Is a person-affecting solution to the nonidentity problem impossible? Axiology, accessibility and additional people

Abstract

This paper considers two objections based in axiological considerations against the position that whether a given outcome, or possible future or world, is morally worse than a second world may depend in part on what is going on at a third world. Such a wide-angled approach to determining worseness is critical to the solution I have previously proposed in connection with the nonidentity problem. I argue that both objections fail.

Keywords:

• Nonidentity problem
• axiology
• accessibility
• person-affecting intuition

Notes

1. I have elsewhere proposed substantially the same solution to this type of nonidentity problem. See Roberts (2007, 2009). It has not been widely endorsed, perhaps as a result of the objections I consider, and argue fail, in Section 4 below.

2. Temkin (1987, 2012).

3. ‘Utilitarianism does not take seriously the distinction between persons.’ Rawls (1972, 27).

4. For discussion of the separateness objection to totalism, see de Lazari-Radek and Singer (2014, 137–140).

5. Feldman thus introduced a concept of adjusted utility, which he proposed as helpful in solving the repugnant conclusion without rejecting aggregation. See Feldman (1995a and 1995b). Broome’s concept of the personal good functions similarly. Broome (2015).

6. This is not to say that a person’s moral status depends on his or her modal status. Rather, it’s to note that the moral status of a given loss might depend on whether the person who incurs the loss does or will exist at the world where the loss is incurred. See Section 2.5 below; see also Roberts (2011a, 2011b).

7. To insist that we shouldn’t aggregate well-being across the population is not to accept that we shouldn’t aggregate. It’s just to accept that, if we aggregate at all, we should aggregate something other than that which makes the particular existence at the particular world so precious to – or so miserable for – the one who exists.

Thus, we can take the position that what is to be aggregated across the population at a given world is not well-being but rather what Broome calls the personal good, which, according to Broome, we are free to define in a way that recognizes the values of equality, priority and fairness. Broome (2015). We might, for example, take the position that a low well-being level that one suffers as a result of an inequality in the distribution of well-being across the population yields a personal good that is lower still (lower, that is, than we might have thought it would be given the well-being level alone).

I have argued elsewhere that, if the concept of the personal good is indeed that flexible, then it is also flexible enough to accommodate our existential values as well, that is, to recognize that the positive well-being level one has at a world where one exists and is happy may itself translate, not to a positive level of the personal good, but rather to a zero level. On that view, the existence of the additional happy person doesn’t make the world morally better. See Roberts, Modal Ethics, Part II (unpublished manuscript).

That latter thesis is sometimes associated with what Broome calls the neutrality intuition. Broome (2004). According to the neutrality intuition, within a certain ‘neutral range,’ a given existence doesn’t make the world either morally better or morally worse. Broome argues, however, that the neutrality intuition leads to inconsistency. I accept Broome’s argument but point out that we might reject the neutrality intuition itself but accept an alternate principle, that is narrow neutrality. According to narrow neutrality, the existence worth having doesn’t make the world morally better but does, often, make it worse. For additional discussion, see Roberts, Modal Ethics, Part II.

8. Thus, for purposes here, I accept that the deontic and the telic projects are linked in important ways. In general, we determine the permissibility of a given act performed at a given world by reference to the other possible worlds (or futures or outcomes) that agents have the resources, ability and power to bring about in place of the one; the accessibility of any better alternate outcome means a wrong act. I also think, however, that the fact that an agent, working on his or her own, does not have the resources, the ability or the power to bring about a better outcome does not necessarily mean that what the agent as an individual has done is permissible. Participating in a firing squad consisting of nineteen other armed agents isn’t necessarily permissible, even if not participating would have failed to bring about any better outcome. For further discussion, see Section 4.3 below.

I take for granted here that the term person includes many non-human animals but may well exclude some humans.

9. This is not to say that we won’t in the end aggregate. See note 7 above. It is just to say that we’ll first determine what well-being gains and well-being losses are relevant to whether one world is morally worse than another. And only then might we aggregate – not well-being, but rather what Broome calls the personal good.

10. Now, the fact that our principle explicitly includes that limitation may create some confusion. If our core principle explicitly limits the range of people the condition applies to, then it might be assumed that all our further person-affecting principles must do so as well. But that’s just an assumption. And – as we shall see in Section 2.4 below – it’s an assumption we should resist.

11. For further discussion of accessibility, see Sections 3.2 and 4.3 below; see also Feldman (1986, 10–11 and 16–25). The notion of accessibility I use here owes much to but is also more liberal than the concept Feldman has developed. For my purposes, if agents working together at a world w have the power, resources and ability to bring about an outcome w′, then w′ is accessible to w even if an individual agent had no means of making things any better than they in fact turn out to be.

12. Parfit ([1984] 1987, 363).

13. See e.g. Temkin (1987, 167), cl. (ii) of what he calls the ‘person-affecting principle’ (or ‘PAP’); Temkin notes that he bases his own formulation on Parfit’s.

14. For an example of how the telic component of the person-affecting intuition is standardly formulated, see Holtug (2010, 158).

Holtug and many other philosophers go even further: w is better than w′ only if there is a person who does or will exist in w and w is better for that person than w′. Holtug (2010, 158), Fleurbaey and Voorhoeve (2015) [in Hirose and Reisner, eds.], p. 102 (I assume that when the authors write a ‘social situation cannot be better than another if it is not better for someone,’ they mean, ‘better for someone who does or will exist in that situation).

But the principle that one outcome is better than another only if there is a person p who does or will exist in the one outcome and the one outcome is better for p than the other outcome is itself subject to instant counterexample. Thus, in cases involving the miserable child half of the procreative asymmetry, where the person’s life is clearly less than worth living and we want to say that, for that person, it would have been better never to have existed at all, the outcome that excludes that person is the better outcome even though no person who does or will exist in that better outcome is such that that outcome is better for that person.

See also Arrhenius (2015) [in Hirose and Reisner]. Arrhenius thus explores the principle that an ‘outcome A is better (worse) than B’ only if ‘A is better (worse) than B for at least one individual in A or B’ (p. 111). I take it that principle implies that A is better than B only if A is better for at least one individual in A. If so, that means that this principle too is subject to the instant counterexample of wrongful life.

But even eliminating the references to betterness and restricting the principle so that it functions just as a necessary condition on when w is worse than w′ fails to save the principle. It won’t, then, be subject to instant counterexample. But it will still fail – for my purposes, at least – by virtue of the fact that it forces the result that one world isn’t worse than another on the basis of an unduly narrow inquiry – and, as such, would completely undermine any attempt to forge a plausible person-affecting solution to the nonidentity problem. See Section 4.1 below; see also Roberts (2015).

It’s worth nothing that philosophers who focus on the deontic component of the person-affecting intuition also often (if perhaps not quite standardly) formulate that intuition as well in a way that deems an act permissible on the basis of an unduly narrow inquiry. Thus: a1 performed at w1 is wrong, only if there does or will exist a person p in w1 such that had a1 not been performed, p would have been better off. See e.g. Dasgupta (forthcoming), part 1 (combination of his claims (2) and (3)). Equivalently, the deontic intuition may be spelled out in terms of harm, where harm itself is then defined in a defective way (most commonly, by reference to a simple counterfactual ‘but for’ account of harm). See Boonin (2015, 3) (discussion of premise ‘P2’) and p. 52ff. (Chap. 3). (‘The Counterfactual Account is the commonsense account of harm.’ (Boonin, p. 52)) See also Mulgan (2006, 8). The shoot-you-in-the-arm case shows, I believe, that such accounts are problematic (whether or not they are formulated in terms of harm). See Section 3.2 above.

15. Thus, a workable Pareto-like person-affecting sufficient condition on when one world is worse than another includes the following:

P** (telic):

A world w is worse than another world w′ if there is a person who does or will exist at w such that w is worse for that person than w′ is and there is no person who does or will exist in w′ such that w′ is worse for that person than any accessible world w″ is, where w″ may but need not be identical to w.

This principle can be contrasted with still another Pareto-like principle, Px. I describe Px in Section 2.4 below but immediately reject it.

16. Temkin is probably best understood as intending to include Px as clause (i) in his own formulation of the ‘person-affecting principle’ (or ‘PAP’). See Temkin (1987, 166).

17. de Lazari-Radek and Singer (2014, 367).

18. It might be argued that Px yields that result only on the assumption that it is sometimes better for a particular person never to have existed at all. I do not think, however, that that assumption is one we should contest.

19. de Lazari-Radek and Singer (2014, 367–368).

20. de Lazari-Radek and Singer (2014, 368).

21. Roberts (2011a, 2011b).

22. If might be objected that, if the nexus between existence and loss makes things worse, then the nexus between existence and gain must make things better. It seems, however, that to defend variabilism against this objection, we need simply deny that the implication must in fact hold. Thus, variabilism asserts that the nexus between existence and loss and only that nexus makes the person’s loss morally significant. It follows – as a conceptual matter, given that a loss a person incurs at a world w compared against another world w′ is identical to the gain that same person accrues at w′ compared against w – that certain gains are also morally significant – but only those that reverse morally significant losses. There’s nothing self-contradictory in that view – unless we import into that view some further view, one that insists that the nexus between existence and gain makes the person’s gain morally significant. When we do that, we end up with a view that contradicts itself. But that’s not the view that, starting out, was under attack or that we aimed to defend.

Variabilism may legitimately be considered a version of the complaint view. But it’s a non-Singer-esque version of that view. Thus, the objection de Lazari-Radek and Singer put forward against the (Singer-esque) version of the complaint view that they explore doesn’t work against variabilism. (See de Lazari-Radek and Singer (2014, 269–370).) Variabilism isn’t, that is, applied to addition plus, committed to the position that, in virtue of the applause agents get in w3, that w3 is better than w1. Variabilism, rather, is committed only to the position that the applause agents get in w3 shows that w3 is better than w2. We should agree that w3 is better than w2 in virtue of the fact that w3 is better in a morally significant way for q. But we can still deny that w3 is better than w1 by denying that w3 is better for q than w1 is in a way that has any moral significance at all.

23. See Roberts (2007 and 2009); see also Roberts and Wasserman (2017).

24. I have called it the nonidentity fallacy. See note 23 above.

25. Some philosophers favor a non-comparative account of harm and argue, accordingly, that harm, or loss, can be imposed in a case where the person’s well-being has been maximized – even, that is, in a Type II case. But while I am happy to say that the term ‘harm’ may well have such a non-comparative sense – along with many others; ‘harm’ seems to me to be a multiply ambiguous term – I find it implausible that harm in that sense has any moral significance. It seems plausible to me, in other words, that, for a harm or a loss or a simple diminution in well-being to have moral significance, the option of things being at least somewhat better for the person must exist.

26. My thinking on this point has been greatly influenced by David Wasserman. See note 23 above.

27. Kavka (1981, 98). Explaining the ‘precariousness’ of existence, Kavka, citing Adams, Parfit and Schwartz, writes: ‘Which particular future people will exist is highly dependent upon the conditions under which we and our descendants procreate, with the slightest difference in the conditions of conception being sufficient, in a particular case, to insure the creation of a different future person … This fact forms the basis of a surprising argument …’ Kavka (1981, 93).

28. Kavka (1981, 98).

29. Andy thus is a perfectly good genuine proper name; Ruth a constant.

30. Now, Kavka doesn’t himself rely on that particular counterfactual. Other theorists, however, do; and we can accept it here as a stipulation of the case: in the closest possible world where the couple doesn’t take the pleasure pill, Andy never exists.

31. See Kavka (1981, 100, n. 15).

32. This is just to say that the simple counterfactual ‘but for’ test of harm – that is, of making things worse for – is false. We must reject the principle that asserts that a1 harms a person p only if had a1 not been performed – but for a1 – p would have been better off. The concept of the baseline here is Feinberg’s. Feinberg (1988, 150–151). I question, however, the simple counterfactual account of harm he explores.

33. Of course, if we want to further complicate their efforts, we can simply note that they would have had no way of knowing just which child that would have been.

34. And it’s a fallacy to think that we can. It’s not implausible that what people have been doing with this case and other cases that ground the Type I problem is comparing the subject’s actual well-being in w1 given a1 against the subject’s expected well-being under various alternate choices. Now we might think the important thing is whether a choice, expectationally, is worse than another for a person, in which case we should compare expected value compared against expected value. Or we might think the important thing is whether a world is in fact worse than another for a person, in which case we consider actual value compared against actual value. But we can’t mix things up without facing inconsistency. See note 23 above.

35. Smilansky (2017).

36. It’s the pause, not the pill, that arguably gives Andy whatever chance he has of coming into existence. And, whatever the duration of that pause turns out in fact to be at w1, the probability, just prior to choice, that that pause will have exactly that duration remains very small.

37. We have ample room to reach those felicitous results, in other words, on the basis of, for example, the person-affecting principle P** (not Px). See note 15 above.

38. See note 14 above (discussion of Fleurbaey etc.).

39. I here assume a tripartite theory of outcome betterness: if w2 is neither better nor worse than w3, then w2 is equally as good as w3.

40. That same principle – P**, not, of course, Px – is the same principle we need to solve the nonidentity problem. See note 15 above. In contrast, Ps would close the door to that further principle by forcing us to say that w1 is at least as good as w2 is.

41. In taking this approach, we aren’t eschewing pairwise comparison. It’s just that when we conduct our pairwise comparisons we have to take facts about accessibility into account. We look at w1 – and see that for agents there w3 remains a perfectly live option.

42. The discussion of this Section 4.2 shows that the wide-angle-lens approach to determining betterness we see in P* doesn’t, after all, violate the independence principle. For this approach simply denies that there could be both a case in which w3 exists as an accessible alternative from w1 and w2 and a case in which w3 doesn’t exist as such an alternative; one of the two cases would be ruled out as impossible in view of our understanding that each world has its features necessarily. Specifically, the following accessibility axiom seems both undeniable, given our understanding of worlds, and plausible.

Accessibility axiom:

If wβ is accessible to wα, then necessarily wβ is accessible to wα.

43. See note 11 and Section 3.2 above (on accessibility).

44. Now, in the context of the nonidentity problem, the individual agent often can, on his or her own, bring about the better outcome. Thus, in the pleasure pill case, whoever it is who ingests the pleasure pill was fully capable of ingesting the aspirin instead; the acts of other agents do not block that choice. But we can certainly sketch nonidentity cases where that condition is not satisfied – problems, in other words, that combine the nonidentity problem and a collective action problem. In any case, the goal here is not simply to address the nonidentity problem but also to develop principles that stand up to scrutiny in other sorts of cases as well.

45. I am grateful to Elizabeth Harman for bringing to my attention to the need to address this issue here.

46. de Lazari-Radek and Singer (2014, 366).

47. de Lazari-Radek and Singer (2014, 370).

48. See Temkin (1987, 2012).

49. McMahan has worried about similar inconsistencies, proposing that our betterness results will themselves depend on our ‘order of comparison.’ See McMahan (2006). On the approach suggested here, we shall have no need to relativize our results in that way. If a third world w3 – as in addition plus or the mere addition paradox – shows that w2 is itself morally defective, that defect shall stay with w2, whether it is w3 we are comparing w2 against or w1 we are comparing w2 against. The order in which we consider things is not important; what is important is that we not think we can accurately compare w2 against w1 without first understanding whether w3 exists as an accessible alternative.

50. I am very grateful to Rahul Kumar for his comments on an earlier version of this paper. I presented a still earlier version of the paper at a Population Ethics Workshop organized by the Institute for Futures Studies (Stockholm, May 2016), and I am very grateful for the comments of those who participated in that conference as well.