The Parent Trap: Why Choice-Dependent Moral Theories Fail to Deliver the Asymmetry

Abstract

According to the asymmetry, creating a miserable person is morally impermissible but failing to create a happy person is morally permissible, other things being equal. Some attempt to underwrite the asymmetry by appealing to a choice-dependent moral theory according to which the deontic status of an act depends on whether the agent performs it. We show that all choice-dependent moral theories in the literature are vulnerable to what we call ‘The Parent Trap’. These theories imply that the presence of morally impermissible options can generate a moral requirement to create happy people, even at the cost of the procreator’s well-being. We consider two new choice-dependent theories that avoid this result but show that they generate an implausible moral permission to create miserable people. Choice-dependent theories therefore fail to do justice to the intuitions that motivate the asymmetry.

1 Introduction

Intuitively, creating a miserable person is wrong, but failing to create a happy person is morally permissible, other things being equal. This is known as ‘the procreation asymmetry’, or ‘the asymmetry’ for short. ¹ Many find the asymmetry intuitively plausible. ² It can explain, for instance, why ordinary individuals (usually) have no moral obligations to procreate, even if their offspring would have happy lives.

However, there is disagreement about which moral theory is best suited to underwrite the asymmetry. In this paper, we focus on a class of theories, what we call ‘choice-dependent moral theories’, that purport to underwrite the asymmetry. According to these theories, the deontic status of an act (whether this act is permissible, impermissible, or required) depends, in some way or other, on whether the act is performed by the agent. Perhaps the main example of a choice-dependent theory is

Moral Actualism (MA): An act is permissible iff its outcome is at least as good for actual people as that of any alternative. ³

MA sometimes carries the label ‘Strong Actualism’ and is contrasted with another choice-dependent theory developed by Caspar Hare (2007), which he labels ‘Weak Actualism’, and which we address in §3. For reasons we explain in §3, these labels are misleading, and we shall introduce new terminology to provide a more accurate classification of the theories.

For the remainder of this section and the next, we carefully unpack MA and the problems that it faces before moving on, in §3, to critically discuss two extant offshoots of MA. These alternatives emerged during the search for a choice-dependent theory that avoids the problems with MA. The central problem that we shall discuss applies with equal force to all these theories.

According to MA, the deontic status of different options (different possible acts that an agent can choose in a choice context) depends on how they impact the well-being of actual people. MA is just act consequentialism with a restricted domain for the morally relevant goods, the domain being the set of all actual people. But which people are actual depends on what the agent does. If the agent creates an additional person, then the set of actual people includes this person, and her well-being is morally relevant. But if the agent refrains from creating an additional person, then the set of actual people excludes this person; and how this person would have fared if the agent had created her is morally irrelevant. Hence, according to MA, the deontic status of the agent’s options depends on which option the agent actually chooses.

To illustrate how MA seems to capture the asymmetry, suppose your options are AWFUL and NONE, and you choose AWFUL. See Table 1. ⁴

Table 1. Miserable addition

Then there is an actual person, Adam, whose well-being you negatively impact. Let’s suppose that for this person, non-existence is better than a miserable existence. ⁵ According to MA, if other things are equal, you have acted wrongly. You failed to do what is best for actual people.

Next, suppose your options are instead NONE and HAPPY, and you choose NONE. See Table 2.

Table 2. Happy addition

Then according to MA, if other things are equal, you have acted permissibly, as there is no actual person for whom NONE is worse than HAPPY. So MA implies that it’s wrong to create a miserable person, but permissible to refrain from creating a happy person. Both conjuncts of the asymmetry seem accounted for.

But, as some have pointed out (Harman, 2004; Bykvist, 2007; Cohen, 2019; Hare, 2007; Spencer, 2021), and as we demonstrate below, MA does not do full justice to the intuitions that motivate the asymmetry, and faces further objections based on cases where one’s only options involve the creation of miserable people. Recently, some have defended MA against these objections, such as Daniel Cohen (2019), and others, such as Caspar Hare (2007) and Jack Spencer (2021), have developed alternative choice-dependent theories that do not face these objections.

We will show that none of these choice-dependent theories adequately explains the asymmetry. These theories face what we call ‘The Parent Trap’ – the presence of impermissible options can give you a moral duty to create a happy person, even when doing so would make you worse off than you would be if you were to create no one. This flies in the face of the asymmetry’s second conjunct. If there is no general duty to create happy people, then it is hard to see how you can have a duty to create happy people and incur harm in the process. Insofar as our aim is to build a general moral theory around the asymmetry, we have strong grounds to reject choice-dependent theories in favor of some alternative theory that can underwrite the asymmetry. ⁶ Such alternatives include harm-minimization theories, avoid reasonable objections theories, and bearer-regarding theories. ⁷ These alternative theories face various challenges. ⁸ But unlike choice-dependent theories, they at least capture the core intuition underlying the asymmetry.

The paper proceeds as follows. In §2, we review the objections to MA. In §3, we consider the choice-dependent theories that have been proposed as alternatives to MA. In §4, we show how the choice-dependent theories, including MA, fall into the Parent Trap, and hence fail to honour the intuition that motivates the asymmetry’s second conjunct. In §5, we consider whether the Parent Trap can be dismantled by appealing to our protagonist’s agent-centred permissions, or prerogatives. In §6, we present two new choice-dependent theories, inspired by a recent article by Jack Spencer (2021). While both avoid the Parent Trap, they give insufficient weight to avoiding the creation of miserable people, and hence fail to honour the intuition that motivates the asymmetry’s first conjunct. We conclude in §7.

2 Problems with moral actualism

To see why MA doesn’t do full justice to the asymmetry’s first conjunct – the injunction against creating miserable people – suppose you choose NONE over AWFUL in the case of Miserable Addition. According to MA, although you have acted permissibly, you would not have acted impermissibly if you had instead chosen AWFUL. Since you chose NONE, the miserable person you would have created if you had chosen AWFUL is non-actual, so her well-being doesn’t matter morally. But intuitively, the wrongness of AWFUL is modally robust; it doesn’t depend on which world is actualized (Harman, 2004, p. 106).

There is a further problem. Suppose your options are AWFUL and GODAWFUL. See Table 3.

Table 3. Awful or godawful

Intuitively, AWFUL is permissible and GODAWFUL impermissible, regardless of which you choose. Yet according to MA, if you choose AWFUL, then AWFUL is impermissible but GODAWFUL is permissible, and if you choose GODAWFUL, then GODAWFUL is impermissible but AWFUL is permissible. The permissible option is elusive. You know whichever option you choose will be wrong and that the option you could choose, but won’t, is your only permissible option (Bykvist, 2007; Cohen, 2019; Hare, 2007; Spencer, 2021). ⁹

3 Other choice-dependent theories

The problems with MA have motivated some to look for alternative theories that account for the asymmetry.

For example, Hare (2007) suggests that the aim is not necessarily to maximize value for actual people, but to maximize value for those who would exist if the act in question were performed. He therefore introduces a theory that he calls ‘Weak Actualism’, using the label ‘Strong Actualism’ for MA. However, as we flagged in the introduction, this labelling is misleading. This is because Weak Actualism is not actualist; on this theory, deontic statuses of acts are not fixed relative to the actual world. For this reason, we shall use the label ‘Conditional Maximization’ for Weak Actualism. Here is the idea. For any act a, let w _a be the world that would obtain if a were performed, and let w _a -value be the total value of the lives of the people who would exist if w _a were to obtain. According to

Conditional Maximization (CM): a is permissible iff were a to be chosen it would maximize w _a -value.

To apply this criterion, one looks at the world that would obtain if a certain act were to be performed, then evaluates one’s alternatives relative to that world.

To see how CM captures the first conjunct of the asymmetry better than MA, let’s revisit Miserable Addition. CM, like MA, implies that if you choose NONE, then NONE is permissible and that if you choose AWFUL, then AWFUL is impermissible. But unlike MA, CM implies that if you choose NONE, then AWFUL is impermissible. For, if you choose NONE, then it is true that if you had instead chosen AWFUL, value for those who would have existed conditional your choice of AWFUL would not have been maximized. If you had chosen AWFUL, it would have been the case that there existed a miserable person for whom the AWFUL-world was worse than the NONE-world. Hence, unlike MA, CM is consistent with the modal robustness of the moral injunction against creating miserable people.

CM is not without its problems. For instance, consider Awful or Godawful. In this case, if you choose AWFUL, then for those who exist in the AWFUL-world, the GODAWFUL-world is better, but if you choose GODAWFUL, then for those who exist in the GODAWFUL-world, the AWFUL-world is better. Hence, according to CM, AWFUL and GODAWFUL are both impermissible, regardless of which you choose. This seems like a serious problem, though some, such as Spencer, do not consider it a decisive objection. ¹⁰

Spencer (2021) proposes an offshoot of MA that he calls ‘Stable Actualism’. Like CM, and unlike MA, Stable Actualism avoids elusive permissibility. But it also has what Spencer perceives to be an advantage over CM. According to CM, you act permissibly if you choose HAPPY in Happy Addition. But you would also have acted permissibly if you had chosen NONE, for then there would have been no one for whom the NONE-world was worse than the HAPPY-world. We find this implication of CM plausible, but Spencer disagrees. He thinks that if HAPPY is chosen, then the counterfactual choice of NONE should be morally evaluated by considering the perspective of the happy person who exists in the HAPPY-world. Given that for this happy person the HAPPY-world is better than the NONE-world, we should conclude that it would have been wrong not to create her. This verdict and the avoidance of elusive permissibility jointly constitute the motivation for Stable Actualism.

Unlike MA and CM, Stable Actualism is not a complete theory; it states only a sufficient condition for permissibility. However, as we discuss below, there are different ways of completing it, each with different theoretical advantages. To define Stable Actualism, and distinguish it from both MA and CM, we need to introduce some formalism. ¹¹ Let A = {a ₁ , …, a _n } be the set of options available to the agent at the time of choice, M(a _i ) the set of options that maximize a _i -value, i.e., value for those who exist if a _i is chosen, a _@ the option that is actually chosen, M(a _@) the set of options that maximize value for actual people, and C(a _i ) the set of options that are permissible given the choice of a _i . Following Spencer, let us say that a _i stably maximizes value at a given world w just in case, at w, a _i ∈ M(a _@) ∩ M(a _i ); in other words, a _i maximizes value both for actual people and for those who exist conditional on its performance. According to

Stable Actualism (SA): for any option a _i and world w, if at w, a _i ∈ M(a _@) ∩ M(a _i ), then at w, C(a _i ) = M(a _@) ∩ M(a _i ).

If a _i stably maximizes value at w, then the only permissible options at w, given the choice of a _i , are those that stably maximize value at w.

In contrast to SA, MA and CM can be defined as follows:

MA: for any a _i , C(a _i ) = M(a _@)

CM: for any a _i , C(a _i ) = M(a _i )

SA differs from CM in precisely the way Spencer intends. SA implies that if your options are NONE and HAPPY, then regardless of what you actually choose, you will act permissibly; but it also implies that if you choose HAPPY, then you would have acted impermissibly if you had instead chosen NONE. For if you choose HAPPY, then the actual world is the HAPPY-world, and hence, the act of bringing about the NONE-world is evaluated as suboptimal for the actual people, i.e., for those who exist in the HAPPY-world.

SA also avoids elusive permissibility. To illustrate this, we can revisit the case of Awful or Godawful. AWFUL ∉ M(AWFUL), and GODAWFUL ∉ M(GODAWFUL). Hence, if you choose AWFUL, then at the AWFUL-world, AWFUL ∉ M(a _@), and if you choose GODAWFUL, then at the GODAWFUL-world, GODAWFUL ∉ M(a _@). Regardless of which option is chosen, neither stably maximizes value. Hence, SA is silent.

However, silence is not a solution. To render permissibility verdicts even when no option stably maximizes value, Spencer (2021, p. 3839) offers two possible completions of SA:

Hardline Actualism: The permissible options at w are all and only those that stably maximize value at w.

Hierarchical Actualism: If some option stably maximizes value at w, then the permissible options at w are all and only those that stably maximize value at w. If no option stably maximizes value at w, then the permissible options at w are all and only those that minimize regret.

Spencer defines the ‘regret’ of an option a _i as the difference in a _i -value between a _i and an option a _imax that maximizes a _i -value. ¹² The regret of a _i is therefore greater the worse a _i is, in terms of a _i -value, relative to a _imax . The better you could have done for those who exist given the choice of a _i , the more regret a _i carries.

Hardline and Hierarchical Actualism render verdicts in the choice between AWFUL and GODAWFUL. Since neither option stably maximizes value at either the AWFUL-world or the GODAWFUL-world, Hardline Actualism implies that regardless of which is chosen, both are impermissible. The case is a moral dilemma – a context in which one cannot avoid choosing wrongly. Hierarchical Actualism, on the other hand, implies that AWFUL is permissible and GODAWFUL impermissible. While both AWFUL and GODAWFUL have positive regret, only AWFUL minimizes regret. The negative impact on the well-being of the person who would exist if you chose GODAWFUL would be greater than the negative impact on the well-being of the person who would exist if you chose AWFUL.

4 The Parent Trap

We think that choice-dependent theories are fatally flawed as accounts of the asymmetry. To see why, consider the following case. ¹³

Wilma’s Conundrum. Wilma could remain childless or conceive a child named Pebbles. Wilma knows she has a genetic disease that would cause any offspring she produces to have a miserable life. But she also knows that having a child, even a miserable child, would fulfill her emotional needs. There is a cure available for Wilma’s disease. If Wilma receives the cure and conceives Pebbles, then Pebbles will have a happy life. However, receiving the cure would impose severe financial and physical costs on Wilma, leaving her substantially worse off than she would be if she were to remain childless. If Wilma decides to receive the cure, she must receive it prior to conceiving Pebbles.

This case is summarized in Table 4.

Table 4. Wilma’s conundrum

(For simplicity, we ignore the possibility of Wilma receiving the cure and remaining childless, which we can assume would impose the same costs on Wilma as a ₃ but would not create any new person. To rule out this possibility, one could imagine a variation of the case in which the decision whether to receive the cure and whether to create Pebbles must be made at a single point in time.)

All the choice-dependent theories we’ve considered imply that a ₂ is impermissible if actually chosen. But what matters for our purposes is the following claim:

Permissible: If a ₂ is impermissible if actually chosen, then a ₁ is permissible if actually chosen.

We think any asymmetry-friendly theory should imply Permissible. Given that a ₂ is impermissible if chosen, rejecting Permissible commits one to the claim that Wilma will act permissibly only if she chooses a ₃. But this flies in the face of the asymmetry. If it is generally permissible not to create happy people, then surely, in this case, it is permissible for Wilma not to create Pebbles with a happy life when this would involve making Wilma worse off than she would be if she were to refrain from creating anyone.

We will argue that the choice-dependent theories cannot accommodate Permissible, and that therefore these theories cannot adequately capture the spirit of the asymmetry’s second conjunct, the permission to refrain from creating happy people. We will then consider an objection to our argument that asymmetry-friendly theories should imply Permissible, which appeals to the idea that Wilma’s permission to choose a ₁ is explained by the judgment that she has an agent-relative permission to give additional weight to her own well-being. Finally, we will present two new completions of SA that accommodate Permissible, but we will show that both give insufficient weight to avoiding the creation of miserable people and therefore fail to capture the asymmetry’s first conjunct.

Coming back to Wilma’s Conundrum, suppose Wilma chooses a ₁. Then M(a _@) = M(a ₁) but a ₁ ∉ M(a ₁), so both MA and CM imply that Wilma acts impermissibly. But as we saw, these theories also imply that if Wilma chooses a ₂, she acts impermissibly. So both MA and CM are incompatible with Permissible.

What about SA? We see that a ₁ ∉ M(a ₁), a ₃ ∉ M(a ₁), and a ₂ ∈ M(a ₁) but a ₂ ∉ M(a ₂). Hence, at w ₁, {a ₁ ,a ₂ ,a ₃} ∉ M(a _@) ∩ M(a ₁). None of a ₁–a ₃ stably maximizes value at w ₁. We need to look at the proposed completions of SA to determine whether a ₁ is permissible at w ₁.

Hardline Actualism implies that all three options are impermissible at w ₁ since none is stably maximizing at w ₁. If Wilma chooses a ₁, not only does she act wrongly, but she would have acted wrongly no matter what she had chosen.

Hierarchical Actualism implies that a ₁ is impermissible at w ₁, since a ₁ fails to minimize regret. Recall that the regret of option a _i is the difference in a _i -value between a _i and whatever option maximizes a _i -value. In this case, a ₂ maximizes a ₁-value, and the difference in a ₁-value between w ₁ and w ₂ is 2 (i.e., two more units of well-being for Wilma). So, a ₁ has regret of 2. What about the regret of a ₂? In this case, a ₃ maximizes a ₂-value, and the difference in a ₂-value between a ₃ and a ₂ is 1. So, a ₂ has regret of 1 ((2+10)−(12−1)). However, a ₃ has regret of zero. This is because a ₃ maximizes a ₃-value. Hence, a ₃ is the option that minimizes regret, and so Hierarchical Actualism entails that at w ₁, a ₃ is morally required. If Wilma chooses to remain childless, then on Hierarchical Actualism, she violates her moral requirement to choose a ₃, even though fulfilling this requirement would have made Wilma worse off.

Indeed, no matter what Wilma chooses, she is not permitted to remain childless on either Hardline Actualism or Hierarchical Actualism. Suppose Wilma chooses a ₂. Then M(a _@) = M(a ₂), a ₂ ∉ M(a ₂), a ₁ ∉ M(a ₂), but a ₃ ∈ M(a ₂), and hence, a ₃ ∈ M(a _@). Moreover, a ₃ ∈ M(a ₃). So only a ₃ stably maximizes value at w ₂, and therefore a ₃ is required at w ₂. Next, suppose Wilma chooses a ₃. Then M(a _@) = M(a ₃), and a ₃ ∈ M(a ₃), and hence, a ₃ ∈ M(a _@). But a ₁ ∉ M(a ₃) and a ₂ ∉ M(a ₃), so only a ₃ stably maximizes value at w ₃, and therefore a ₃ is required at w ₃.

On all the views we’ve considered, Wilma will avoid wrongdoing only if she chooses a ₃. But if the asymmetry is true, then this is implausible. Morality demands that if Wilma decides to create Pebbles, she sacrifices some of her well-being to ensure Pebbles has a happy life. Yet, Wilma is morally prohibited from remaining childless because if she chooses the childless option, a ₁, she will be the only actual person whose well-being matters in this case, and she will be worse off than if she had instead chosen the option that is worst for Pebbles, a ₂.

Wilma’s Conundrum exposes a problem for choice-dependent theories that we call ‘The Parent Trap’. Let x and y be any well-being values, and let i and k be any positive values, and j any non-negative value. Now suppose your options are those in Table 5:

Table 5. Parent trap

For any x, y, i, j, and k, such that k > (i + j), i.e., any values such that the wellbeing gain for your child in w ₃ (relative to w ₂) is greater than your well-being loss in w ₃ (relative to w ₂), the choice-dependent theories that we’ve considered imply that you can avoid wrongdoing only by choosing a ₃, even though you will then be either equally well off (if j = 0) or worse off (if j > 0) than you would be if you were to choose a ₁.

To illustrate the problem with a realistic case, suppose that a certain married couple, living in a developing country, would be better off having and raising a child than remaining childless. Perhaps cultural norms favour having and raising children, and the couple’s preferences align with these norms. But now suppose the couple has the option of sending the child to live with distant relatives in a rich country where she would be better off than if she remained in her home country, due to the economic opportunities she would have in the rich country. If the couple sends the child away, they will be either equally well off, or worse off, than if they raise the child themselves. For example, if the couple sends the child away, not only will they lament the lack of the family life that they desire, but also they will miss the child dearly. If this well-being loss for the couple would be outweighed by the child’s well-being gain in moving to the rich country (and if the options described above are exhaustive), then according to the choice-dependent theories, the couple will avoid wrongdoing only by having the child and shipping her away. They may not remain childless.

This isn’t what proponents of the asymmetry signed up for. One of the virtues of the asymmetry is that it has intuitively plausible implications regarding procreative duties. It implies, for instance, that an individual may refrain from creating a person, even if that person would have a good life. Indeed, this connection between the asymmetry and procreative ethics is why we framed Wilma’s Conundrum so that Wilma is both the procreator and the agent in that case. (We revisit this issue in §5.1.) We doubt that proponents of the asymmetry will welcome the result that one can be morally required to have kids and make oneself worse off (or at least, no better off) just because one happens to have the option of having kids and doing worse by them.

The upshot is that the choice-dependent theories cannot adequately account for the asymmetry’s second conjunct – the permission to refrain from creating happy people.

Apart from choice-dependent theories, most asymmetry-friendly theories in the literature accommodate Permissible. For instance, according to harm-minimization theories, in Wilma’s Conundrum, a ₁ is permissible and a ₂ impermissible because only a ₁ minimizes total harm. ¹⁴

Similarly, on Michael McDermott’s (2019) Objection Minimization, a ₁ is permissible and a ₂ impermissible. On this theory, an act is permissible iff no one can reasonably object to it. And a person can reasonably object to an act iff it makes her worse off than some alternative that would impose no greater harm. While a ₁ makes Wilma worse off than a ₂, on McDermott’s theory, a ₂ imposes greater harm than a ₁, since it makes Pebbles worse off than a ₃, and the extent to which a ₂ is worse for Pebbles than a ₃ is greater than that to which a ₁ is worse for Wilma than a ₁.

Joe Horton (2021) and Abelard Podgorski (2023) have also recently defended asymmetry-friendly theories that reconcile the asymmetry with Permissible.

Importantly, we are not endorsing any of these alternative asymmetry-friendly theories. We are not even claiming that these theories are overall more plausible than any choice-dependent theory. ¹⁵ We are merely arguing that unlike choice-dependent theories, these alternative asymmetry-friendly theories do what they are intended to do; they adequately capture the asymmetry. ¹⁶

5 Agent-relative permissions

One objection to our argument against choice-dependent theories is that the central example on which our argument depends, Wilma’s Conundrum, is framed in a misleading way, and that once framed properly, we will see that choice-dependent theories can accommodate Permissible.

Permissible states that if a ₂ is impermissible if actually chosen, then a ₁ is permissible if actually chosen. Wilma is not required to create Pebbles at her own expense; she may remain childless. A critic might insist that this can be explained by assuming Wilma has an agent-relative permission to prioritize her own wellbeing. ¹⁷

Not all choice-dependent theories we’ve considered can recognize agent-relative permissions. MA, for instance, is a general theory of permissibility. It’s simply maximizing consequentialism restricted to actual people. Indeed, all choice-dependent theories we’ve criticized are presented as general theories; none includes a relevant domain restriction.

In any case, we find the objection unconvincing. If the plausibility of Permissible depends on Wilma having an agent-relative permission to prioritize her well-being, then Permissible should lose its plausibility when we imagine that the agent is someone else.

But it doesn’t. Suppose that you are the agent in Wilma’s Conundrum. Whatever you do, your well-being will be unaffected, so agent-relative permissions to prioritize one’s well-being are irrelevant. For simplicity, suppose you are the procreator and Wilma is unrelated to you. If you refrain from creating Pebbles, a ₁, Wilma will be well off; if you create Pebbles with a miserable life, a ₂, Wilma will be (for whatever reason) somewhat better off, and if you create Pebbles with a good life, a ₃, then Wilma will be (again, for whatever reason) worse off than if you had chosen a ₃. Whether the agent is Wilma or you, the choice-dependent theories give the same verdicts. They imply that you will act permissibly only if you choose a ₃.

But this goes against the spirit of the asymmetry. We suggested in §4 that if it’s permissible not to create happy people, then it’s permissible not to create a happy person when doing so would make the procreator worse off. Suppose we replace ‘the procreator’ in this statement with ‘some other person’: if it’s permissible not to create happy people, then it’s permissible not to create a happy person when creating that person makes some other person worse off. The revised statement seems about as plausible as the original. And the implications of choice-dependent theories seem no less puzzling when we imagine that the harm is imposed on someone other than the agent. This suggests that the problem we’ve raised for these theories doesn’t hinge on assumptions about agent-relative permissions. ¹⁸

6 Two new actualist variants

In this section, we present two new completions of Stable Actualism (SA) that can accommodate Permissible – the claim that if Wilma chooses to remain childless in Wilma’s Conundrum, then she acts permissibly – and that thereby avoid the Parent Trap. However, we show that each of these completions gives insufficient moral weight to avoiding the creation of miserable people. These completions therefore fail to adequately capture the ssymmetry’s first conjunct – the injunction against creating miserable people.

6.1 No dilemmas actualism

Perhaps the most obvious completion of SA to consider here is the reverse of Hardline Actualism, which we call

No Dilemmas Actualism: If some option stably maximizes value at w, then the permissible options at w are all and only those that stably maximize value at w. If no option stably maximizes value at w, then every option is permissible at w.

In the absence of stably maximizing options, Hardline Actualism condemns all, whereas No Dilemmas Actualism permits all.

Now recall Wilma’s options in Wilma’s Conundrum. (We reproduce Table 4.)

As we saw, if Wilma chooses a ₁, the childless option, then none of a ₁–a ₃ stably maximizes value at w ₁. No Dilemmas Actualism therefore implies that a ₁–a ₃ are all permissible at w ₁. This accommodates Permissible, and hence, avoids our objection.

6.2 No Regrets Actualism

The second completion of SA that we will consider avoids the Parent Trap by building on the following insight. ¹⁹ In Wilma’s Conundrum, although there is a possible world, w ₂, in which Wilma creates Pebbles and is better off for it, Wilma can foresee that morality won’t let her bring about that world. If she creates Pebbles, then morality requires her to choose a ₃, which is worse for her than either a ₁ or a ₂. But then it seems morality should allow Wilma not to create Pebbles. Intuitively, Wilma shouldn’t be required to board a train that skips past her preferred stop.

Let’s say that an option a _i is defeated if a _i doesn’t maximize a _i -value and there is some alternative a _j which maximizes a _i -value. According to

No Regrets Actualism: If some option stably maximizes value at w, then the permissible options at w are all and only those that stably maximize value at w. If no option stably maximizes value at w, then the permissible options at w are all and only those defeated by alternatives which are themselves defeated.

No Regrets Actualism accommodates Permissible. In Wilma’s Conundrum, a ₁ is defeated by a ₂, which is defeated by a ₃, so No Regrets Actualism implies that a ₁ is permissible at w ₁. One difference between No Regrets Actualism and No Dilemmas Actualism is that the former implies that a ₂ is impermissible at w ₁. Although a ₂ is defeated by a ₃, there is no option that defeats a ₃. Hence, a ₂ is not defeated by any defeated option, and is therefore impermissible at w ₁ according to No Regrets Actualism.

Unfortunately, both No Dilemmas Actualism and No Regrets Actualism are unacceptable. To see this, consider Easy Moral Choice. See Table 6.

Table 6. Easy moral choice

Wilma can either create no one or create a person with a hellish life. Indeed, we can imagine this person with an arbitrarily large amount of misery. Suppose Wilma chooses a ₄. Then M(a _@) = M(a ₄), a ₄ ∉ M(a ₄), and a ₁ ∈ M(a ₄), but a ₁ ∉ M(a ₁). Neither option stably maximizes value at w ₄. Hence, No Dilemmas Actualism implies that Wilma acts permissibly by choosing a ₄, which is absurd. Moreover, each option is defeated by an option that is defeated, since each option defeats the other; a ₁ maximizes a ₄-value and a ₄ maximizes a ₁-value. Hence, No Regrets Actualism has the same absurd implication as No Dilemmas Actualism – a ₄ is permissible at w ₄.

The claim that a ₄ is permissible violates the spirit of the first conjunct of the asymmetry. If it is wrong to create a miserable person, then surely it cannot be permissible to create an (arbitrarily) miserable person just because doing so provides the slightest benefit to the procreator.

7 Conclusion

We have argued that all choice-dependent theories in the literature are vulnerable to the Parent Trap. These views imply that one can be morally required to procreate even though this makes one worse off than if one were to remain childless. They therefore fail to adequately capture the intuition that motivates the asymmetry’s second conjunct – the permission to refrain from creating happy people. We presented two new variants, two new completions of Spencer’s SA, that don’t trigger the trap. But we showed that these views fail to capture the intuition that motivates the asymmetry’s first conjunct – the injunction against creating miserable people. There may be a choice-dependent theory that threads the needle; but showing that there is such a view is a burden that proponents of the choice-dependent tradition will have to discharge.