The Asymmetry: A Solutiontheo_1117

Transcription

1 THEORIA, 2011, 77, doi: /j x The Asymmetry: A Solutiontheo_1117 by MELINDA A. ROBERTS The College of New Jersey Abstract: The Asymmetry consists of two claims. (A) That a possible person s life would be abjectly miserable less than worth living counts against bringing that person into existence. But (B) that a distinct possible person s life would be worth living or even well worth living does not count in favour of bringing that person into existence. In recent years, the view that the two halves of the Asymmetry are jointly untenable has become increasingly entrenched. If we say all persons matter morally whether they exist or not and on that basis try to explain the first half of the Asymmetry, we lose the second half of the Asymmetry. If we say that some persons do not matter morally and some do and on that basis try to explain the second half of the Asymmetry, we lose the first half of the Asymmetry or else find ourselves with a principle that is either inconsistent or otherwise deeply troubled in some way that has nothing to do with the content of the Asymmetry itself. In this article, I propose an alternative approach to the Asymmetry which I will call Variabilism. By understanding each and every person, whether existing or not, to matter morally but variably such that the moral significance of any loss incurred by any person is considered to depend, not on who incurs that loss and whether that person matters morally, but rather on where that loss is incurred in relation to the person who incurs it we can both nicely ground the two halves of the Asymmetry and avoid the conceptual difficulties that have plagued competing approaches. Keywords: the Asymmetry, Totalism, Prior Existence View, Neutrality Intuition, Moral Actualism, loss or harm, procreative choice, value of nonexistence 1. Introduction THE FOLLOWING TWO claims constitute what Jeff McMahan (1981, p. 100; 2009, p. 49) has called the Asymmetry. 1 A. That a possible person s life would be abjectly miserable less than worth living; a genuinely wrongful life counts against bringing that person into existence; that a person would be abjectly miserable if that person were to exist can, without more, make the otherwise permissible choice to bring that person into existence wrong. and: B. That a distinct possible person s life would be well worth living does not count in favour of bringing that person into existence; that a person would be 1 See also Singer (1976, p. 92). Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden, MA 02148, USA.

2 334 MELINDA A. ROBERTS happy if that person were to exist cannot, without more, make the otherwise permissible choice not to bring that person into existence wrong. Taken one at a time, these two claims seem highly plausible. We might even ask whether they are more than plausible. We might ask whether they are important constraints on any adequate moral theory. That is a hard question. But it is also one we may not ever need to answer. Before we ever get to that question, we might instead find ourselves convinced that one of these two claims simply rules out the other. After all, if the miserable child matters morally for the purpose of evaluating the procreative choice under scrutiny, how can the happy child not matter morally for the purpose of evaluating exactly that same kind of choice? And if we think that avoiding misery is a matter of some moral import, then how can we also think that creating happiness is not? Of course, if we must choose between making things better for the miserable child and making things better for the happy child, plausibly we must favour the miserable child. But those are not the facts of the Asymmetry. The two halves of the Asymmetry are two different cases, and it is part of each that what we do for the one child does not come at a cost to anyone, including the other child. In recent years, more sophisticated attempts at asymmetrical reconciliation have been made. And still more telling obstacles have been put in their path. Such attempts generally have tried to forge a connection between a person s moral status and that person s modal status. 2 The difficulty is that the various principles that try to decide whether a person matters morally on the basis of whether that person will exist independently of the act under evaluation or whether that person exists at the uniquely actual world or whether that person exists under the act under evaluation, etc. cannot be made to work. It is thus no surprise that many contemporary moral philosophers have reached the conclusion that the two halves of the Asymmetry cannot be situated within any principle that we are not compelled to reject, either on grounds of inconsistency or as a result of other serious conceptual concerns that have nothing to do with the content of the Asymmetry itself. The view that the Asymmetry must be abandoned in favour of some form of symmetry has become deeply entrenched. 3 2 Derek Parfit s explanation of the Asymmetry, for example, considers the critical factor to be whether the person the existence is (Parfit agrees we can say) good or bad for is among the people who ever live (Parfit, 1982, p. 150). Parfit here alludes to the person-affecting intuition the idea, that is, that what is bad must be bad for someone (Parfit, 1987a, p. 363). In the 1982 paper, Parfit seems to find such a person-affecting account of the Asymmetry conceptually unproblematic if ultimately, he thinks, an account we are forced to reject on grounds of the nonidentity problem. See also Parfit (1987b). Elsewhere, however, he takes a more critical view (Parfit, 1987a, pp (Jack and Jill case)). I discuss the nonidentity problem itself briefly in section 9.2 below. 3 See, e.g., Persson (2009, pp. 30, 37 45) ( if we take morality to consist in nothing but reasons of beneficence, purged of the reference to rights, it will be symmetrical in the sense of affirming that we have

3 THE ASYMMETRY: A SOLUTION 335 Certainly, philosophers are entitled to be suspicious that the one half of the Asymmetry simply rules out the other. But suspicion on its own will never settle the question of consistency. We may sense a tension, but, without proof of inconsistency, we have to leave open the possibility that other critical factors are quietly at play factors that in point of fact have made it perfectly sensible all along for us to say exactly what we intuitively want to say that it is wrong to bring the miserable child into existence but perfectly permissible to leave the happy child out. I believe that just such a critical factor is at play in the context of the Asymmetry. The purpose of this present article is to identify it. We will briefly review arguments that challenge more sophisticated accounts of the Asymmetry in what follows. My own view, however, is that they generally succeed in showing just what they aim to show. (The more naïve accounts, accounts that instruct us to do what we can for miserable children but not for potentially happy children, or that consider misery and happiness to be anything other than two sides of the same coin, or that see a doing-allowing distinction that is both questionable on its face and that the statement of the Asymmetry nowhere itself suggests, I simply set aside for purposes here.) What is especially interesting to me, however, is the fact those more sophisticated but ultimately flawed accounts share a single common strategy for making sense of the Asymmetry. According to that strategy, the moral significance of a given loss is entirely a function of whether the person who incurs that loss matters morally or not. If the person matters morally, so do all of that person s losses matter morally. And if the person does not matter morally, neither do any of that person s losses. This all-or-nothing strategy may seem so uncontroversial that we may not even notice that we have adopted it. We then overlook our alternatives. In contrast, the principle I call attention to here, which I will call Variabilism, makes use of a very different strategy. According to Variabilism, all persons matter morally but they all matter variably. Specifically, according to Variabilism, the moral significance of any particular loss is a function not of the moral status of the person who incurs that loss but of and this is the above-referenced critical factor just where that loss is incurred in relation to the person who incurs it. Incurred at a world where that person does or will exist, a loss has full moral significance. Incurred elsewhere, a loss has no moral significance at all. According to Variabilism, then, each of us you, me and the merely possible matters morally some of our losses, that is, have full moral significance but reasons to create good lives, proportionate to their goodness, just as we have reasons against creating bad lives, proportionate to their badness ); and McMahan (2009, pp. 49, 67) (the Asymmetry is intuitively compelling [but] extraordinarily difficult to defend or justify ; the prospects for finding a compelling theoretical defense of the Asymmetry are not promising ). See also Singer (1999, pp ); Sprigge (1968, p. 338).

4 336 MELINDA A. ROBERTS variably some of our losses have none at all. My argument then will be that Variabilism nicely grounds both halves of the Asymmetry and avoids the consistency and other conceptual problems that plague its competitors. It has thus been a mistake to think that we are compelled to reject the Asymmetry. If we want it, we can have it. Do we want it? Are the two halves of the Asymmetry as Variabilism interprets them important constraints on any adequate moral theory? That is another question, and not one I will try to answer here. I will just note here that I find the Asymmetry highly intuitive. And so do many other philosophers, including many who inaccurately, I believe take themselves to be compelled to reject it. 4 Just what we do say about the Asymmetry, moreover, will likely have important implications for many other problems in both moral theory and applied ethics. It is difficult, for example, to take the position that we have an obligation to bring the happy child into existence when we can do so at no cost to others or even that we have a bona fide moral reason in some form or another to bring that child into existence and also to claim that the early abortion or indeed the non-conception is perfectly permissible. But at the same time as we shall see it will not work at all to say that the happy child we never bring into existence has no moral status whatsoever Preliminaries 2.1 Basic Assumptions I begin with some less controversial assumptions. I take a four-place betterness-for relation to be fundamental: act a performed in a given scenario, or possible future or world w, isbetter (or worse) for a person p than an act a performed at a world w is for a person q. 6 I then understand the term loss as nothing more than shorthand 4 Broome thus draws our attention to an intuition that many people find strongly attractive. I am one of those people. We think intuitively that adding a person to the world is very often ethically neutral (Broome, 2004, p. 143). The view that Variabilism takes is more limited. According to Variabilism, adding a person whose life then goes well does not make things better but may make things worse. Still, Variabilism provides strong support to the part of the intuition that seems most clearly correct the part, that is, that says that adding a person whose life goes well does not make things better. 5 I consider the question of whether the permissibility of the early abortion can be reconciled against the moral significance of merely possible people in more detail in Roberts (2010). My argument there is that Variabilism easily shows that the two positions are consistent that we can recognize that the merely possible have exactly the same moral status we ourselves have, but also take the position that the early abortion, the abortion that does not end the life of an existing person but instead causes an additional person not to be brought into existence at all, is perfectly permissible. 6 I understand the individual betterness relation to be both transitive and asymmetric. Moreover, the ranking it establishes does not vary depending on the alternatives available to the agents at a given time.

5 THE ASYMMETRY: A SOLUTION 337 for that relation. Thus, to say that a person p incurs a loss at a given world w as a result of a given act a is to say that there was still another world w accessible to agents at the critical time such that their performance of an alternate act a at w is better for p than their performance of a at w is. I take it for granted that a given person can fail to incur a loss at a given world w in comparison to second world w but still incur a loss at w in comparison to still a third world w. Moreover, to say that a person incurs a loss at w relative to w but not relative to w is just to say that that w is better, and w is not better, for that person than w is. I will use downward arrows in the figures throughout this article to show specific losses. The significance of a given arrow s being solid rather than broken we will come back to later. This account of loss includes a strong maximizing element. If agents create less well-being for a person when they had the alternative of creating more, that person incurs a loss. Of course, it is controversial whether such an account really does provide an understanding of loss as that term is often used (along with harm) either in ordinary language or in the effort to articulate moral law. 7 Many theorists will argue that a maximizing account of loss is too broad. It discerns loss when I fail, e.g., to write you a cheque for $50. But for purposes here I set that controversy to the side. For purposes here, loss is just shorthand for talk about the betterness-for relation. Whether we want to call the differences in well-being levels that we will want to focus on here losses or not, it will be critical that we do not make our focus too narrow. We want to insure that we do not miss the moral significance ofagiven difference agivenloss in virtue of the fact that we have missed the loss.wedo not want to miss the moral significance of a situation in which agents have created (In that respect, the relation conforms to the Axiom of the Independence of Irrelevant Alternatives.) I take it for granted here that acts are to be evaluated at the worlds at which they are performed. To do otherwise means that we must move into an analysis of expected well-being. That move may be important at some point but would be a distraction for purposes here. 7 The maximizing account of loss I make use of here has close affinities with the account Gustaf Arrhenius develops under the heading of comparative harm. See, e.g., Arrhenius (forthcoming, ch. 9; 2009, pp ). I have argued elsewhere that a maximizing account of harm (or loss) has distinct advantages over its competitors. Among other things, a maximizing account of harm is an important first step in resolving a certain type of nonidentity problem (the can t-expect-better type). Such an account positions us to identify the harm that we intuitively think we should be able to discern in that problem type, which includes Parfit s depletion and risky policy examples, Kavka s slave child and pleasure pill cases and cases involving historical injustices and climate change. See section 9.2 below and Roberts (2009, pp ). See also Roberts (2007). Temporal and counterfactual (or but for ) accounts of harm seem less plausible than the maximizing account for reasons that have nothing to do with the nonidentity problem. The difficulties with those accounts of harm are well documented. List and threshold accounts also seem problematic. Hard to state, they are also difficult to test (Roberts, 2009, pp ).

6 338 MELINDA A. ROBERTS less well-being for a person when they could have created more in virtue of the fact that the shorthand we have used to talk about such situations is conventionally narrow. Rather, we want to make all the losses that are even arguably relevant to the moral analysis visible, so to speak, so that we can then decide, for each such loss, whether that loss is morally significant or not. 2.2 Comparability I also adopt the more controversial assumption that the betterness-for relation is defined even if the person p exists in one world but not the other. I assume, that is, that it is both cogent and will often be true to say that never existing at all is better (or worse) for a given person than is a particular existence. Thus a person can incur a loss or, equally, a gain at a world where that person never exists at all. We can call this the Comparability of Existence Against Nonexistence Assumption or Comparability for short. For purposes here, Comparability is just an assumption and will not be argued for in any detail. 8 On the other hand, Comparability is controversial enough that my 8 I have elsewhere argued in favour of Comparability (Roberts, 2003). McMahan, however, calls a principle akin to Comparability strained (McMahan, 2009, p. 55). More generally, McMahan rejects the idea that bringing a miserable person into existence or leaving a happy person out of existence could make things better or worse for that person. [T]o cause a miserable person to exist cannot be worse for that person, since worse for implies a comparison with an alternative that would be better...[but] there is never anyone for whom that alternative is better... (McMahan, 2009, p. 59, emphasis added). See also McMahan (1999, p. 168). Krister Bykvist s argument that creating a person does not harm or benefit that person in any comparative sense, and that leaving a person out of existence does not harm or benefit that person in either a comparative or an absolute sense, appears to be based on a very strict form of modal actualism. See Bykvist (2007, pp ). I say more about Bykvist s approach in n. 9 below. For additional criticism of Comparability, see Broome (1999, p. 168); Cohen (1996, p. 22); Heyd (1992, pp ); Dasgupta (1995, p. 383) ( It makes no sense to attribute a degree of well-being, low or high or nil, to the state of not being born. Non-existence is like nothing for us, not even a very long night, because there is no us to imagine upon. One can t be asked what it would be like to experience one s own nonexistence, for there is no subject of experience in non-existence. ). For counterarguments, see Holtug (2001, pp ), and Roberts (2003). See also Fred Feldman, who seems willing to grant the intuition behind Comparability without argument. See Feldman (1991, p. 210) ( I stipulate that if s fails to exist at w, thenv(s, w) = 0. ). Broome has more recently suggested that the quest for cogency may not be hopeless. Thus, thanking Wlodek Rabinowicz for the point, [i]t is analytic that [...isbetter for p than ] can hold only between pairs of histories in which p exists. But it may be analytically possible for [ p is better off in...than in ] to hold between two histories even if p does not exist in both of them (Broome, 2004, p. 63). For a more recent discussion, see Arrhenius and Rabinowicz (2010). See also Johansson (2010). Johansson challenges the particular defences of Comparability that appear in Roberts (2003) and Holtug (2001). But on other grounds Johansson nonetheless reaches the result that your existence can indeed be better or worse for you than your non-existence.... Johansson s defence, however, is restricted in a certain way that the defence I have suggested is not. On his view, we seem unable to obtain that it is better (worse) for a person p never to have existed at all than for p to exist at (say) world w unless we are in a case where p does or will exist at w and is, in that weak sense, actual at w. We thus seem unable to explain why, in a case where agents have made things worse for

7 THE ASYMMETRY: A SOLUTION 339 discussion of Variabilism and its competitors may be of little more than passing interest unless I say a bit about the role it plays in this article and why I think it would be a mistake to dismiss it along with the particular articulation of Variabilism, which takes Comparability for granted, that I provide here as obviously false. My aim here is to defend the Asymmetry. Why, then, am I interested in assuming Comparability, which if anything seems to favour symmetry? Why not instead take what seems to be a promising start to building a defence of the Asymmetry and just claim that the happy child incurs no loss when left out of existence? 9 Why not go existing people at w as a result of their failure to bring an additional and, relative to w, never existing person p into a lesser existence at w when they could have brought p into a better existence at w, the agents failure at w is in fact perfectly permissible. Addition Plus, which I discuss in section 9.1 below, is just such a case. Johansson seems to concede the point but argues that we have the option of taking as the operative normative fact not whether the world w where p exists is worse for p than w is, but rather whether w would be worse for p than w is if that world [w ] were to obtain (Johansson, 2010, pt VII). I won t, however, examine closely here whether that distinction is viable. For purposes here, the most important point is that the debate surrounding Comparability remains lively. 9 Bykvist (2007, pp ) takes this approach. According to Bykvist, failing to create a person cannot harm or benefit a person because, he writes, there is no one who would have been directly affected for worse or better if we fail to create that person (Bykvist, 2007, p. 335). This rule holds, according to Bykvist, for both the senses of harm and benefit that he recognizes that is, the comparative and the absolute sense. Bykvist s position on this point appears to be based on a very strict form of modal actualism a theory he assumes for purposes of discussion. Bykvist can thus offer a straightforward account of the second half of the Asymmetry. Leaving the happy child out of existence cannot harm that child, in either the comparative or the absolute sense, since there is no child there to be harmed. Bykvist s metaphysical doctrine means, however, that he cannot appeal to the comparative sense of harm for purposes of accounting for the first half of the Asymmetry. But in his view it is enough to note that bringing the miserable child into existence would harm that child in the absolute sense. Causing someone to exist who will lead a wretched life is instrumentally bad for that person and would lead to a life that is itself intrinsically bad a life that would be bad in itself for, and hence bad for in an absolute sense that person (Bykvist, 2007, p. 340). Bykvist s discussion can plausibly be taken to suggest still another point: that if an act is bad for a person in the absolute sense, then that badness, in at least some cases, is enough to ground a finding of wrongdoing. On that basis, we are, arguably, able to obtain the first half of the Asymmetry. So far, Bykvist s results coincide with the results we obtain from Variabilism. But Bykvist s approach also has certain counterintuitive features that Variabilism itself avoids. Just as Bykvist recognizes an absolute sense of harm, he also recognizes an absolute sense of benefit. If the former has moral import, then so, surely, does the latter. Now, we are not, in Bykvist s view, supposed to see that import at the world say, w3 where the agent does not create the happy child. That is so since, according to Bykvist, the happy child is not there in w3 to be harmed, in either the comparative or the absolute sense. It is, accordingly, permissible not to create the happy child at w3. The difficulty for Bykvist is that it seems that we do see that import at the world w4 where the agent does create the happy child. Specifically: if absolute harm is bad enough that it gives rise to the obligation not to create the miserable child in the first half of the Asymmetry, then it seems that absolute benefit is good enough that it gives rise to the obligation to create the happy child at w4 in the second half of the Asymmetry. Of course, since that is just what agents have done at w4, there is no basis for a finding that what agents have done at w4 is wrong. But let s now return to w3, where agents have done quite otherwise. If at w4 the agent is obligated to create the happy child, then it cannot be permissible at w3 for the agent not to create the happy child. That is what obligation means: if one is obligated to do one thing, one is prohibited from doing any alternative. We now face an

8 340 MELINDA A. ROBERTS further and just deny that we can cogently say anything at all about the never existing happy child including that that child has interests we must take into account for purposes of deciding what we ought to do or that the choice that leaves the happy child out of existence imposes a loss on that child or that it constitutes a wrong? It seems to me that we have some very good reasons for seeing whether we can work things out in some other way for seeing whether, that is, we can both assume Comparability and rescue the Asymmetry. For one thing, it is clear that any arguments against Comparability are of necessity going to be highly speculative and highly technical. Moreover, the claims that Comparability blesses seem perfectly understandable. We seem perfectly able to understand, for example, the claim that it really is worse from the perspective of the miserable child that agents bring that child into existence rather than not that things would have been better for that child had he or she never existed at all. And similarly we seem perfectly able to understand that it is better from the perspective of the happy child that agents bring that child into existence rather than not that existence is better for that child than never having existed at all. Under these conditions, a defence of the Asymmetry that does not require Comparability to be false should be of interest to anyone who has an interest in defending the Asymmetry to begin with. It is a plus, in other words, not a minus, that the defence of the Asymmetry that I will provide holds up even under the assumption of Comparability. Of course, if it can also be established that the Asymmetry is true even if Comparability happens to be false and I would not rule out that that point can be established, despite the fact that efforts to do so have not so far met with much success so much the better for the Asymmetry. But that is not my task here. Moreover, when we give up Comparability, we give up what is perhaps our strongest and certainly most intuitive ground for insisting that it is wrong to bring the miserable child into existence. 10 We want to say that the reason it is wrong to inconsistency: not bringing the happy child into existence becomes both permissible and prohibited. I argue elsewhere that Elizabeth Harman s Actual Future Principle faces an analogous difficulty (though the breach that she sees between moral reason and moral obligation enables her to avoid technical inconsistency). See Harman (2000, pp ); Roberts (2010, pp ). 10 Abandoning Comparability creates particular difficulties for any intuitive account including any plausible variation on the person-based account of why it is wrong to bring the miserable child into existence. If we decide that existence does not make things better for the happy child and that nonexistence does not make things worse for that child, then it is hard to see how nonexistence makes things better for the miserable child or how existence makes things worse for that child. But a part of the person-based approach is the idea that acts can be wrong only if they make things worse for some particular person that a bad act must be bad for someone (Parfit, 1987a, p. 363). If we can t say that existence makes things worse for the miserable child, then it is not clear that we can establish that the choice to bring that child into existence truly is bad for that child in any morally relevant sense. Of course, we can resort to list or threshold accounts of harm, or loss. But such accounts come with their own difficulties. See n. 7 above.

9 THE ASYMMETRY: A SOLUTION 341 bring the miserable child into existence is that it is worse from the perspective of that child that agents bring that child into existence rather than not. The wrongness of the choice seems not just loosely rooted in the idea that it is worse simpliciter worse, that is, for the universe for that child to exist. Nor can the wrongness of an act ever clearly be a matter of agents creating a situation that is deeply unfortunate for a given person in a case where agents could not have done any more for that person than they have. There are too many situations that strike us as deeply unfortunate that we will die; that the sun will stop shining without evoking in us any sense that a wrong has been done. The account of why it is wrong to bring the miserable child into existence that we construct from Comparability clarifies that that choice is not simply deeply unfortunate but vividly wrong: itmakes things one way for the miserable child when agents at no cost to anyone else could have done much more than that for that very same child than they have. Of course, nothing that I have said here establishes that Comparability will not, on much closer inspection, turn out to be defective and something we must reject. But the fact is that we can even now sketch a rudimentary foundation for Comparability. Let s thus accept that we cannot compare what things are like (in a literal sense) for a person p at a world where p never exists against what things are like for p at a world where p does or will exist. When we try to make that comparison, we face the rejoinder that things are not like anything at all for a person who never exists. But there are other relevant comparisons that we do seem able to make. Specifically, we can say that our well-being level at worlds where we never exist is plausibly zero. In such worlds we accrue no benefits and incur no burdens; such worlds assign to us no well-being and no ill-being whatsoever. Why, then, isn t zero exactly the level of well-being that we have at any world where we never exist at all? But if it is, then it seems that we should be able to make the comparisons we intuitively think we can make in the context of the Asymmetry. After all, if we scale well-being in such a way that, on an overall, net, lifetime, basis, the existing, miserable child has a negative well-being level and the existing, happy child has a positive well-being level, then, in reaching the result that it is better for the one child, and worse for the other child, that that child never exist at all, we are not really comparing an existing child against a non-existing child an existing child, that is, against nothing at all. We are rather comparing, in each case, two quantities, a negative well-being level, or a positive well-being level, against a zero well-being level. Even the most demanding modal actualist should concede that numbers, if not concrete individuals, exist in every possible world. The very comparisons we intuitively think that we can make when understood in this way, numbers against numbers, and not children against nothings seem well-grounded. Of course, if it is numbers against numbers we are comparing, then we have the question of how we are to connect, in our expressions at any world w of the claims

10 342 MELINDA A. ROBERTS we want to make, the one number zero with any particular child who never exists at all at w but who is, we want to say, either better or worse off for that fact. This is the concern that the subject term of our claim is missing a referent and that the expression itself is not, accordingly, fully cogent. 11 In any such case, our reference cannot, of course, be achieved by means of a singular term; it cannot be de re. But it does not follow that we are talking about nothing at all. After all, the very child we want to say something about at w at least does exist at still another world accessible to agents just prior to choice the world w that is said to be either worse, or better, for that child than w is. In talking about that child, we need not know, specifically, just who that child is or whose plight we are talking about. It is enough to know that an alternative choice would bring some additional person p into existence and make things either worse or better than they in fact are for that person p. We do not need names; descriptions will do. The very strict modal actualist might insist at this point that there is no such person p for us, even by way of description, to talk about. On that view, if p does not exist at a given world w, then nothing we want to say at w about p s situation even that p never exists at w and thus has zero well-being at w, or that p has less well-being, or more, at some alternate world w where p does exist can be true. 12 But this particular form of modal actualism seems implausibly strict. 13 After all, even the modal actualist agrees that we must find some means of recognizing as both meaningful and true the claim that JFK could have had another child who was a senator but could have been an astronaut As a referee for this journal put the point, well-being presupposes, so to say, being. 12 This may be Bykvist s view. See, e.g., Bykvist (2007, pp. 339, ). That we can say nothing about p is supposed to be so, (i) even if p does or will exist at w and w is accessible to agents just prior to choice and (ii) even though the term p can itself be understood to function, not as a singular term, but rather as an arbitrary constant that we have introduced in the context of an existential instantiation. 13 It is unclear, that is, that our best actualist theories will rule out such constructions. Thus, ways of talking about merely possible individuals within an actualist framework are proposed in Rosen (1990) and Sider (2002). 14 See McMichael (1983). Nor can the implication from well-being to being plausibly be accepted as a simple logical inference. We could just as well say that well-being is often but not always correlated with being. We could say, in other words, that the correlation between well-being and being ordinarily holds, but that, when a person s well-being level is zero, that correlation will sometimes fail. Thus, a person might have zero well-being at a given world in virtue of the fact that that person s life is (on an overall, net, lifetime basis) without value at that world or in virtue of the fact that that person never exists at that world at all but does exist at some other world accessible to agents prior to choice. One final point. Let s just suppose that we do, on balance, find a very strict form of modal actualism persuasive. And suppose that at the same time we are forced to concede that our best moral theory instructs that a correct evaluation of acts for their permissibility must take into account the plights of persons who never exist at all at one or the other of the two worlds we seek to compare. What are we to do? More generally, what should we do, if our best metaphysical theory says that we can t say something that our best moral theory requires? We stipulate. Rather than accept an unproved metaphysics whose moral implica-

11 THE ASYMMETRY: A SOLUTION 343 There is, in any case, a much bigger picture that moral philosophers need to keep in mind. We are going to have to find a way of talking about never existing people and their never existing plights. As we shall see in what follows and as the cases themselves clearly suggest, merely possible people and their merely possible plights undeniably exert a certain cross-world moral pull on how the actual acts that we actual people perform at the actual world are to be evaluated. 15 An actual act that would otherwise be deemed wrong can be made perfectly permissible in cases in which the only way of making things better for actual people involves bringing additional merely possible people into existence and then treating them badly. Unless we make reference to the plights of such merely possible people and, specifically, to the fact that agents could have done more for those people than they have; could have, that is, both brought them into existence and created more well-being for them we cannot explain the permissibility of creating less wellbeing for actual people. Thus, just as the modal actualist must find some means of talking about the children JFK never produced, so must we whether we are modal actualists or not find some means of recognizing as both meaningful and true claims about a person even when the worlds at which those claims are asserted are worlds at which that person never exists at all. 16 tions we know are false, we put into operation a language that we cannot establish is not merely artificial and yields moral truths (Feldman, 1991, p. 210) ( I stipulate that if s fails to exist at w, then V(s, w) = 0. ). 15 Addition Plus, discussed in section 9.2 below, clearly makes this point, as does Parfit s Jack and Jill case (what I call Double Wrongful Life in Roberts, 2010, ch. 2, and Hare calls a negative symmetrical case in Hare, 2007, p. 504). 16 It might seem that the intuition that we are generally under no obligation to bring the happy child into existence could be affirmed without our going to all the trouble of adopting the Asymmetry in its original form. It might seem, in other words, that we could adopt a revised asymmetry, according to which (A )we have a prima facie obligation to bring the happy child into existence, or that we have a moral reason to do so, or that it is morally better for us to do so, but that reasons of procreative autonomy at least often mean that we have no absolute moral obligation to do so, and (B ) we have an absolute moral obligation not to bring the miserable child into existence. A referee for this journal pointed out to me the relevance of this point for purposes here. See also Harman (2000, pp , ) (arguing that we have a moral reason not to have the early abortion but no obligation not to have the early abortion, even in the case where we have no reason at all, moral or otherwise, to have the early abortion and no reason at all, moral or otherwise, to do anything other than continue the pregnancy). Such a revised view and let s call the view that comprises (A) and (B) the Original Asymmetry or, as we have so far, just the Asymmetry, and the view that comprises (A ) and (B ) the Revised Asymmetry is very complicated. In developing it, we must figure out just when (and how) procreative autonomy annihilates what would otherwise be an absolute moral obligation; that it does so at least often is nothing we can test. For another, we must ask why we are not obligated to bring the happy child into existence whenever it remains, notwithstanding considerations of procreative autonomy, morally better to do just that. If we have a moral reason to bring the happy child into existence, and no moral reason not to bring the happy child into existence (other than our own disinclination, if that counts as a moral reason at all), and no moral reason to do anything other than bring the happy child into existence, then why aren t we obligated to do just that?

12 344 MELINDA A. ROBERTS 2.3 The Otherwise Plausible Permissibility Theory Many of the principles we will examine in what follows including Variabilism itself suggest evaluations of but do not, on their own, evaluate acts for their permissibility. Rather, they divide losses up according to whether those losses have moral significance for the purpose of evaluating a particular act. Each principle, in other words, serves as a loss rule that defines a particular collection of loss data, with competing loss rules defining competing collections of loss data. But each loss rule leaves the work of saying just what those loss data mean, for the purpose of evaluating a given act, to some other theory what we can call a permissibility theory. 17 Interestingly, once we make the distinction between the loss rule and the permissibility theory, it begins to seem that the locus of the controversy that the Asymmetry has created is rooted not in the permissibility theory per se but rather in the loss rule. In other words as we shall see once we determine just which losses are morally significant and which are not, the permissibility issues we will need to resolve almost settle themselves. Things will be much less straightforward when we turn to complex tradeoff scenarios. But the Asymmetry is not a complex tradeoff scenario and neither are any of the other cases we will examine for purposes here. For purposes of testing, then, we combine the particular loss rule we want to test with an otherwise plausible permissibility theory a permissibility theory, that is, that will strike each party to the debate as perfectly plausible but for the particular loss rule with which we are putting it together. 18 It is thus clear that the Original Asymmetry and the Revised Asymmetry are two very different views. If the Original Asymmetry raises certain issues, the Revised Asymmetry raises still others. The Original Asymmetry captures our strong intuition that it is permissible not to bring the happy child into existence free and unclear, while the Revised Asymmetry must engage in a great deal of manoeuvring to capture the vaguest semblance of that same strong intuition. And if the RevisedAsymmetry has certain advantages, the OriginalAsymmetry has still others. Thus, the advantage of the Revised Asymmetry would be the apparent ease of reconciling its two halves against each other. After all, the first half of the Revised Asymmetry can be read as conceding that the happy child or perhaps just the happiness of the happy child matters morally; it counts in favour of the choice to produce that child. That same fact means that, consistent with the first half of the RevisedAsymmetry, we can also say some soothing things about some of the cases that create problems for the principles that we will consider in what follows. But it also means that the task of reconciling the two halves of the Original Asymmetry is the more challenging task, which should add to, not take away from, the value of the current project. 17 The classic utilitarian principle (Totalism) is an exception. It contains both a loss rule and a permissibility theory. According to Totalism, the permissibility of a given act is determined by whether that act maximizes well-being on an aggregate basis. Totalism can thus be understood to deem all losses incurred by all people including the losses incurred when agents have the ability to bring a given person into existence and fail to do so to have full moral significance. 18 Pareto principles, tailored to take into account improvements for people that avoid morally significant losses, would be just one element of the otherwise plausible permissibility theory. A second element would be the idea that an act is wrong only if it imposes some morally significant loss on some person. A third element would address tradeoff scenarios. Thus depending on other facts the otherwise plausible

13 THE ASYMMETRY: A SOLUTION The Asymmetry We can now graph the Asymmetry as follows: a1 at w1: a2 at w2: a3 at w3: a4 at w4: cause Meg to exist at cause Meg not to cause Hans not to cause Hans to exist w1 exist at w2 exist at w3 at w Hans 0 Meg Hans 100 Meg Claim 1/first half of the Asymmetry: a1 is wrong Claim 2/second half of the Asymmetry: a3 is permissible Figure 1. The Asymmetry Acts a1 through a4 are performed within the context of the particular scenarios, or possible futures or worlds, w1 through w4. The performance of a1 leads to the miserable child Meg coming into existence at w1, and the performance of a4 leads to the happy child Hans coming into existence at w4. The performance of a2 leads permissibility theory would presumably require agents to avoid the greater of two morally significant losses. For a more detailed discussion of what the otherwise plausible permissibility theory will look like, see Roberts (2010, pp ). For purposes here, what is most important is that the content of the otherwise plausible permissibility theory avoid begging the question against any of the approaches we will be testing here, including, for example, Totalism. Thus, the otherwise plausible permissibility theory is perfectly consistent with the Totalist idea that the loss imposed on a person when agents fail to bring that person into a life worth living can make what agents have done wrong. And it is perfectly consistent with the idea that an act can be wrong, not just in virtue of what it does to a person p who does or will exist, but also in virtue of the fact that it brings a less well-off person p into existence in place of a more well-off but distinct person q. That is: the loss rule we see in Totalism implies that both such losses have full moral significance. Given those loss data, the otherwise plausible permissibility theory and, in particular, a component of that theory that addresses tradeoffs can be expected to imply that, since the morally significant loss that q faces is greater than the morally significant loss that p faces, agents ought to bring q into existence in place of p. In contrast, the very same otherwise plausible permissibility theory, combined with the loss data generated by each of the competing principles we will examine here, yield very different permissibility results.

14 346 MELINDA A. ROBERTS to Meg never existing at all at w2, and the performance of a3 to Hans never existing at all at w3. 19 Boldface type means that the indicated person Meg, Hans does or will exist at the indicated world. Italics Meg, Hans means that the indicated person is merely possible relative to the indicated world that is, that that person never exists at all at that world. It is part of both cases that the indicated choices make things neither better nor worse for any person other than those indicated, and that the alternatives displayed in figure 1 a1 and a2 in the one case, and a3 and a4 in the other exhaust the alternatives available to the agents at the critical time just prior to choice. The vertical positions of the names in the figure express the four-place betterness-for relation a relation that holds, according to Comparability, even when the comparison is between how things are for a person at a world where that person does or will exist and how things are for that same person at a world where that person never exists. Thus, a2 performed at w2 is better for Meg than a1 performed at w1 is; and a4 performed at w4 is better for Hans than a3 performed at w3. The numbers at the left other than perhaps zero can be understood to have their ordinal values only. But so that we do not lose the sense of the two cases we should also understand +100 to indicate a life that is unambiguously well worth living and 100 a life that is unambiguously much less than worth living. To say that agents have imposed a loss, as noted earlier, is just shorthand for the claim that agents have created less well-being for a person at one world when they had the alternative of creating more well-being for that same person at another world. As the downward arrows thus indicate, Meg incurs a loss at w1 where she exists and Hans incurs a loss at w3 where he does not exist. 4. The Prior Existence View; Totalism Peter Singer (1999, p. 103) suggests the Prior Existence View (PEV) as a way of grounding the intuition that a couple need not take into account the likely future 19 If one prefers, Meg and Hans can be thought of not as genuine proper names but rather arbitrary constants selected for purposes of instantiating the relevant existential claims. And these constants will be duly eliminated in accordance with the rules for existential derivation before we conclude the argument. That is: the permissibility principles that we use for purposes of testing the loss rules and ultimately for analysing the Asymmetry later in the article the principles collected as components of the otherwise plausible permissibility theory make reference, not to whether Meg rather than, e.g., Molly has incurred a loss, but rather to whether some person has incurred a loss. Similarly looking ahead to Addition Plus they refer not to whether Fen rather than, e.g., Franklin s loss cuts deeper at one world than Etta s does at another, but rather to whether one person s loss at one world cuts deeper than does a distinct person s loss at another world.

15 THE ASYMMETRY: A SOLUTION 347 pleasure of their children as a significant reason for having children. According to PEV, we count only beings who already exist, prior to the decision we are taking, or at least will exist independently of that decision (Singer, 1999, p. 103, emphasis added). As Singer notes, the account PEV provides of the second half of the Asymmetry seems plausible. PEV instructs that, in evaluating a3 or a4, we are not to count Hans; we are not to take Hans s prospective happiness into account. In our terms: we are not to count the loss that Hans incurs in w3 against a3 or in favour of a4. Suppose, then, that a4 is permissible. Then, since a3 by hypothesis has exactly the same effects on exactly the same persons as a4 does other than in the case of Hans, and since the loss Hans incurs is to be disregarded altogether, any otherwise plausible permissibility theory can be expected to instruct that a3 is permissible as well. In other words, if a3 is otherwise permissible, Hans s loss at w3 cannot make a3 wrong. As Singer argues, however, this way of looking at things collapses when we turn back to the first half of the Asymmetry. We are persuaded that it would be wrong to choose a1 in place of a2 and that a2 is obligatory. Yet it cannot be the consequences a1 has for the people who do or will exist in both w1 and w2 those people who do or will exist independently of how the choice between a1 and a2 is made that make us think that a1 is wrong and a2 obligatory. By hypothesis, all those persons fare just as well in w1 as they do in w2. So it must be the consequence for the miserable Meg at w1that motivates the first half of the Asymmetry. The loss that a1 imposes on Meg, in other words, is what makes the otherwise permissible a1 wrong. But just as PEV does not count Hans s loss against a3 or in favour of a4, nor does it count Meg s loss against a1 or in favour of a2. After all, Meg s existence depends just as much on how the choice between a1 and a2 is made as Hans s existence does on how the choice between a3 and a4 is made. The otherwise plausible permissibility is, accordingly, left with no grounds on which to say that a1 is wrong (Singer, 1999, p. 104). At this point in his discussion, Singer s aim is to compare PEV against a more classic utilitarian principle. According to that principle Totalism the permissibility of a given act is determined by whether that act maximizes well-being on an aggregate basis. Totalism is thus indifferent whether [aggregate pleasure is increased] by increasing the pleasure of existing beings, or increasing the number of beings who exist, or we can add reducing the number of beings who exist and are abjectly miserable (Singer, 1999, p. 103). On this view, Meg s misery Meg s loss matters morally even though Meg s coming into existence depends on how the choice under scrutiny is made. By refusing to blind itself to the plight of Meg at w1, Totalism can easily accommodate the first half of the Asymmetry and generate the result that a1 is wrong.