Is the difference between making something happen and allowing it to happen morally relevant?
Goal, and Definitions
In this essay, I shall argue for what sounds like a very controversial conclusion: that the essential distinction between making some upshot ‘U’ occur and merely allowing it to occur is in one important sense morally irrelevant. To make very clear what I am and am not arguing for, I shall begin this essay by defining the phrases ‘essential distinction’ and ‘moral irrelevance’. I will then give my reasons for postponing a proper definition of ‘making and allowing’, and outline the path the rest of the essay will trace.
Many discussions of the making/allowing distinction fail to make their definition of ‘moral relevance’ clear. Are we discussing the distinction’s relevance to practical reasoning, to moral praise and blame, or to something else altogether? I would guess that there is no one definition of ‘moral relevance’ to which everyone would agree. In this essay, I consider only the distinction’s relevance to our moral decisions about how to behave: for instance, whether we should allow U to occur in a case where we should not make it happen.
Some people argue that the distinction is ‘morally relevant’ insofar as it helps us apportion blame. I am not tackling this question; the implications my conclusion would have for praise and blame are a matter for another essay. And then, as mentioned, there are other ways in which ‘moral relevance’ could be understood. My conclusion is thus more limited than it may appear: when I speak of the distinction being ‘morally irrelevant’, I mean that it is irrelevant to (moral) decisions about how we should behave. Please bear this in mind throughout.
In my terminology, to give the ‘essential distinction’ between making and allowing is to explain what makes us describe one way of behaving as “making U happen” and another as “allowing U to occur”. This involves giving one set of properties found only in cases of “making”, and another set found only in cases of “allowing”.
Frances Kamm uses similar-sounding terminology in a slightly different way. She gives the example of a doctor pulling the plug on one of his life support machines, and says that though it is a case of “killing” (i.e. making someone die), it has what she calls an “essential property” of allowing someone to die: the patient dies only because the doctor fails to prolong her life. (This fact lets her make an exception and endorse the “killing”.)
Obviously, this could not be an ‘essential property’ of allowing as I define the phrase. I am demanding properties found only in cases of allowing. Whether the making/allowing distinction is morally relevant in my sense depends on whether these properties help us decide how we should behave. If a property found in some cases of making and some cases of allowing proved morally relevant, that would not vindicate the moral relevance of the making/allowing distinction – only that of the property. An example might be “behaving with the intention that U occur”. The moral relevance of this is not my topic, as it is separate from the question of whether the making/allowing distinction helps us decide how to act, given that we can make U happen and allow it to happen with the same intentions. Again, if my intuitions about thought experiments ever seem to clash with yours, bear in mind that I mean to hold intentions constant. (Naturally enough, some difference between us may still remain.)
‘Making and allowing’... and a plan
We all have an intuitive understanding of the difference between ‘making’ and ‘allowing’. In Part I, I will show how those who rely on this understanding are able to give thought experiments which suggest the making/allowing distinction is morally relevant. I will then suggest that only a proper analysis of the essential distinction between ‘making and allowing’ can show that this is really so. I turn to this task in Part II, arguing that, whatever its imperfections, Jonathan Bennett’s analysis gives us some provisional reason to suspect that, if there is an essential distinction between making and allowing, it is not one that it is morally relevant.
It is not too hard to give thought experiments where the making/allowing distinction seems morally relevant. Consider a case in which we must choose between killing one person and letting another die. Our first intuition is that we should choose the second course of action over the first. If this intuition is based only on the essential difference between making and allowing, killing and letting die, it provides prima facie reason to doubt my conclusion.
But do we get this intuition by considering only the essential difference between killing and letting die, the difference between making and allowing? Perhaps not. We may also be thinking about various ‘inessential’ properties of killing which merely tend to go along with it. In general, we associate killing with calculated, cold-blooded murder and letting die with hurrying past the local Oxfam shop, without really intending that anyone die. We tend to assume that the killing in the above case involves more effort, which is not necessarily the case. More seriously, we might assume it will have catastrophic social consequences, undermining people’s sense of security, etc. (Cases which involve doctors certainly suggest this would happen, and muddy the waters with issues of medical ethics. They are best avoided for this reason.)
These inessential properties might have ‘moral relevance’, in the above sense – though many of them seem relevant in deciding only who to blame
One suggestion a defender of thought experiments might make is that we can give a pair of cases in which it seems like only the essential properties of killing/letting die can have been varied. Rachels purports to do just this, in his story of Smith and Jones. Both plot to kill their bathing cousins, and creep up on them to do so. But, while Smith does so, Jones finds his cousin already drowning, and though he stands by, ready to kill the child himself, he does not have to do so.
Here our intuition is that both Smith and Jones behave equally badly. But even defenders of the making/allowing distinction like Warren Quinn accept this. It does not, as Rachels suggests, prove that the making/allowing distinction is morally irrelevant in every situation – only that in one case, making something happen is no worse than allowing it to happen. Some suggest makings are only worse than allowings when they violate someone’s negative rights: rights which can be violated only by making something happen, not merely by allowing it to occur. This suggestion is of little help here, since if any such rights exist, one of them will presumably be the ‘right to life’, which gets violated by Smith but not Jones. But the general point stands.
Thankfully, I do not need to embark on a lengthy examination of rights – that would take me far from the question at hand. If the argument in Part II succeeds, no principle which forbids making U happen but not allowing it to happen can be morally relevant. If it fails, it fails before we even start considering such principles. But at least in training our intuitions on the making/allowing distinction itself, rather than a particular thought experiment, it is not vulnerable to the above criticism.
Before shifting to this approach, I will level a final criticism at the ‘method of thought experiment’ as it applies in ethics (later in the essay, I will use a thought experiment or two to test some linguistic analyses; this is not hypocrisy, as my current point applies only when our moral intuitions are being tested, not our semantic ones). This is that the emotive nature of the thought experiments normally used may distort our intuitions. For instance, our horror at Rachels’ story may make us loathe to call either way of behaving “better”, swamping our intuitions with so much wickedness that they cannot detect a morally relevant difference when they see one.
There is sound empirical evidence that our intuitive reactions to thought experiments are affected by the way in which they are framed, even when this is morally irrelevant. We are much more reluctant to endorse killing when it is done face-to-face, but though such killing may reflect a chilly character, making blame more appropriate, it is hard to see why we ought to prefer courses of action in which our victims are more distant. Some react violently to letting a beggar on their doorstep die, but consider allowing faraway foreigners to perish to be a more permissible course of action. I find differences of nationality and proximity morally irrelevant, but I am aware that some political philosophers disagree – my point is only that people’s intuitions get affected by a difference other than the making/allowing distinction. Other examples could be given, but I shall stop here, trusting that I have shown the desirability of an approach which promises to free us from the messy dialectic we have seen, in which one side offers one thought experiment, the other offers another, and neither can agree which reveals the moral relevance of the essential distinction between making and allowing.
What would such an alternative approach look like? As mentioned, it would involve analysing the terms ‘making’ and ‘allowing’ so that we can see the essential distinction between them. This analysis cannot be given in terms which imply a moral judgement has already been made, for instance by saying that it is an essential property of allowings that they be permissible. That would prevent the making/allowing distinction from being any help in deciding how we may behave in the first place (this, recall, is the question I am examining) – we could only apply the terms ‘making’ and ‘allowing’ once we had settled that question.
So we are after an analysis of ‘making’ and ‘allowing’ in non-moral terms. If no such analysis is available (as some suspect) then that will only support my conclusion that the making/allowing distinction cannot by itself help us decide how to behave. However, I shall support this conclusion by arguing that if there is a non-moral distinction between ‘making’ and ‘allowing’, it is something like the one suggested by Bennett. I will not claim that Bennett’s analysis is perfect, only that it heads towards capturing the truth. I will then conclude by endorsing Bennett’s claim that this suggests the making/allowing distinction to be morally irrelevant, but explain why this can only be a provisional conclusion.
Why, you may ask, am I supposing that we cannot rely on our intuitive understanding of ‘making’ and ‘allowing’? My main reason is that, to paraphrase Russell, common sense embodies the semantics of the stone age. If someone removes a bathplug, we find it natural to say that he “allows” the water to run away, but on reflection this is a case of making something happen. Kamm recognises this in her discussion of pulling the plug on a life support machine, which I mentioned earlier. Another way in which our initial semantic intuitions can let us down is that when we find a case of letting someone die abhorrent, we use the charged word “killing” (or, if we feel particularly strongly, “murdering”) to describe the behaviour in question.
Warren Quinn came close to suggesting that we can rely on our intuitive understanding of ‘making’ and ‘allowing’ (in the paper to which I refer, he spoke instead of ‘doing and allowing’, but the difference is only terminological). He claimed that we say agent A ‘allowed’ upshot U to occur just when the least specific fact about A’s behaviour sufficient to explain U’s occurring is that A did not behave in a certain way. For instance, if upshot U were the death of B, and if all we need say about A’s behaviour to explain U happening is that A did not save B’s life, we will say that A allowed B to die. If, by contrast, we also have to mention that A poisoned B in the first place, we will say that A made B die.
This sounds plausible. But consider the following puzzling case, to which we shall return later. Unless A moves, U will occur. A does not move. U occurs. Does A make U happen, or merely allow it to?
The trouble is that on Quinn’s account it all depends on whether we describe “the least specific fact about A’s behaviour sufficient to explain U’s occurring” as “A not moving” or as “A standing still”. The first description says that A did not behave in a certain way, the second that he did. Both descriptions are equally unspecific, so it seems we have no reason to prefer one rather than the other. But in that case Quinn’s analysis would not let us recognise whether this is a case of making or allowing.
What Quinn needs is some account of why it normally (though not, as we shall see, always) seems appropriate to describe standing still as “not behaving in a certain way”, even though it can equally be described as “behaving in a certain way”. I suggest that Bennett provides the most plausible such account. Let us see what it would add to Quinn’s analysis.
Put briefly, Bennett’s suggestion is that we say agent A ‘allowed’ upshot U to occur just when the least specific fact about A’s behaviour sufficient to explain U’s occurring is highly unspecific. This means that most of the ways in which A could have behaved would have resulted in U occurring. Where the least specific fact about A’s behaviour sufficient to explain B’s dying is that she did not rescue him, Bennett’s analysis will (rightly) count her as allowing him to die, because saying she did not rescue B does little to pin down how she did behave. She might have been playing pinball, or reading a poem. Presumably, there are many ways not to rescue B, and only a few ways to rescue him.
Considered in the context of these cases, Bennett’s analysis seems to capture how we speak of ‘making’ and ‘allowing’. Granted, his technical account of what it is for a fact about behaviour to be ‘highly unspecific’ runs into difficulties. If there are infinitely many ways of behaving which fit the descriptions “rescuing B” and “not rescuing B”, how can we compare them? There is, I think, an answer to this, but it requires some complex maths.
But no such problem really matters. What matters is just that we base our use of ‘making’/‘allowing’ language on an intuitive grasp of what descriptions of A’s behaviour are ‘highly unspecific’. I and ten of the eleven people I have surveyed (some of them philosophers!) agree that that is the case. If we lose our grasp on this notion when confronted with technical problems, that is bad for us, but it does not make Bennett’s analysis false. He claims only that we describe behaviour as “making U happen” or “allowing U to happen” on the basis of the notion.
I hope that the cases above, as well as my (admittedly anecdotal) evidence, show that this is plausible. Now, Bennett’s opponents have found just one case in which it is not clear that his analysis captures the difference between ‘making’ and ‘allowing’. This is the case given above, in which A’s standing still is required for U to happen. For “A stands still” is a relatively specific description of A’s behaviour: to put it rather informally, there are more ways of moving than of standing still. Accordingly, Bennett’s analysis has A down as “making” U happen. But, Bennett’s opponents claim: “Surely A just allowed U to happen?”
My response (and that of many people I have asked) is that our semantic intuitions are not quite so clear. If we imagine that A deliberately forced herself to keep still, ignoring a terrible itch, in the knowledge that U would result, we might well say that she made U happen after all (I have already said that intention, effort and knowledge are essential features of neither makings nor allowings, and asked you to assume that they feature in all the cases I present). Perhaps our intuition that A allows U is based on the fact that, in almost all cases, what happens as a result of A standing still would have happened unless A had behaved in a highly specific way? This suggestion is not ad hoc – I know that my own semantic intuitions are unreliable in this way, and when I consider them in the above case, I suspect that this is what is going on.
However, so as not to be open to charges of begging the question, I shall instead defend an alternative suggestion: that our semantic intuitions are also fed by considering whether U would have happened had A not exercised her agency. Alan Donagan suggests that it is in just these cases that we say “A allowed U”. This analysis seems better placed to satisfy those whose semantic intuitions diverge from mine and Bennett’s in the immobility case: unless A’s body had a natural tendency to move by itself, A’s standing still would always be a case of allowing. Which analysis you think captures your use of the terms ‘make’ and ‘allow’ better will depend partly on what your own semantic intuitions are. But, as I said, it seems plausible to think that Bennett and Donagan both identify criteria which feed our semantic intuitions, explaining why these intuitions are confused in one of the few cases in which Bennett and Donagan deliver different verdicts.
This can only be a provisional conclusion. It seems to be the best explanation of how we use the terms ‘making’ and ‘allowing’, but someone could always come along and give a better suggestion. I note only that I have found no such suggestion in the literature so far. Thomson, for instance, takes up Bennett’s challenge in Nous (Volume 30, No. 4): she claims that we make U happen only if U comes about by a process we carry out. But Bennett would surely demand to know what counts as a process we carry out. For if “not rescuing B” does, Thomson’s analysis cannot distinguish making from allowing. Bennett’s analysis was an attempt to explain how we tell that “not rescuing B” does not count as a positive action, or for that matter as what Thomson calls a ‘process’. Thomson has offered us no way to improve on it.
So if we think that the making/allowing distinction is best captured by Bennett, Donagan or a mixture of the two, can we think it morally relevant? Where we understand ‘moral relevance’ as relevance to deciding how we should act, it does not seem that we can. Suppose I favour allowing U over making it happen, where the amount of effort involved, and my intention that U happen, remains the same. Then on Bennett’s analysis the only essential difference is that when I allow U the least specific description of my behaviour sufficient to explain it is very unspecific. But Bennett is surely right to say that this is “obviously morally irrelevant” (at least in the above sense). On Donagan’s analysis, meanwhile, the only essential difference is that when I allow U it would have happened had I not been around to exercise my agency – but this is not relevant when the time comes to decide how I should act, for then I shall be exercising my agency.
Most of Bennett’s rivals concede these points. They suggest instead that his approach is wrong, or that his analysis is imperfect (cf. Thomson). I have tried to defend Bennett against these charges as best I can. I believe he gives us reason to think that any successful analysis of the making/allowing distinction will show it to be morally irrelevant, at least in the limited sense I have been employing throughout. Limited and provisional though this conclusion must necessarily be, it has powerful implications – exploring these is a difficult task, in which self-deception seems common. Thankfully it is not my task here. I do not expect to have stopped anyone from caring whether they make things happen or merely allow them to do so – that is a basic part of human nature. I have only tried to show that, seen in the coolest analytic light, the distinction does not appear to be morally relevant.
1. When we decide how we should act we can treat our motives and knowledge of the situation as constant – our reactions to thought experiments may suggest we care less about how we should act, and more about avoiding blame by keeping our hands clean, than we would like to admit.
2. If you want them, read Peter Unger’s careful and challenging book, Living High and Letting Die.