Consider two actions that leave the parties they affect equally well (or badly) off, but differ in whom they affect in the following way: the first action affects parties who would have existed even if it had not been performed, while the second action affects parties who would not have existed had it not been performed. Is this a morally significant difference? This is one of the central questions raised by the Non-Identity Problem. Derek Parfit answers no, defending what he calls the ‘No-Difference View’. In this article, I argue that Parfit is mistaken and that sometimes this difference is morally significant.
I begin in section I with a brief description of the Non-Identity Problem and continue in section II with a formulation of the No-Difference View. In section III, I present a case – the deafness decision – which, I argue, shows that non-identity sometimes makes a difference. This is followed in section IV by a discussion of Parfit's own medical programmes case that employs some findings from social psychology to help deflect its force as a counterexample to my view. Then, in section V, I respond to Parfit's more recent argument, based on an example he calls Case Six, in favour of the No-Difference View. I conclude, in section VI, with a brief discussion of some questions my argument raises for consequentialist moral theory.
I. NON-IDENTITY
If your parents did not conceive you when you were in fact conceived, you would not exist. You are the product of a unique sperm and a unique egg. Had your parents successfully mated a month before your actual conception, a different sperm would have met a different egg, in which case a different child – a sibling to a non-existent you – would have been created instead. Our identity depends on the timing of our conception.Footnote 1
To see how the time dependence of conception may be relevant to moral reasoning, consider two cases described by Derek Parfit. In the first, we may object to a 14-year-old girl having a baby now rather than in ten years because it will be worse for the baby. But strictly speaking this judgement involves a mistake. The girl's choice is not between having a baby at one level of welfare now or the same baby at a higher level of welfare a decade hence. It is a choice between having two different babies. On the assumption that each baby would have a life that is ‘worth living’,Footnote 2 having the first baby rather than the second does not harm the first baby any more than giving you $5 rather than giving someone else $10 harms you. In the second case, depletion, Parfit points out that different environmental policies will, over many generations, create different sets of future people. Thus, again on the assumption that, whatever we choose, the resultant people would have lives worth living in some minimal sense, it would be a mistake to object to a policy of depleting resources on the grounds that it will harm future people. It won't. Rather, it will create one set of future people with a lower quality of life than would have been had by a different set of future people had a policy of conservation been implemented instead.
We thus confront the non-identity problem. We are tempted to treat the choices in these cases as choices between making the same people better or worse off. The truth is that the comparisons we use to justify our choices are between non-identical parties, preventing us from saying that the parties are harmed. In non-identity cases like these, we cannot make moral judgements in the way that at first glance seems to make the most sense: that an act would be wrong because it harms someone. In these cases no one is harmed at all.Footnote 3 What if anything is wrong, then, with the 14-year-old girl choosing to conceive a child now, or with us choosing depletion? More generally, the non-identity problem is the problem of figuring out what, if anything, is wrong with an act that leaves some people (who have lives worth living) badly off in some respect, but is also responsible for the creation of those people and their generally worthwhile lives.Footnote 4
II. THE NO-DIFFERENCE VIEW
When we initially think an act is wrong because it harms someone, and then we realize that because of non-identity it does not cause that harm, we might come to think either that (a) the act is not really wrong, or (b) we were mistaken about what made the act wrong. Thomas Schwartz picks the first option, arguing no harm, no foul.Footnote 5 Parfit picks the second. Non-identity, he claims, should not make a difference, calling his position the No-Difference View.Footnote 6
To what is non-identity not supposed to make a difference? First, non-identity should not make a difference to what we should do. We should still choose conservation over depletion, and the 14-year-old girl should still wait. But Parfit also asks whether, when we learn about the non-identity problem, it makes us ‘less concerned’ about the effects of the actions in question.Footnote 7 I take it that the reason to be less concerned would be that we come to think of the effects as less bad. He takes his answer – that he did not become less concerned – as supportive of the No-Difference View. So it seems that, second, according to the No-Difference View, non-identity should not make a difference about how concerned we are about the effects of our actions. This means that non-identity should not make a difference to our view of how bad or good those effects are.Footnote 8
III. WHEN NON-IDENTITY MAKES A DIFFERENCE
Parfit seems right about non-identity not making a difference in the depletion case. Yet there are other cases in which non-identity does make a difference. Consider the deafness decision.Footnote 9 Assume that it is better to be hearing than deaf.Footnote 10 Now consider three scenarios:
1. Conceiving Deafness: a lesbian couple seeks a deaf sperm donor to conceive a deaf baby. Successful, they name the incurably deaf baby Conan.
2. Inducing Deafness: instead of finding a deaf donor, the couple gets sperm from a sperm bank which does not accept sperm from deaf donors. The resultant baby can hear. They then give the baby a medicine that makes him permanently deaf. They name the baby Indy.
3. Avoiding Deafness: the couple acquires the same sperm as in 2. The resultant baby can hear. They decide that Indy is not a good name and instead call the baby, who grows up with normal hearing, Norman.
Of these options, 3 is the most preferable. Given our assumptions, Norman is better off than Indy and Conan, who we will stipulate are equally permanently deaf. But now let us make some comparisons to find out how much more preferable 3 is than the other options. Think about choosing 1 over 3. This is a non-identity case. The sperm donors differ in 1 and 3, and thus, so does the sperm, and thus, so do the resultant babies: Conan and Norman are different people. Now think about choosing 2 over 3. This is an identity-preserving case, one of harming or not harming the same person. Indy and Norman have different names, but they are the same person. The only difference is that Indy suffers from induced deafness and Norman is hearing.
According to the No-Difference View, the fact of non-identity itself should not make a difference to our assessment of the value of the acts between which we are choosing. If the No-Difference View is correct, then the choice to make Conan rather than Norman is morally equivalent to the choice to make Indy rather than Norman. Is it? Which choice is worse: choosing 1 over 3, or choosing 2 over 3?
I think that choosing 2 over 3 is worse – much worse – than choosing 1 over 3. It is much worse to deafen a hearing child than it is to create a child who could not have existed without being deaf. My conversations with others about these cases indicate widespread agreement with this judgement.
If the choice of 2 over 3 is indeed worse than the choice of 1 over 3 then there is a difference between these cases. What explains this difference? I submit that the best explanation for the difference is that when the parents choose 2 over 3 they are making someone worse off than he otherwise would have been, and they are not doing that when they choose 1 over 3. In other words, the explanation for the difference in our judgements is that if the parents choose 2 over 3 they have harmed a child, and if they choose 1 over 3 they have not.
What else could the explanation be? It could not be that Indy is in a worse condition than Conan, for we have stipulated that that is not the case. Nor is the difference to be found in the intentions of the parents; in both cases they are the same – to create a deaf child – and we can presume that in both cases they believe they are being benevolent.Footnote 11 We can even assume that Indy will never find out that he could have been hearing, so Indy and Conan will feel the same about their parents and the opportunities they think they should have had. The difference comes down to non-identity. Therefore, the No-Difference View must be false. At least in some cases, non-identity matters.
IV. PARFIT'S MEDICAL PROGRAMMES
It is worth reconsidering a thought experiment structurally similar to the deafness decision, a case Parfit calls the medical programmes. He takes this case to support the No-Difference View.Footnote 12
In the medical programmes, we must decide between detecting and addressing 1,000 cases of Condition J, which causes the child a pregnant woman is already carrying to have some disability, or detecting and addressing 1,000 cases of Condition K, which causes any child who would be conceived by a woman with it to have the same disability. (We will assume that no women have both conditions.) Condition J can be eliminated with a simple treatment to the pregnant mother. Condition K cannot be cured, but it lasts only two months, so its effects can be eliminated by warning women to delay conception by two months. Either option would cost the same, and in the end the result of either option would be that there are 1,000 fewer children born without the disability.
If we choose the medical programme that addresses Condition J, 1,000 children who would have been born with the disability will instead be born without it. This is what we can call an ‘identity-preserving benefit’; treating Condition J makes these specific children better off than they otherwise would have been. Meanwhile, 1,000 children of mothers with Condition K will be born with the disability. There was no way for these specific children to be born without it. This is a ‘non-identity loss’, as these children are worse off than other children would have been had we implemented the alternative programme.
If we choose the medical programme that addresses Condition K, 1,000 children will be born without the disability instead of 1,000 different children born with it. This is a ‘non-identity benefit’; addressing Condition K creates children who would not otherwise have existed. Meanwhile, 1,000 children of mothers with Condition J will be born with a disability they would not have were the other medical programme chosen instead. This is an ‘identity-preserving loss’, that is, a harm, because it makes these children worse off than they would have been had we instead implemented the alternate programme.
Which programme should we choose? Parfit thinks that there is no moral difference between the two programmes, that they are ‘equally worthwhile’.Footnote 13 His judgement is based on the fact that no matter which programme is chosen, the same number of children will have the disability. Likewise, in our choice between Conan and Indy in the deafness decision, the same number of children will be deaf. In that case, though, the fact about similar outcomes did not have the force Parfit thinks it has here.
The cases can be made more similar. Certainly it does not matter whether it is 1,000 or just 1 instance of disability that is prevented by each of the medical programmes, so long as each prevents the same. Let us make it 1, then. And it does not matter what the disability is, so long as it is fairly serious. Deafness fits the bill.Footnote 14 Now, if we choose the Condition J programme, we will save a child from being deaf. A consequence of this programme is that a child (whose mother has Condition K) will be born deaf, rather than some other hearing child of this mother being conceived later. So, in terms of how deafness is produced, choosing the Condition J programme will cause a deaf child who otherwise would never have existed to be born. Call this latter child Conan Junior. Alternatively, if we choose the Condition K programme, a hearing child will be conceived rather than a deaf child. A consequence of this programme is that a child (whose mother has Condition J) will be born deaf instead of hearing. So, in terms of how deafness is produced, choosing the Condition K programme will cause a child who otherwise would have been hearing to be deaf instead. Call this latter child Indy Junior.
Now that we have tweaked what should be morally irrelevant details of Parfit's example, the structural similarity between it and the deafness decision is clearer. We can ask: is there a moral difference between (J) causing a deaf child, Conan Junior, to exist who otherwise would not have existed, and (K) causing a child who would have been hearing, Indy Junior, to be deaf instead? Framed this way, our judgement about the medical programmes seems just like our judgement in the deafness decision: yes, there is a difference, and K is worse.
We might ask why the adjustment to morally irrelevant details in Parfit's example leads to a result contrary to his expectation, and whether these adjustments affect the reliability of the judgements they produce. In the remainder of this section I briefly discuss three psychological phenomena that may explain the change. Each speaks in favour of thinking that our judgements about my version of the example should be considered more reliable than our judgements about Parfit's version.
Prospect Theory, from behavioural economics, teaches us that we care more about potential harms than potential benefits.Footnote 15 The question Parfit asks about his example has to do with the beneficial part of the programmes’ effects: their prevention of disability. The question I ask about the modified version of his example has to do with the cost-side of the programmes: whom they leave deaf. Though our questions are about the effects of the programmes, they are framed differently. If we do care more about potential harms, it is plausible to think we will pay more attention to assessing outcomes described as harmful. We should give greater credence to judgements about things we pay more attention to than to things to which we pay less attention.
The Identifiable Victim Effect is exemplified in the saying, ‘When one dies, it is a tragedy; when a million die, it is a statistic.’Footnote 16 We tend to pay less attention to bad effects brought to our attention when such effects happen to very large numbers of people, and pay more attention when such effects happen to small numbers of people we can identify.Footnote 17 Parfit's example asked about 2,000 people; my version of his example asked about 2. With just 2 we are likely to think more carefully about the details of the affected parties’ lives. Again, if we should give more credence to judgements about things we pay more attention to and think more carefully about, we have a reason to give greater credence to our judgements about my version of the example.
Order Bias is the phenomenon that the order in which we read and respond to questions and thought experiments affects our responses to them.Footnote 18 If we suspect a judgement is influenced by Order Bias, we may have reason to discount it. In his book, Parfit introduces the medical programmes case immediately following a discussion of the depletion case.Footnote 19 As I stated earlier, I agree with Parfit that in depletion, non-identity does not make a difference. Depletion is a very powerful intuition pump in favour of Parfit's view, and it primes the reader to come to that view in subsequent cases. So it may be that initial agreement with Parfit about his medical programmes case is owed less to the morally relevant details of the case, and more to its following the very compelling depletion. Absent a sympathetically priming context like this, it is not surprising that his example does not speak in favour of non-identity mattering.
Note that I am not saying these phenomena are rational or desirable. However, their occurrence is well-confirmed and they may play a role in the explanation for why changing trivial details about an example can affect our moral judgements about the example. There is a larger point concerning intuitions about thought experiments here, but that is a lesson for another time.
V. PARFIT'S NEW DEFENCE OF THE NO-DIFFERENCE VIEW
In On What Matters, Parfit defends the No-Difference View against a rival view which assesses the value of a state of affairs in terms of both its total amount of value and the extent to which any particular person in it is made worse off, relative to the available alternatives. This rival view, which we can call the ‘Equal Weight View’,Footnote 20 is one way of taking into account the element I have been arguing is sometimes relevant to our judgements about the value of states of affairs: how much better or worse off specific people are in them compared to how they otherwise would be.
To illustrate the problem with the Equal Weight View, Parfit offers us the following scenario, which he calls ‘Case Six’:Footnote 21
Assuming that the better state of affairs is the one with the most years lived, A, with a total of 110 years lived, is superior to B, with a total of 100 years lived, which in turn is superior to C, with a total of 70 years lived. On the No-Difference View, the particular identities of the parties and their relative advantages across different states of affairs do not matter to our assessment of those states of affairs; outcome A is the best, and that seems consistent with our intuitions.
However, on the Equal Weight View, we subtract from the value of a state of affairs how much better off the person would have been in the available alternatives. So, first, compare A and B. If we choose A over B, we are making Bernard worse off, by 50 years, than he otherwise would have been, so the revised value of A is 110 − 50 = 60. Second, compare B and C. If we choose B over C, we are making Charles worse off, by 40 years, than he otherwise would have been, so the revised value of B is 100 − 40 = 60. Meanwhile, our choice of C over its alternatives makes no one worse off than they otherwise would have been, so there is no need to subtract anything from its initial valuing of 70. According to the Equal Weight View, then, if our choices are between A, B and C, the best alternative is C. That is, on the Equal Weight View, it would be better to choose an outcome in which two people exist and live for a total of 70 years, rather than an outcome in which, instead, two other people exist and live for a total of 110 years. That seems false, and so is bad news for the Equal Weight View.
Is it bad news for my argument, too? No, for at least two reasons. First, Parfit's argument, at best, shows that there are some cases in which non-identity does not matter. I have argued that non-identity sometimes matters, which is consistent with there being cases in which non-identity does not matter. Second, Parfit's argument is an argument against one type of explanation for how non-identity should make a difference in our evaluation of states of affairs. There could be other explanations.Footnote 22
To my first reason, Parfit could reply that Case Six is relevantly similar to the cases in which I claim that non-identity makes a difference, and so were I to concede that it makes no difference in Case Six, my view would be inconsistent. As it turns out, though, I make no such concession. I believe non-identity does make a difference to our evaluations of A, B and C in Case Six. At this point, Parfit might call for an explanation of my second reason, asking how non-identity makes a difference. This would have to be a rather compelling explanation, he might press, in order to overcome the counterintuitiveness of claiming that C is better than A. Note, however, that this assumes that if one claims that non-identity makes a difference in Case Six, it makes a difference to our judgement of which outcome is the best outcome. That is not necessarily true. Non-identity may make a difference to our (cardinal) evaluations of outcomes without making enough of a difference to change their relative (ordinal) rankings. That is, one can claim both that non-identity matters and that, nonetheless, in Case Six, A is better than C. Parfit's argument gives us no reason to think such a view could not be developed. In fact, such a view has been developed, but an elaboration and defence of it would take us beyond the scope of this article.Footnote 23
VI. TWO QUESTIONS
I have argued that non-identity sometimes makes a moral difference. In the deafness decision and the modified medical programmes cases, the actual consequences of the options are the same, yet we judge the options that involve identity-preserving losses – i.e. harms – to be worse. Typical consequentialist theories (that is, act-consequentialist theories) hold that an act's morality is to be judged solely on the basis of the goodness and badness of its consequences. If the consequences of two acts are the same, so must be our judgements of them. This consequentialist idea at least seems to conflict with our intuitions in these cases. We can ask: is consequentialism compatible with our judgement that non-identity sometimes matters?
I have also claimed that Parfit is correct in saying that in cases like depletion, non-identity does not make a difference. A second question that arises, then, is: what explains when non-identity makes a difference and when it does not?
I believe the answer to the first question is yes; there is a version of consequentialism that is compatible with the idea that non-identity sometimes matters. The explanation of this version of consequentialism will require answering the second question, that is, it will give an account of what explains when non-identity makes a difference. Such an account, I believe, will involve a theory of value that holds both that identity-preserving losses (i.e. harms) can make outcomes worse, but also that losses in impersonal value can sometimes outweigh the badness of identity-preserving losses. Parfit considers and rejects one version of this type of view, but his argument depends on mistakenly assuming that whenever non-identity makes a difference, it will make a difference to our ordinal ranking of outcomes. It is possible to develop a theory which denies this. What such a theory looks like in detail, and whether it is ultimately defensible, are questions for another paper. With this article I only set out to show that non-identity matters, sometimes.Footnote 24
