The Procreation Asymmetry, Improvable-Life Avoidance, and Impairable-Life Acceptance

1. Introduction

Many philosophers are attracted to the procreation asymmetry in population ethics, according to which it is always wrong to create a person who would have a bad life (all else equal) but never wrong not to create a person who would have a good life (all else equal).⁠This is the deontic version of the asymmetry, for which see Roberts (2011a), Cohen (2020: 70), and Spencer (2021: 3819–20). The asymmetry can also be formulated in terms of reasons, for which see McMahan (1981: 100), Frick (2020: 53–54), and Bader (2022: 15), and in terms of value, for which see Holtug (2004: 138) and Mogensen (2021: 570). And many philosophers are attracted to the following explanation of this asymmetry: creating a person with a bad life is wrong because that person can complain about your act, whereas declining to create a person who would have a good life is not wrong because that person never exists and so cannot complain about your act.

There is something deeply appealing about this perspective, but as it stands the view is incomplete. The procreation asymmetry does not tell us what to do in cases where we can create more than one person, or where creating a person would benefit or harm existing people. And attempts to complete the asymmetry face serious difficulties. In this paper, I present two: the problem of impairable-life acceptance and an especially acute version of the problem of improvable-life avoidance. I show how these problems afflict two recent attempts to spin out the asymmetry into a complete complaints-based theory, from Joe Horton (2021) and Abelard Podgorski (2021).

2. Avoid Reasonable Objections

Horton calls his view Avoid Reasonable Objections (ARO). ARO begins with an account of complaints: a person can complain about an act if and only if she exists after the act, she does not consent to the act, and the act is worse for her than some available alternative. So, for example, Amy can complain about my creating her with a barely good life (represented in what follows by a well-being score of \(1\)⁠) if I could have instead created her with a wonderful life (represented by a well-being score of \(100\)⁠). Horton assumes that living a bad life can be worse for a person than not existing, which means that Amy can also complain about my creating her with a bad life (represented by a negative well-being score) if I could have instead not created her. Horton notes, however, that this assumption is not essential to ARO. If we doubt that living a bad life can be worse for a person than not existing, we can instead augment our account of complaints.⁠For such doubts, see Heyd (1988), Broome (1999, 168), and Bykvist (2007). We can claim that living a bad life when one need not have existed at all is distinct grounds for complaint, in addition to the grounds given by being worse off than one could have been.

That completes the account of complaints. In order for a person’s complaint to qualify as a reasonable objection, three more conditions must be met. First, the alternative that is better for the person must give a greater sum of well-being to the set of people who currently exist. It would not be reasonable, for example, for Amy to object that her well-being is \(99\) when it could have been \(100\) if the only way to make her well-being \(100\) is to reduce every other currently-existing person’s well-being by \(10\)⁠. Second, the alternative that is better for the person must give a greater sum of well-being to the set of people who exist conditional on that alternative. It would not be reasonable, to give another example, for Amy to object that her well-being is \(99\) and not \(100\) if the only way to make her well-being \(100\) is to create Bobby with an awful life at \(- 500\) (and affect no one else). Third, it must be that either (a) no one else can reasonably object to the alternative that is better for the person, or (b) whether anyone else can reasonably object to the alternative that is better for the person does not depend on whether the person’s own objection is reasonable.⁠Horton uses clause (b) to cover cases in which there is circularity in the dependence relations between reasonable objections (2021: 497–99). The clause plays no role in my discussion below. ARO’s final component is as follows: you should act in a way to which no one (at any time) determinately can reasonably object.

Here is Horton’s statement of ARO quoted in full, as a recap:

A person can reasonably object to an act if and only if she exists, she has not consented to the act, and there is or was an alternative act satisfying 1–4.

1. The alternative is, or would have been, better for her.

2. The alternative gives, or would have given, a greater sum of well-being to the set of people who currently exist.

3. The alternative gives, or would have given, a greater sum of well-being to the set of people who exist conditional on the alternative.

4. Either (a) no one can, or would have been able to, reasonably object to the alternative, or (b) whether (a) holds does not depend on whether this person can reasonably object to this act.

You should act in a way to which no one determinately can reasonably object. (Horton 2021: 499)

As a prelude to the problem of improvable-life avoidance, I now give an objection to the most natural reading of ARO. This objection motivates a move to Horton’s clarified version of the view, presented to me in personal communication.

3. The Evil Conclusion

The objection is that ARO, on the most natural reading, does not generate the negative half of the procreation asymmetry: it does not entail that creating a person with a bad life is always wrong, all else equal. In fact, ARO implies what I will call the Evil Conclusion:

All else equal, it is not wrong to create an arbitrarily large number of people living arbitrarily bad lives.

Here is an example. Suppose that Amy currently exists with a wonderful life. You can create either an enormous number of people living awful lives or no one at all. Either way, Amy will be unaffected. So, your options are as follows:

(1) Amy \(100\)
(2) Amy \(100\)⁠, Bobby \(- 500\)⁠, Carly, \(- 500\)⁠, …, Zac \(- 500\)

ARO implies that Amy cannot reasonably object to (1) because there is no alternative which is better for her. Amy also cannot reasonably object to (2) for the same reason. And on the most natural reading of ARO, Bobby, Carly, …, and Zac also cannot reasonably object to (2). Although (1) is better for each of them, it does not give a greater sum of well-being to the set of people who exist conditional on (1): Amy is the only person who exists conditional on (1) and her well-being conditional on (1) is equal to her well-being conditional on (2). So, in this case, no one can reasonably object to (1) or (2), and ARO implies that you can permissibly choose either option. But choosing (2) is evil: it means creating an enormous number of people living awful lives for no gain whatsoever. So, this natural reading of ARO is false.

In personal communication, Horton writes that the problem stems from the interpretation of condition 3. ARO fails to generate the negative half of the asymmetry and implies the Evil Conclusion if we interpret 3 as follows:

The alternative gives, or would have given, a greater sum of well-being to the set of people who exist conditional on the alternative than the act under consideration gives to the set of people who exist conditional on the alternative.

However, Horton intended that condition 3 be interpreted as follows:

The alternative gives, or would have given, a greater sum of well-being to the set of people who exist conditional on the alternative than the act under consideration gives to the set of people who currently exist.

On this interpretation, ARO generates the negative half of the asymmetry along with its complaints-based explanation. It also avoids the Evil Conclusion: Bobby, Carly, …, and Zac can each reasonably object to (2) once they exist, because (1) would have been better for each of them, would have given a greater sum of well-being to the set of people who currently exist, would have given a greater sum of well-being to the set of people who exist conditional on (1) than (2) gives to the set of people who currently exist, and would have been such that no one could reasonably object to (1).

This clarified version of ARO (I will call it ARO+) thus improves on the natural reading. However, like the natural reading, it still faces a serious problem.

4. The Problem of Improvable-Life Avoidance

ARO+ implies that, all else equal, you should avoid creating improvable lives. Horton illustrates this implication with his Case 9 (2021: 501):⁠The original problem comes from Ross (2015).

(1) Amy \(1\)
(2) Amy \(1\) and Bobby \(1\)
(3) Bobby \(100\)

Amy cannot reasonably object to (1) because there is no alternative that is better for her. But Bobby can reasonably object to (2) once he exists, because (3) would have been better for him, would have given a greater sum of well-being to the set of people who currently exist, would have given a greater sum of well-being to the set of people who exist conditional on (3) than (2) gives to the set of people who currently exist, and would have been such that no one could reasonably object to (3). ARO+ thus implies that (1) and (3) are the only permissible options.

ARO+’s verdict in this case might seem implausible. It might seem intuitive that, if choosing (1) is permissible in Case 9, then choosing (2) is permissible as well. Call this claim the Intuition. If the Intuition is true, then ARO+ is false.

Horton suggests that the Intuition follows from another intuitively appealing claim, the Deontic Person-Affecting Principle (DPAP):

If an act \(A\) is permissible and an act \(B\) is better than \(A\) for some people and worse for no one, \(B\) must be permissible as well. (2021: 501)

Horton then argues against the DPAP using his Case 10, in which ‘—’ represents creating no one (2021: 501):

(1) —
(2) Amy \(1\)
(3) Amy \(100\)

In this case, choosing (2) is wrong. If you are going to create Amy, you should choose (3). And given that the procreation asymmetry is correct, choosing (1) is permissible. So, Horton concludes, since choosing (1) is permissible and choosing (2) is wrong, the DPAP must be false.

There are three reasons to be dissatisfied with Horton’s discussion here. The first is that the DPAP only has the implications that Horton suggests — both the Intuition and the parallel verdict in Case 10 that if choosing (1) is permissible, then choosing (2) is also permissible — if we assume Better to Exist:

Existing with a good life is better for a person than not existing.

And if we assume Better to Exist, then it is hard to hold on to the procreation asymmetry. For suppose that we accept the following dominance principle:

If an act \(A\) is at least as good as an act \(B\) for each person, \(A\) is better than \(B\) for at least one person, and performing \(A\) neither costs you too much nor violates any moral constraints, it is wrong to perform \(B\)⁠.

Then we must conclude that it is wrong not to create a person who would have a good life (all else equal) in cases where doing so would neither cost you too much nor violate any moral constraints. Given that there are such cases, the procreation asymmetry is false. So, advocates of the asymmetry should be wary of assuming Better to Exist.⁠Of course, those inclined towards both the asymmetry and Better to Exist (e.g. Roberts 2011b: 338) could reject the dominance principle. They could claim that the principle is compelling only if we interpret the second clause as follows: ‘\(A\) is better than \(B\) for at least one person who exists in \(B\)⁠.’ This version of the dominance principle is compatible with both the asymmetry and Better to Exist.

In personal communication, Horton offers a revised DPAP:

If an act \(A\) is permissible, an act \(B\) is worse than \(A\) for no one, and \(B\) does not violate any moral constraints, \(B\) must be permissible as well.

This revised DPAP serves Horton’s purposes without any commitment to Better to Exist. However, it does not allay the second reason for dissatisfaction, which is that rejecting the DPAP (revised or not) does not compel us to reject the Intuition. The revised DPAP is sufficient for the truth of the Intuition (which, recall, states that if choosing (1) is permissible in Case 9, then choosing (2) is also permissible), but it is not necessary for the truth of the Intuition. So, even if the revised DPAP is false, that does not imply that the Intuition is false, and hence does not imply that ARO+’s verdict in Case 9 is acceptable after all.

We might think that the Intuition is robust enough to stand on its own two feet, unsupported by any principle. Certainly, there are intuitions in the vicinity that are sufficiently robust. And that brings us to the third reason to be dissatisfied with Horton’s discussion: he does not consider the most acute version of the problem of improvable-life avoidance. That is because ARO+ does not only imply that choosing (1) is permissible and choosing (2) is wrong in Case 9. It also implies that choosing (1) is permissible and choosing (2) is wrong in Case 9*:

(1) Amy \(1\)
(2) Amy \(49\) and Bobby \(49\)
(3) Bobby \(100\)

This case is like Case 9 except that Amy’s and Bobby’s lives are much better in (2): their well-being is each \(49\) rather than \(1\)⁠. Nevertheless, ARO+ implies that Bobby can reasonably object to (2) once he exists, because (3) would have been better for him, would have given a greater sum of well-being to the set of people who currently exist, would have given a greater sum of well-being to the set of people who exist conditional on (3) than (2) gives to the set of people who currently exist, and would have been such that no one could reasonably object to (3). Amy cannot reasonably object to (1), because the only alternative that is better for her is (2) and Bobby can reasonably object to (2). Hence, ARO+ implies that (1) and (3) are the only permissible options. But this verdict is implausible: if choosing (1) is permissible, then choosing (2) should also be permissible.⁠Horton might reply with a modified version of Case 10: (1) —
(2) Amy \(99\)
(3) Amy \(100\) Here one might intuit that (1) is permissible and (2) is (slightly) wrong. That might be taken as support for the corresponding verdict in Case 9*. After all, Amy’s life conditional on (2) is much better than her life conditional on (1), and Bobby’s life conditional on (2) is as good as Amy’s. This intuition seems robust enough to stand on its own, but if we want a principled basis on which to challenge ARO+, we can note that it violates Weak Normative Dominance Addition, no matter how small we make \(x\) (so long as it is non-negative) and how large we make \(y\)⁠:

Suppose that every person who exists conditional on an act \(A\) has well-being at least \(0\) and at most \(x\)⁠, and that every person who exists conditional on \(A\) also exists conditional on an act \(B\) where they have well-being at least \(y\)⁠, with \(y > x\)⁠. Suppose also that every person who exists conditional on \(B\) but not \(A\) has well-being at least \(y\)⁠, and that the distribution of well-being conditional on \(B\) is perfectly equal. Then if \(A\) is permissible, \(B\) is also permissible.⁠This principle is a weakening of Arrhenius’s Normative Dominance Addition principle (Arrhenius 2022: 192).

ARO+ violates this principle in the following case (with \(y > x \geq 0\)⁠), since it implies that (1) and (3) are the only permissible options, no matter how small we make non-negative \(x\) and how large we make \(y\)⁠:

(1) Amy \(x\)
(2) Amy \(y\) and Bobby \(y\)
(3) Bobby \(2y + 1\)

5. UCV-Defeat-Uncovered

Podgorski’s view (2021) begins with an account of relative complaints: complaints against an option relative to another option. Here and below, I present minor rephrasings of Podgorski’s principles.

Common Existence Complaints*

If a person exists conditional on options \(A\) and \(B\)⁠, then she has a complaint against \(A\) relative to \(B\) iff she is worse off conditional on \(A\) than on \(B\)⁠. The strength of her complaint is the difference between her well-being conditional on \(A\) and on \(B\)⁠.

No Ghostly Complaints*

If a person does not exist conditional on option \(A\)⁠, then she has no complaint against \(A\) relative to any other option \(B\)⁠.

Existential Harm Complaints*

If a person exists conditional on \(A\) but not \(B\)⁠, then she has a complaint against \(A\) relative to \(B\) iff her well-being conditional on \(A\) is negative. The strength of her complaint is the magnitude of her negative well-being.

Existential Benefit Answers*

If a person exists conditional on \(A\) but not \(B\)⁠, then she generates an answer to complaints against \(A\) relative to \(B\) iff her well-being conditional on \(A\) is positive. The strength of this answer is the magnitude of her positive well-being. (2021: 12)

Podgorski defines ‘the unanswered strength of complaints against \(A\) relative to \(B\)⁠’ as the total strength of complaints against \(A\) relative to \(B\) minus the total strength of answers to those complaints (to a minimum of zero). He then adds a principle of defeat:

Minimize Aggregate Unanswered Complaints*

\(A\) defeats \(B\) iff the unanswered strength of complaints against \(A\) relative to \(B\) is less than the unanswered strength of complaints against \(B\) relative to \(A\)⁠. (2021: 12)

Podgorski calls the conjunction of these claims UCV-Defeat (with ‘UCV’ standing for ‘Unanswered Complaints View’). He then rounds off the theory with a deontic principle:

Uncovered

\(A\) covers \(B\) iff \(A\) defeats \(B\) and any option that \(B\) defeats. An option is permissible iff there is no option that covers it. (2021: 18)

We can call the complete theory UCV-Defeat-Uncovered.

6. The Problem of Impairable-Life Acceptance

With all that noted, consider the following case:

(1) Amy \(100\)
(2) Amy \(0\) and Bobby \(2\)

The unanswered strength of complaints against (2) relative to (1) is 98: Amy has a complaint of strength 100, but Bobby generates an answer of strength 2. Conversely, the unanswered strength of complaints against (1) relative to (2) is 0: no one has negative well-being conditional on (1) and no one is worse off conditional on (1) than on (2). So, (1) defeats (2). Since these are the only options, (1) covers (2). Therefore, only (1) is permissible. This seems like the right verdict. Amy has a very strong complaint against (2) relative to (1) and Bobby’s answer is weak.

But now suppose that (3) is also an option:

(3) Bobby \(1\)

In this new case, (1) defeats (2) as before. Meanwhile, the unanswered strength of complaints against (1) relative to (3) is 0: no one has negative well-being conditional on (1) and no one is worse off conditional on (1) than on (3). The unanswered strength of complaints against (3) relative to (1) is 0 as well, for parallel reasons. So, neither (1) nor (3) defeats the other.

The unanswered strength of complaints against (2) relative to (3) is also 0: no one has negative well-being conditional on (2) and no one is worse off conditional on (2) than on (3). However, the unanswered strength of complaints against (3) relative to (2) is 1: Bobby is slightly better off conditional on (2) than on (3) and no one else exists conditional on (3) to answer the complaint. So, (2) defeats (3).

Therefore, with (3) introduced, (1) no longer covers (2). Although (1) defeats (2), (1) does not defeat (2) and anything that (2) defeats: (2) defeats (3), and (1) does not. So, in our three-option case, (3) is the only covered option. UCV-Defeat-Uncovered thus implies that (1) and (2) are permissible.

Podgorski (2021: 16) considers a case with this structure and notes that such cases are tricky. But I claim that the case above is more than just tricky for UCV-Defeat-Uncovered. The verdict that (1) and (2) are permissible is very hard to accept, especially for those inclined towards complaints-based theories. Amy has a very strong complaint against (2) relative to (1) and Bobby’s answer is weak. Amy lives a wonderful life conditional on (1) and a life that is not even good conditional on (2).⁠To add some colour to the case, we can imagine that (2) gives Amy the life that she would have had conditional on (1) plus enough torture at the end to bring her well-being down to 0. Bobby’s life conditional on (2) is mediocre. Nevertheless, UCV-Defeat-Uncovered implies that (2) becomes permissible when we introduce (3): an option on which Bobby’s life is slightly worse.

Call this the problem of impairable-life acceptance, since it is the possibility of making Bobby’s life worse that makes (2) permissible. UCV-Defeat-Uncovered implies this problem no matter how strong Amy’s complaint against (2) relative to (1) and no matter how weak Bobby’s complaint against (3) relative to (2).

7. Conclusion

Although the procreation asymmetry is appealing, attempts to complete it face grave difficulties. For Horton’s ARO+, the problem of improvable-life avoidance remains serious. For Podgorski’s UCV-Defeat-Uncovered, the problem of impairable-life acceptance presents a new challenge.⁠I would like to thank James Evershed, Tomi Francis, Hilary Greaves, Teru Thomas, and two anonymous referees for Analysis. I am especially grateful to Joe Horton and Abelard Podgorski, for multiple rounds of helpful comments.

References

Arrhenius, G. 2022. Population paradoxes without transitivity. In The Oxford Handbook of Population Ethics, eds. G. Arrhenius, K. Bykvist, T. Campbell and E. Finneron-Burns, 181-203. Oxford: Oxford University Press.

Bader, R. 2022. The asymmetry. In Ethics and Existence: The Legacy of Derek Parfit, eds. J. McMahan, T. Campbell, J. Goodrich and K. Ramakrishnan, 15-37. Oxford: Oxford University Press.

Broome, J. 1999. Ethics out of Economics Cambridge: Cambridge University Press.

Bykvist, K. 2007. The benefits of coming into existence. Philosophical Studies 135 (3): 335-62.

Cohen, D. 2020. An actualist explanation of the procreation asymmetry. Utilitas 32 (1): 70-89.

Frick, J. 2020. Conditional reasons and the procreation asymmetry. Philosophical Perspectives 34 (1): 53-87.

Heyd, D. 1988. Procreation and value: can ethics deal with futurity problems? Philosophia 18 (2-3): 151-70.

Holtug, N. 2004. Person-affecting moralities. In The Repugnant Conclusion: Essays on Population Ethics, eds. T. Tännsjö and J. Ryberg, 129-61. Dordrecht: Kluwer Academic Publishers.

Horton, J. 2021. New and improvable lives. Journal of Philosophy 118 (9): 486-503.

McMahan, J. 1981. Problems of population theory. Ethics 92 (1): 96-127.

Mogensen, A.L. 2021. Moral demands and the far future. Philosophy and Phenomenological Research 103 (3): 567-85.

Podgorski, A. 2021. Complaints and tournament population ethics. Philosophy and Phenomenological Research

Roberts, M.A. 2011a. An asymmetry in the ethics of procreation. Philosophy Compass 6 (11): 765-76.

———. 2011b. The asymmetry: a solution. Theoria 77 (4): 333-67.

Ross, J. 2015. Rethinking the person-affecting principle. Journal of Moral Philosophy 12 (4): 428-61.

Spencer, J. 2021. The procreative asymmetry and the impossibility of elusive permission. Philosophical Studies 178 (11): 3819-42.