified concern on the part of either of the parties involved. We might, for example, say that Wilfred has a stronger claim against being killed than being let die, even though he has no reason to care about whether he is killed vs. let die (my best guess is that something like this is the standard view amongst non-consequentialist philosophers, even if it’s not stated explicitly). Rights and claims, in this world, aren’t beholden to things like what actually matters to people. Rather, they are just, as it were … there. They’re the rules of the game, the furniture of the moral universe. We might not like them. We might wish we could change them. They might compel us to destroy things of beauty and value for reasons that seem, from the perspective of what we actually and justifiably care about, arbitrary and mysterious on reflection. But thems the rules.
This seems like an objectionably alienated conception of morality to me. If differences in rights/claims are justified neither by appeal to something directly significant to any actual moral patients/agents, nor by some pragmatic story about norms, heuristics, coordination, division of responsibility, etc, then it’s just very hard for me to see why they would have any weight for us, the actual humans who are supposed to obey their dictates. On whose alter must Maria be sacrificed in Case 2, but not Case 1, if not Wilfred’s? Whence the rhythm of this deontological dance? If morality isn’t about the interests/rational concerns of moral patients/agents, or about finding and upholding ways for those patients/agents to live well together, then what is it about?
In my experience, many non-consequentialists normative ethicists, especially in the Rossian pluralist tradition, proceed without deep and satisfying answers to this question. For them, first-order intuitions about cases are the most important data, and the central project is looking for principles that capture these cases. Most of the action is in this back and forth, between data points and initial efforts at fitting (over-fitting?) the curve; the most influential objections are counterexamples. True, one hopes that the eventual curve will have a deep and morally resonant explanation, and sometimes one steps back to wonder; but the project of explaining the curve is generally secondary to the project of fitting it, and one is prepared, if necessary, to take or treat the resulting principles as brute — justified simply by their ability to capture our first-order intuitions, even if there is little else to say to make sense of them.
Because one stops demanding, at least immediately, that one’s principles make abstract sense, this methodology can lead, at least in my experience, to a type of alienation akin (though not identical) to the type I discussed at the end of my post on helium-maximizers. That is, one begins to feel and act as though morality could end up quite alien and arbitrary. You might wake up in Hell, and learn that it was because you set up an auto-pay to a charity and then cancelled it, instead of just not setting it up in the first place. “What?” you say. “Why does that matter?” “It just does,” the demons reply. “Thems the rules.” Morality, here, can’t be anything (e.g., maximize helium): it needs to fit with our intuitive judgments about cases, which is a very substantive constraint. But it doesn’t need to make any kind of deep or abstract sense on reflection. This seems to me pretty unappealing (though of course, “X is unappealing” is distinct from “X is false”).
VI. Generalizing the objection
I’ve discussed three categories of intrinsic justification for the difference between Case 1 and Case 2: appeal to some interest/concern of Wilfred’s, appeal to some interest/concern of Maria’s, and appeal to something else. The “something else” category is clearly broad, and I don’t claim to have done it justice (for example, we might also attempt to appeal to something about the relationship between Maria and Wilfred that Maria’s killing him sets up, though my suspicion is that this will end up uninformative/circular as well. My thanks to Ketan Ramakrishnan for suggesting this category of explanation).
My main hope here is to point to a general intuition that currently plays an important role in my relationship to cases of this kind, and which I expect various others to share. The intuition is: if there is something intrinsically important at stake in the distinction between doing and allowing harm, it ought to reflect something that the potential victim of the harm in question has reason to care about. As far as I can tell, the doing and allowing distinction fails this test: Wilfred has no reason to care about whether he dies by doing or dies by allowing. And if Wilfred has no reason to care — and especially if Maria, too, has no reason to care — then it’s very hard for me to see why morality should care, intrinsically, either.
I want to emphasize that this is not a fully general objection to all forms of non-consequentialism. The worry here is not: “but if the doing-allowing distinction doesn’t track differences in the impersonal goodness of outcomes, then how could it be justified?” If, for example, it turned out that Wilfred did have some stronger interest in avoiding death by doing vs. death by allowing, then nothing I’ve said so far would conflict with allowing this interest to serve as the basis of an agent-relative deontological constraint against killing Wilfried, vs. letting him die, in order to create a better world. That is, the demand isn’t that intrinsically important moral distinctions always track axiological differences, or differences in the impersonal weighings of the interests of the different agents involved; rather, it’s that such distinctions track something that someone has some reason to care about. And I don’t feel like I’ve heard a non-question-begging case for this in the context of doing and allowing.
I do think, though, that something like this objection plausibly applies to a variety of other deontological distinctions as well. For example, it seems plausible to me that we do not have reason (of the type relevant to intrinsic justifications) to care whether we die by being pushed in front of trolleys, or by having trolleys turned on us; whether we die via intended harm, vs. foreseen harm; and so on. And I think something like this objection plays a role in the intuitive problem with the so-called “paradox of deontology,” on which one cannot violate a deontological constraint C with respect to one victim, even to prevent e.g. five other people violating that same constraint with respect to five other victims (see Scheffler (1994) for discussion). Deontological constraints, one feels, ought to be about protecting some legitimate interests/concerns of the potential victims — but then, one wonders, why don’t the interests/concerns of the five victims outweigh the identical interests of the one?
That said, as I believe Ketan Ramakrishnan points out in an unpublished paper, objections to the effect that a given feature of deontology is too alien from what we have reason to care about need not apply to other characteristic features of non-consequentialist thought, such as options that make it permissible to give moral priority to e.g. one’s own family, friends, etc, or duties that require one to do so. Whatever their other merits, these options/duties seem more plausibly grounded in something the agents involved have reason to care about — namely, the special relationships these options/duties reflect.
I’ll close by emphasizing that intrinsic justifications of deontological distinctions like doing/allowing are not the only possible justifications; that “pragmatic” justifications of the type I gestured at earlier need not be cashed out in naively instrumentalist ways, nor need they be “weaker” than intrinsic justifications; and the line between pragmatic and intrinsic, consequentialist and non-consequentialist, can, in my experience, get blurry fast. Even in the context of intrinsic justifications, I am most confident that the objection I’ve raised here applies to a kind of brute Rossian pluralism, on which there just are stronger rights/claims against doings than against allowings, and I’ve thought much less about how it generalizes to other views. What’s more, I do expect that something like Rossian pluralism will ultimately be the theory best suited to capturing many core and reflectively persistent intuitions that animate much of our moral life and practice, and this seems like a substantive argument (indeed, the argument) in its favor, whatever its other problems. I take these intuitions seriously in my own life, even where I do not understand them, and I think there’s a lot of room for not-having-yet-understood.
What’s more, regardless of how we feel about Rossian pluralism, or intrinsic justifications more broadly, or what future improvements in our understanding of morality’s foundations might be available, the case for upholding various deontology-flavored norms seems to me quite overdetermined, at least in most cases relevant to everyday life. Indeed, in my experience, commitment to such norms in practice seems at most weakly correlated with the types of abstract justifications a person will offer for abiding by them — a fact that makes some sense when we bring to mind that most deontologists are threshold deontologists rather than absolutists, most consequentialists accept various types of pragmatic stories anyway, and moral uncertainty can play an additional role in smoothing out the differences. Partly for related reasons, my aim here is not to query the amount of weight deontological distinctions like doing and allowing should be given. Rather, it’s to point to the type of vision of morality that it seems to me should underly the weight they are given: namely, a vision in which justifications for moral distinctions ground out — perhaps in complicated ways — in things that moral agents/patients have reason to care about.
This is an interesting essay, but I feel the lack of focus on norms and outcome probabilities is what really drives the distinction in intuition between the two cases, rather than a difference in what matters to the victim or an omission/commission distinction.
* In case 1, Maria is imminently dying and Wilfred is imminently dying. Both need to make it to the hospital to live. In the real world, this means both have a pretty good chance of dying - medical care isn't that great, there's no guarantee either will survive even if they make it. If Maria's already the one in the car, might as well be her that makes it as getting her to stop to pull over and switch out for Wilfred probably just increases the chance neither will survive on net.
* In case 2, Maria is imminently dying but Wilfred is not. Given Maria has a pretty good chance of dying soon anyway (as medical care isn't that great), but Wilfred will be fine if he can get off the road (seems eminently plausible), it seems like a bad choice from a consequentialist perspective to kill Wilfred and try and save Maria.
To reframe the case in a way that doesn't lead to such strong intuitions:
* Case A: Maria and Wilfred are both dying of thirst on an island, but know rescue is imminent in a few days. There is enough water to sustain one, but not both of them, for that time (if they split the water they will both die). Maria and Wilfred are matched in every way that would predict life expectancy. Maria already has the water bottle. I have no intuition about who should get the water, but if Maria already has it, it doesn't seem unreasonable for her to be the one to live.
* Case B: Same as Case A, but this time Wilfred has the water bottle among his possessions but hasn't realised. Maria notices, and wants to take the water bottle and drink it to survive. Again, I don't have the intuition that it is unreasonable for her to do this and be the one to live.
In general, I think what guides intuitions in these scenarios is norms and outcome probabilities sneaking in, even when we're supposed to assume these people live in a (social) frictionless vacuum with no broader consequences for society of their actions. Reframing the situation often leads to a different intuition, which implies something fishy is going on...