54 Counterfactual Utilitarianism: A New Metaphysical Framework for Consequentialist Ethics

Faculty Mentor: Matt Haber (Philosophy, University of Utah)

Some traditional utilitarians (following “average utilitarianism” [3]) might carry out their moral calculations with the aim of maximizing overall average utility. As an illustration of an (admittedly fanciful) scenario in which calculating under this rule would lead us astray, consider the following. Assuming briefly for the sake of argument that well-being can be realistically and easily quantified this way, let us imagine a universe containing only one person—let’s call him Tyler—who has an overall well-being of 7/10. Suppose we are presented with only two possible ways to make a moral improvement in this universe. Action 1 would increase Tyler’s well-being to 8/10. Action 2 would bring into existence five more people, all with well-being 9/10, without changing Tyler’s well-being. If we choose to take Action 1, the resulting average utility in this universe becomes 8/10. If we choose to take Action 2, the resulting average utility in this universe becomes 8.67/10. So, if maximizing overall average utility is the goal we ought to strive for in our moral calculations, Action 2 is the morally better one. My intuition, however, strongly disagrees. I believe that while Action 1 results in a positive moral change, Action 2 results in no moral change whatsoever—it is entirely morally neutral. I believe, then, that we in fact should opt for the former action, and counterfactual utilitarianism will provide us the reason why.

Some other traditional utilitarians (following “Classical utilitarianism” [4]) might carry out their moral calculations with the aim of maximizing overall total utility. As an illustration of an (again, admittedly fanciful) scenario in which calculating under this rule would lead us astray, consider the following. Again, assuming well-being can be realistically and easily quantified this way, let us imagine a universe full of people with 10/10 well-being. Since they are all perfectly well-off, there is no way to make a moral improvement in this universe by affecting the people who already exist. So, instead, similar to in Action 2 from the previous scenario, perhaps we should just spawn more people with 10/10 well-being. If maximizing overall total utility is the goal we ought to strive for in our moral calculations, this clearly would be a moral act. However, like with Action 2 in the previous scenario, I believe this would result in no moral change whatsoever. There is no moral difference between a universe of 1,000 perfectly well-off people and a universe of 1,000,000 perfectly well-off people, and counterfactual utilitarianism will provide us the reason why.

This example may be somewhat reminiscent of Derek Parfit’s thought experiment which led him to the “Repugnant Conclusion” [5]. One way to think of the procedure which leads to the Repugnant Conclusion is like this. We start by imagining a universe with a small number of very well-off people. We then imagine increasing the size of the population significantly, while also slightly decreasing each person’s well-being. This seems to make a moral improvement, since the universe’s total utility is greatly increased. If we make this move many, many more times, though, we eventually arrive at a universe with an absolutely massive population—and therefore with more total utility than any previous iteration of this universe—“even though its members have lives that are barely worth living” [6]. If we are classical utilitarians, though, then we have to accept that this final iteration of the imagined universe is the best one. Although this is uncontroversially an undesirable moral conclusion, and many philosophers have attempted to provide reasons to reject it, “it has been surprisingly difficult to find a theory that avoids the Repugnant Conclusion without implying other equally counterintuitive conclusions” [7]. One theory which avoids it is average utilitarianism, which would of course find the average well-being to be highest in the first iteration of this imagined universe, and therefore identify that iteration as the morally preferable one. However, I have already mentioned that I think average utilitarianism makes errors in its moral evaluations elsewhere. Instead, I will soon argue that counterfactual utilitarianism provides us the reason why we do not have to accept the Repugnant Conclusion.

One way to describe the mistake that traditional utilitarians make in all the above three examples is to say that they are attempting to make things better from the point of view of the universe. But, of course, the universe itself has no point of view. It is not a sentient being; it does not have interests. Consequentialists concerned with maximizing pleasure, well-being, positive phenomenal experiences, etc. must be concerned with maximizing pleasure, well-being, positive phenomenal experiences, etc. for individual sentient beings: the subjects for whom such a maximization will make an experienced moral difference. In other words, when arguing that an action is a morally good one, the question we must be able to answer is, For whom? This is not to say that we cannot talk about making things better in general, but rather that making things better in general can only be accomplished by making things better for individual sentient beings. In many cases, it may turn out that the most effective way to achieve this goal is by instrumentally aiming to maximize overall average utility, for example. But there are also cases in which too heavy a focus on something like overall average utility—which should only be used as an intermediary goal—might distract us from the ultimate goal of attending to the well-being of individual sentient beings, and therefore produce erroneous ethical conclusions. We have already seen some abstract examples of when I think this is the case, so let me now explain the method of counterfactual utilitarianism which produces the results contrary to traditional utilitarian analyses.

2.    Counterfactual Utilitarianism

2.1.  The Basic Method

In effect, the metaphysical framework counterfactual utilitarianism instructs us to employ is this: when trying to determine the morality of an action, we first 1) imagine the different possible futures (or counterfactuals) which would result if the action were taken versus if it were not, then 2) compare the well-being of the sentient beings in the first counterfactual to the well-being of their counterparts in the second. It is step 2 of this procedure which sets counterfactual utilitarianism apart from other forms of utilitarianism, and developing the details of how it works will be much of the project of this paper. For now, though, as a rather trivial example, say my brother is deciding whether or not to punch me in the shoulder. He therefore needs to compare the counterfactual in which he punches me with the counterfactual in which he does not. Then he asks the question, How is Milo doing in each one? Comparing the version of me that exists in each counterfactual, one is experiencing significant pain in his shoulder while the other is not. So, he realizes it is morally preferable to bring about the counterfactual in which I am not experiencing that pain, and therefore he decides not to punch me [8].

We must also be clear about the important ethical questions that counterfactual utilitarianism does not address. Most prominently, it does not say anything about how we ought to evaluate the comparisons among counterfactuals, other than the generic utilitarian counsel that morally better actions are the ones that result in higher well-being or more positive experiences for sentient beings. For example, counterfactual utilitarianism does not provide any instruction on how to value different kinds of well-being or positive experiences (e.g. the physical pleasure of tasting a cake versus the intellectual satisfaction of solving a difficult math problem). In essence, counterfactual utilitarianism is a framework that mainly reveals what ethical judgements need to be made without supplying very many specifics on how said judgements should be made. It is meant primarily as a means of making sure moral agents are on the same page about how to conceptualize the basic form of the ethical evaluations at hand, so that they can then collectively move forward focusing their energy on questions such as how to value different sorts of well-being in different scenarios, which are precisely the kinds of important details that seriously engaged moral agents concerned about complex ethical issues should be earnestly discussing.

Hidden within the apparent simplicity of the counterfactual utilitarian method, though, are some further guidelines that occasionally come to the forefront in important ways under the right conditions. So, in order to much more fully flesh out how counterfactual utilitarianism works, let’s first revisit the examples from section 1.3, then apply it to a couple relevant bioethical cases.

2.2 Earlier Examples Revisited

Recall the case in which there exists a person—Tyler—with well-being 7/10. Articulated using the language of counterfactual utilitarianism, taking Action 1 would result in a counterfactual that contains Tyler with a well-being of 8/10. Taking Action 2 would result in a counterfactual that contains Tyler with a well-being of 7/10 along with five more people each with well-being 9/10. The counterfactual utilitarian method now dictates that we compare the well-being of each sentient individual across counterfactuals. Let’s start by focusing on Tyler. In the first counterfactual he has well-being 8/10, and in the second he has well-being 7/10, so taking Action 1 is definitely preferable for him. Now let’s think about the other five people we could create if we took Action 2. In that counterfactual they all have well-being 9/10, and in the first counterfactual… they have no well-being. I do not mean that they have well-being 0/10. I mean that they do not exist at all, and therefore that they have no well-being in this counterfactual to compare with their well-being in the second counterfactual. This is a key point that counterfactual utilitarianism reveals: when a sentient being does not exist in some counterfactual, that counterfactual cannot be said to be better or worse for that being compared to any other counterfactual. It makes no sense at all to compare the well-being of a person in one scenario to the well-being of the same person in a scenario in which they don’t exist; existing in both scenarios is a prerequisite for being able to make that comparison in the first place.

Remember, when arguing that an action is a morally good one, the question we must be able to answer is, For whom? In the example currently at hand, if someone were to claim that taking Action 2 is morally better, they certainly cannot claim that it is better for Tyler—his well-being is worse in the counterfactual associated with taking Action 2. They also, however, cannot claim that it is better for any of the other five people created by taking Action 2—the counterfactual it produces is the only situation in which these people exist at all, so it can neither be better nor worse for them compared to any other situation. So, if taking Action 2 is not morally better for anyone in particular, on what grounds could we possibly say it is better in general? Counterfactual utilitarianism reveals that Action 1 is morally preferable to Action 2, since its associated counterfactual is better for Tyler than Action 2’s, which is better for no one and worse for Tyler.

The second case in section 1.3 began with a universe made up of people all with perfect 10/10 well-being, and the consideration was whether or not to add a substantial number more of perfectly well-off people to this universe. Let’s say it has 1,000 people to begin with, and we are given the option of increasing it to 1,000,000. So, the two counterfactuals to compare consist, first, of a universe with 1,000 perfectly well-off people, and second, of a version of this universe with an additional 999,000 people. Like with the five people created by Action 2 in our previous example, the 999,000 people in this example only exist in one of the counterfactuals, so their well-being is irrelevant in our moral evaluation. And, because the 1,000 people that do exist in both counterfactuals are equally well-off in both, neither is morally preferable for them. So, these two counterfactuals are morally equivalent, and choosing to actualize either one is completely morally neutral. It is almost an identical line of reasoning that shows Parfit’s Repugnant Conclusion to be false. Although each iteration of his thought experiment adds significantly more people to his hypothetical population, the new people did not exist before and therefore cannot be said to have been made better off, but the people who did exist before have been made worse off. Clearly, then, each subsequent version of Parfit’s hypothetical population is morally worse than the prior one.

Nevertheless, there is a different ethical stance that counterfactual utilitarianism generates which many might find to be “repugnant.” Until now, the only cases I have given that involve comparisons between the existence and nonexistence of certain sentient beings are ones in which said beings did not exist in the original situation. In other words, the actions we have been evaluating have all entailed bringing into existence some number of people who would not have existed otherwise. Psychologically, this might make it easier for us to ignore their well-being in our moral assessments. It might be more difficult, however, were the action in question one which would remove any number of people from existence. But, for counterfactual utilitarianism, this is a distinction without a relevant difference. Whether the action in question will create or eliminate sentient beings, the counterfactual comparison is the same: we examine two possible futures, one in which these beings exist, the other in which they do not.

This generates a potentially very unintuitive conclusion about the morality of death: namely, that death is never a bad thing (or a good thing) for the person who dies. To be clear, dying can still be bad (or good) for them. The process of dying, for example, frequently involves a great amount of pain—an almost definitionally negative experience for the sentient being going through it—which can make the process a bad one for the being who is dying. Additionally, we still have strong reasons to think murder is bad. Even aside from the problem of the pain inflicted on the victim, one person’s death can have enormous negative psychological impacts on other people, such as the family and friends of the victim, which would make this person’s death a bad thing for these other people. So, counterfactual utilitarianism’s conclusion on the amorality of death is only this: if we understand death to be a permanent loss of sentience, when considering the moral impact that the death of a sentient being makes on that sentient being, we necessarily will be attempting to compare a counterfactual in which they exist to one in which they don’t, which makes the comparison incoherent. There can be no reason to prefer either counterfactual over the other, so long as the well-being of only that one particular sentient being is what’s taken into account.

A hypothetical designed to make this conclusion seem maximally repugnant might go something like this. Imagine the perfect torture machine. It holds one person, to whom it constantly administers unthinkable pain. It also keeps the person alive in order to suffer in perpetuity. Anything you do to turn the machine off, break it, or release its prisoner from it, will instantly kill the prisoner. Surely, anyone who happens upon this machine in action should kill the prisoner to end their suffering, right? Counterfactual utilitarianism seems to disagree. When considering whether or not to turn off or break the machine, the relevant counterfactuals are these: one in which the person continues suffering ad infinitum, and the other in which the person doesn’t exist. As we’ve already established, the second of these options cannot be better for this person, since there is no person in this scenario to be more well-off than in the alternative. So, even though it would end their suffering, removing them from the torture machine would be entirely morally neutral because it would also end their sentient experience, which eliminates the possibility of any discussion about their well-being and therefore any discussion about how they’re doing as a result of being removed from the torture machine.

Although my moral instinct is horrified by the prospect of just leaving the person to suffer, I cannot find a reason to disagree with this conclusion. Fortunately, I can still redeem counterfactual utilitarianism by pointing out that this idealized version of the hypothetical is so far removed from anything that could possibly happen in reality that it’s probably not worth dwelling on for long, and that counterfactual utilitarianism would not generate the same conclusion in a more realistic version of this case. Here’s why. First, if you came across something like this machine in real life, there is simply no way you could possibly be certain that the machine works perfectly. In other words, you could never be certain that there is no method of dismantling the machine or removing its prisoner which would not kill them—there would always be a chance (even if you think it’s small) that getting them out would result in at least a short period of time where they would be alive and not being tortured. If you managed to do this, you clearly would have made a positive well-being difference to that person—you would be actualizing a counterfactual that is exceedingly morally preferable for them over the one in which they remain in the machine. So, because of the imperfectness and uncertainty inherent in real life, we would always be right to attempt to carefully remove the prisoner from the torture machine; we might just be able to make a positive moral difference.

Moreover, the parallel but inverse stance to the uncomfortable upshot about the amorality of death is one which many of us would find intuitively right. I’m thinking of the decision not to have children. Say you live a fairly well-off life and would reasonably be able to have and raise a child who would live a happy and fulfilling life. It does not seem, however, like you would be doing a significant moral wrong by choosing not to have a child. Sure, there is a sense in which you are depriving a potential life of experiencing much happiness and satisfaction, but really all you have done is decided to actualize a counterfactual without a particular person in it, instead of one with them (in which they’re happy). Yet again, it makes no sense to say that this person is worse off in the counterfactual where they don’t exist, so it also makes no sense to say that you’ve done a moral wrong by not bringing about the counterfactual where they do. Likewise, killing someone cannot be bad for them even if it deprives them of future happiness, nor can killing someone in pain be good for them even if it ends their suffering. If it is morally neutral to bring a happy person into existence (putting aside the impact on people who already exist and will continue to exist), then it must also be morally neutral to remove a well-off or suffering person from existence (again, putting aside the impact on people who already exist and will continue to exist).

In both these cases, the well-being of a person is compared to a case in which they don’t exist, yet our intuitions are not consistent across cases. My hypothesis about the reason for this is that we are used to thinking about causing or removing happiness or suffering as affecting an existing person and their future. This is a perfectly reasonable and useful tendency that serves us well most of the time, but as such we too easily carry it into cases where it does not apply and thus gives us our warped intuitions. So, although counterfactual utilitarianism generates ostensibly undesirable conclusions in some cases, it in fact aligns closely with our intuitions in others and can even defend those intuitions in practical, real-world cases. For the remainder of this essay, I will give a more in-depth treatment of two applied cases in which counterfactual utilitarianism generates conclusions that align more closely with common intuition than other philosophers’ (even other utilitarians’) conclusions do. A close examination of these cases will also uncover additional important details about how counterfactual utilitarianism works that have not yet been addressed.

3.    Bioethical Applications

3.1.  Gene Editing & Embryo Selection

There is a debate in the field of bioethics about when and to what extent it is morally permissible (or even preferable) for parents or physicians to determine specific traits that future children will have. One method for doing this is embryo selection, in which multiple embryos (typically created by in vitro fertilization) can be genetically tested for a number of high-interest traits [9]. Parents can then select which embryo to implant, thereby determining some of the traits their child will have. Another method, which has become possible to varying degrees since its first successful undertaking in 2015 [10], is gene editing, in which a single embryo has its genome manipulated prior to implantation for the purpose of altering some trait the future child will have. “The dominant view in bioethics is that embryo selection is in general morally preferable to gene editing because the latter involves risks that are absent in the former,” [11] but two ethicists—Jeff McMahan and Julian Savulescu—argued in a paper from just last year that there is in fact a significant moral reason to prefer editing over mere selection.

Their argument, in essence, boils down to something like the following. Let’s say that a couple of parents has decided to use gene editing to remove the gene for cystic fibrosis (CF) from their child’s embryo before implanting it. In this case, the couple has performed a moral good by intervening to remove a harmful disease from a person who, without the intervention, would otherwise have had CF (and therefore been worse off). They have thus taken an action to make a positive moral difference for someone (viz. the child developing from that embryo). If, on the other hand, they were given the choice between implanting an embryo with the gene for CF versus implanting one without that gene, choosing the latter would not be performing a moral good in the same way. This is because selecting the embryo without the disease is not better for that person—it is only allowing for that person to exist (or continue existing) exactly as they already are. So, unlike in the gene editing case, there is no specific person who has been positively affected by the couple’s decision. In McMahan and Savulescu’s own words, “whereas gene editing improves the condition of one and the same individual relative to certain alternatives, embryo selection only causes a better-off individual to exist rather than a different, less well-off individual” [12]. Therefore, “‘editing out’ a genetic disorder is, in one respect, morally better than selecting an embryo that does not have that disorder. This is because editing out the disorder would be better for the subsequent person, whereas selecting the unaffected embryo would not be” [13]. This reasoning even leads them to conclude that gene editing is “morally justifiable even when it involves risks (such as ‘off-target’ mutations) for the individuals it is intended to benefit,” just as long as “the harms it is intended to prevent outweigh the possible harms it might cause” [14].

I think McMahan and Saculescu are mistaken in this analysis. To begin to see why, let us apply the counterfactual utilitarian framework to both the gene editing and embryo selection cases. In the first of these, the action in question is whether or not to edit the embryo before implantation. Let’s again say that the edit would be to remove the gene for CF. Choosing not to edit results in a counterfactual containing a child with CF. Choosing to edit results in a counterfactual containing a child without CF. Now, as per the counterfactual utilitarian method, we need to compare the well-being of the corresponding individuals across counterfactuals, and in this case the relevant individual to consider is the new child created in each counterfactual. If we agree that having CF is worse than not having it, then the child is better-off in the second counterfactual, making that counterfactual the morally preferable one for them, making the action to edit the embryo the morally preferable action for them.

Now let’s think about the case in which there are two embryos—one with the gene for CF and one without. Choosing to select and implant the first embryo results in a counterfactual containing a child with CF. Choosing to select and implant the second embryo results in a counterfactual containing a child without CF. Once again, the next step is to compare the well-being of the corresponding individuals across counterfactuals, and in this case the relevant individual to consider is (once again) the new child created in each counterfactual. If we still agree that having CF is worse than not having it, then the child is better-off in the second counterfactual, making that counterfactual the morally preferable one for them, making the action to select and implant the second embryo the morally preferable action for them.

If this analysis is correct, it should now be clear why there is not the fundamental ethical difference between gene editing and embryo selection that McMahan and Savulescu think there is. While the actions we are evaluating in each case are different, it turns out that the ultimate moral evaluation—the counterfactual comparison—is the same in both cases. In other words, all other factors being equal, gene editing and embryo selection are morally equivalent. If editing out the gene for CF is a morally good action, then selecting an embryo without the gene for CF is a morally good action for precisely the same reasons. It is, therefore, all the ‘other factors’ that should influence the decision to proceed with one method or the other. For example, related to the popular view that prefers embryo selection over gene editing, an assessment of risk is one such relevant factor. If gene editing entails the possibility of realizing a counterfactual in which the child has some unintended, undesirable side-effect, whereas embryo selection does not carry such a risk, this may be enough to prefer embryo selection. Further, this decision should also be made by taking into account other people, other than the child, whose well-being could be differently impacted by electing to edit or merely select an embryo. Choosing one or the other might be associated with any number of financial or personal health implications for the prospective parents, for instance. These are the sorts of details that parents and physicians should consider if presented with a choice between gene editing and embryo selection—not the distinction McMahan and Savulescu have advanced.

3.2. Personal Identity 

Nevertheless, I am certain that McMahan and Savulescu would not be in the slightest bit convinced by my above analysis, and perhaps for good reason—I have left some very important details very conspicuously unaddressed. Surely, they would reply after having read until this point, even if we grant the use of your counterfactual utilitarian framework, there is a crucial distinction between the gene editing and embryo selection cases. Concerning cases of the former method, comparing the well-being of the child in the counterfactual associated with making an edit with the well-being of the child in the counterfactual associated with not making said edit seems perfectly reasonable—the two possible children are different versions of the same person (as we know because they were both derived from the same embryo) and therefore can be compared across counterfactuals. The same cannot be said in cases of embryo selection. In these cases, the two possible children whose well-beings you want to compare across counterfactuals—in the same way you did in the gene editing cases—are derived from two entirely different embryos. As such, they are clearly two different people, not two different versions of the same person, and therefore comparing their well-being across counterfactuals would be a mistake.

The notion of what constitutes a person is doing a lot of work both in this imagined rebuttal and in McMahan and Savulescu’s original argument. But, for as important as the concept of personal identity is for their entire position on the ethics of gene editing and embryo selection, they never explicitly define what concept of personal identity they’re using. At first glance, they seem to understand persons as being defined by a continuous biological process that can be traced all the way back to the embryo they developed from. For instance, in one of their early examples, McMahan and Savulescu write that a hypothetical parent’s decision to wait to conceive at a later date “would have resulted in a different embryoand thus a different child coming into existence” [15]. This tight link between person and embryo evokes a sort of genidentical view of personal identity, in which “the identity through time of an entity X is given by a well-identified series of continuous states of affairs” [16]—in this case, the continuous process of biological development from embryo to adult human.

Yet, McMahan and Savulescu also allow that “if the genes of an early embryo are radically altered, the person who will develop from it will be different from the person who would have developed in the absence of the alteration” [17]. This sounds much closer to a genetic essentialist view of identity, which “says that a living thing remains the same through time in virtue of the fact that it possesses the same genome throughout” [18]. In other words, it is not being biologically continuous with a particular embryo that defines a person’s identity, but rather the content of their genome. Moreover, McMahan and Savulescu hold that some traits “would be insufficient to cause the existence of a different person” [19] if they were altered, whereas others would indeed create a new person if changed. McMahan and Savulescu call procedures that affect the former kinds of traits “identity-preserving” and the latter “identity-determining.” They list eye color as an example of an identity-preserving trait, sex as an example of an identity-determining trait, and deafness as “intermediate” [20]. Though they never address it directly, it is also implied by much of their analysis that cystic fibrosis must be identity-preserving. Finally, they write that “When gene editing is identity-determining, it is relevantly like embryo selection in that it causes one person to come into existence rather than another” [21].

These details of their position raise a number of issues. First, it is not obvious to me that sex should be identity-determining and severe diseases like CF identity-preserving. This, in conjunction with the fact that McMahan and Savulescu acknowledge the existence of intermediate traits, makes the distinction between identity-determining and identity-preserving traits incredibly fuzzy, and therefore the distinction between cases in which gene editing makes a change for a particular person and cases in which it changes which particular person exists becomes similarly fuzzy. This should constitute a non-trivial worry for McMahan and Savulescu’s account, since the crux of their analysis rests on a categorical difference between gene editing and embryo selection, but now it looks like that difference is in fact quite fluid, making the ethical difference between gene editing and embryo selection not one of kind, but one of degree. And the origin of this problem can largely be traced back to a subtle equivocation between genidentical and genetic essentialist concepts of what defines a person.

Crucially, McMahan and Savulescu never provide any justification as to why either of these concepts of personal identity is the morally relevant one, and I want to argue that neither is. Recall that one of the goals of counterfactual utilitarianism is to reapply focus on individual sentient beings. In order to say that an action is good or bad (or, as I prefer, better or worse relative to other possible actions) we need to be able to answer the question of for whom it is such, and our answer to that question must make reference to at least one subject of a sentient experience. We are now much closer to a morally relevant sense of personal identity. While “The genidentity view seems very useful if you want to understand biological identity,” [22] for example, counterfactual utilitarianism hints at a sentience-based view of personal identity that will be useful in the field of ethics. And McMahan and Savulescu clearly care about sentience, too. The whole reason they think that editing out the gene for CF is a morally good action is that they think (all else being equal) a sentient experience associated with CF is worse than one without it. Plus, it seems plausible to assume that the reason they believe significant alterations to an embryo’s genome are identity-determining is because such alterations would result in a sentient experience so radically different than the one that would have existed otherwise, that it merits considering the subject of this new sentient experience a different person than the one who would have existed otherwise. So we have now identified three different senses of personal identity which McMahan and Savulescu covertly switch between, and it is only this last one which is actually morally relevant.

Counterfactual utilitarianism, with its focus on the well-being of sentient individuals, points us toward the idea that a person should not be defined by their genome, nor by their genidentical physical body; rather, they are to be defined by their own, unique, personal sentient experience which is separate from the experience of others. Of course, a person’s genetics and physical body hugely influence their sentient experience, but so do all kinds of environmental factors, and it is the emergent experience itself which is of moral import. Consider a pair of identical twins, for example. Identical twins don’t actually have exactly identical genomes, but let’s assume for a moment that some pair does. Would those two twins then in fact be the same person? No—they are each their own person because each is having their own sentient experience. Or, imagine that we could travel to a parallel universe remarkably similar to our own. Say you found a version of yourself there—someone who shares your name, physical appearance, and most details of your life story. Are the two of you the same person? No—each of you is having your own sentient experience. What about conjoined twins? A pair of conjoined twins shares both a genome as well as a single, genidentical body. Does this make them one and the same person? Again, no—there are two separate sentient experiences and therefore two separate persons. What if we even had the ability to clone a fully developed, adult human being so that two precisely physically identical bodies existed that even underwent precisely the same sentient experiences? Even in the case where their sentient experiences are somehow identical, there are nevertheless two separate instantiations of that experience occurring, and therefore there are two separate persons [23].

Thinking about persons as being defined by their sentient experience is crucial to our moral evaluations. Let’s apply it to the cases of gene editing and embryo selection, starting with the former. If we chose to implant an embryo with the gene for CF without first editing that gene out, this will actualize a counterfactual in which a new person exists whose sentient experience includes having CF. If we do edit the gene out before implantation, the counterfactual we actualize will contain a new person whose sentient experience does not include the effects of CF. These are two different sentient experiences (and potentially radically different ones at that), and therefore two different persons. The situation looks precisely the same in an embryo selection case. Either we implant an embryo with the gene for CF and actualize a counterfactual with a new person whose sentient experience includes having CF, or we implant an embryo without that gene and actualize a counterfactual with a new person whose sentient experience does not include the effects of CF. It is not that, as McMahan and Savulescu would argue, the counterfactuals in the editing case contain two versions of the same person whereas those in the selection case contain two fundamentally different persons. No, it is more accurate to say that in both cases the action in question will determine which of two different possible persons will exist—which of two different possible sentient experiences.

But then the trouble arises of how to justify comparing their well-beings across counterfactuals. Wasn’t the whole point of the counterfactual utilitarian analysis to compare the well-being of a person in one counterfactual to their well-being in other counterfactuals? And doesn’t the possibility for this fly straight out the window as soon as we recognize that each unique sentient experience is a different person? I want to argue that there is a feature of the relationship between sentient experiences that we refer to in our everyday lives as ‘different versions of the same person’ (even if that language is ultimately technically incorrect) which is not true of the relationship between sentient experiences that we only ever refer to as being ‘different people,’ and it is this feature which we can use to ascertain which sentient experiences to compare across counterfactuals.

Imagine a pair of twins, Lux and Joey. We have already established that they are different persons because each of them is having their own, distinct sentient experience. More important for our purposes now, though, is the fact that we are not liable to get confused while performing a counterfactual utilitarian analysis and wind up comparing Lux’s well-being in one counterfactual to Joey’s well-being in a second counterfactual. We would have no trouble seeing that Lux’s well-being in the first counterfactual needs to be compared to Lux’s well-being in the second, and same for Joey. We think the alternative would be a mistake because, in our current, actual world, Lux’s sentient experience and Joey’s sentient experience exist independently and simultaneously, so we consider them different people and give them each their own name. To then go on and compare Lux’s experience in one counterfactual to Joey’s in another would be to ignore this obvious distinction we’ve made between them. Put another way, if we think about the current, actual world along with any counterfactuals in question all as ‘possible worlds,’ there is a possible world where Lux and Joey both exist—making their sentient experiences not mutually exclusive—and this gives us reason not to compare their well-beings across counterfactuals.

On the other hand, consider yet again the example of my brother deciding whether or not to punch me. In the case where he does punch me, this generates a counterfactual containing a person that we, in our everyday language, would refer to as ‘the version of Milo who got punched.’ In the case where he does not punch me, this generates a counterfactual containing a person that we would refer to as ‘the version of Milo who did not get punched.’ These are two different sentient experiences, but unlike Lux and Joey, they are mutually exclusive. The sentient experience of ‘Milo having just been punched’ simply cannot exist concurrently with the sentient experience of ‘Milo having not just been punched.’ There is no possible world in which they both exist, and this mutual exclusivity can be implemented as the criterion for determining the sentient experiences whose well-beings we compare across counterfactuals.

To see how, let’s stay for a moment longer with the case of my brother deciding whether or not to punch me. Let’s call my brother Otto. In the current, actual world we have two sentient experiences which we can call Otto and Milo. The action of not punching would result in a counterfactual containing Otto and Milo as before, and the action of punching would result in a counterfactual containing Otto and a new sentient experience which we can call Milo (Punched). So, across the two counterfactuals we’re comparing, there are three unique sentient experiences: Otto, Milo, and Milo (Punched). There are two possible worlds in which Milo and Otto coexist (the current, actual world and the first counterfactual), so they are clearly not mutually exclusive. Milo and Milo (Punched), however, are indeed mutually exclusive, and therefore they should be the experiences whose well-being we compare in our moral evaluation.

This is not a particularly interesting example of how this procedure works, though, because we could have just noticed that Otto in one counterfactual corresponds to Otto in the other, and the other pair of counterparts would have then been obvious. So imagine a slightly different scenario: Otto and I are deciding which of us should punch the other person. Now the two counterfactuals in play are the following: one which contains Otto and Milo (Punched), and one which contains Otto (Punched) and Milo. Now there are four unique sentient experiences, and we need to figure out how to pair them up as counterparts to be compared across counterfactuals. In this case, it turns out there are three candidates for counterpart pairings: Milo and Milo (Punched), Otto and Otto (Punched), and Milo (Punched) and Otto (Punched) are all mutually exclusive pairs. However, if we were to use the last of these three as one of our pairs of counterparts, that would leave Milo and Otto to be the other pair. But Milo and Otto exist concurrently in the current, actual world, which means they are not mutually exclusive and cannot be compared across counterfactuals. This leaves us with Milo and Milo (Punched) as one pair of counterparts, and Otto and Otto (Punched) as the other, which is clearly the correct result [24].

Let’s now apply this procedure to a gene editing case. A prospective mother has an embryo that has been tested and found to have the gene for cystic fibrosis. If she chooses to have the embryo edited to remove that gene before it is implanted, that action will result in a counterfactual containing a sentient experience that includes the effects of that disease—let’s call this sentient experience CF. If she chooses to implant the embryo without editing, that action will result in a counterfactual containing a sentient experience free from the effects of cystic fibrosis—let’s call this sentient experience ¬CF. Neither CF nor ¬CF exists in the current, actual world, and only one of them exists in each counterfactual. There is no scenario in which they both exist, making them mutually exclusive and thus counterparts to be compared across counterfactuals for the sake of our evaluation of the morality of editing the original embryo.

An analogous analysis can be made in an embryo selection case. Let’s imagine a prospective mother has two embryos, one which has been found to have the gene for cystic fibrosis while the other does not. If she chooses to implant the first embryo, that action will result in a counterfactual containing a sentient experience that includes the effects of that disease—let’s call it CF’. If she chooses to implant the second embryo, that action will result in a counterfactual containing a sentient experience free from the effects of cystic fibrosis—let’s call it ¬CF’. As was exactly our analysis in the gene editing case, neither CF’ nor ¬CF’ exists in the current, actual world, and only one of them exists in each counterfactual. There is no scenario in which they both exist, making them mutually exclusive and thus counterparts to be compared across counterfactuals for the sake of our evaluation of the morality of choosing to implant one embryo or the other.

Once again we find that, contrary to McMahan and Savulescu’s position, gene editing and embryo selection are not inherently morally distinct. It is not the case that the counterfactual utilitarian comparison in the latter case is between the well-beings of two different people while the comparison in the former case is between two versions of the same person. No—since a person is defined by their sentient experience, both cases make comparisons between two different persons. However, because these persons are mutually exclusive (which is the feature that defines those sentient experiences that we do, in our everyday language, typically think of as being ‘different versions of the same person’), they can be compared across counterfactuals for the sake of our moral evaluations.

3.3.  The Non-Identity Problem

The same counterfactual utilitarian procedure advanced in the previous section can also be used to provide and defend a novel solution to the non-identity problem, which has much in common with issues presented by McMahan and Savulescu. The non-identity problem concerns the ethics of bringing a person into existence whose existence is necessarily “flawed” in some way. M. A. Roberts has described the problem like this:

If a person’s existence is unavoidably flawed, then the agent’s only alternatives to bringing that person into the flawed existence are to bring no one into existence at all or to bring a different person – a nonidentical but better off person – into existence in place of the one person. If the existence is worth having and no one else’s interests are at stake, it is unclear on what ground morality would insist that the choice to bring the one person into the flawed existence is morally wrong. And yet at the same time – as we shall see – it seems that in some cases such a choice clearly is morally wrong. The nonidentity problem is the problem of resolving this apparent paradox. [25]

As such, the non-identity problem is usually articulated by presenting a case in which our moral intuition tells us that some person’s action is morally wrong, alongside an argument apparently showing that it is in fact not. Solving the problem, then, becomes a task of showing the argument to be unsound.

David Boonin lays out the problem using a fictional case of a woman named Wilma. Briefly, here are the important details of Wilma’s story. She wants to have a child, but her doctor informs her that if she conceives now, her child will have “a significant, non-terrible, irreversible handicap” [26]. The good news is that she can take a pill once a day for two months before conceiving in order to be sure that her child will not have this handicap. But Wilma decides that having to take a pill for two months is too inconvenient, so she conceives now. Assuming that having this “significant, non-terrible, irreversible handicap” is worse than not having it, what Wilma does seems wrong. But Boonin supplies an argument to show that it “is not even a little bit wrong” [27].

First, he decides to call the child that Wilma would conceive if she did so immediately (i.e. the child with the handicap) Pebbles. He decides to call the child Wilma would conceive if she took the pills first (i.e. the child without the handicap) Rocks. The first (and I think the most important) premise is this:

P1: Wilma’s act of conceiving now rather than taking a pill once a day for two months before conceiving does not make Pebbles worse off than she would otherwise have been. [28]

This is because “if Wilma instead waited two months before conceiving, then Pebbles would not have been conceived in the first place. Rocks would have been conceived instead. And since we are stipulating that the significant and irreversible handicap that Pebbles is born with is not terrible, it follows that the act of conceiving Pebbles does not cause her to live a life that is worse than no life at all (assuming that such a thing is possible), and so does not make her worse off than she would have been had Wilma not conceived her” [29]. The rest of the argument proceeds as follows:

P2: If P’s act harms Q, then P’s act makes Q worse off than Q would have been had P not done the act
…
C1: Wilma’s act of conceiving now rather than taking a pill once a day for two months before conceiving does not harm Pebbles
…
P3: Wilma’s act of conceiving now rather than taking a pill once a day for two months before conceiving does not harm anyone else
…
C2: Wilma’s act of conceiving Pebbles does not harm anyone
…
P4: If P’s act does not harm Q, then P’s act does not wrong Q
…
C3: Wilma’s act of conceiving Pebbles does not wrong anyone
…
P5: If P’s act does not wrong anyone, then P’s act is not wrong
…
C4: Wilma’s act of conceiving Pebbles is not morally wrong. [30]

The non-identity problem arises because “The premises seem right. The conclusion seems wrong” [31].

After carefully examining a wide range of ways philosophers have attempted to solve the problem, Boonin concludes that none of them are satisfactory. Thus, his proposed solution is to “simply accept the argument’s conclusion” [32]. Once we do that, “The non-identity problem, understood as a puzzle that stands in need of resolution, disappears, and is replaced by the non-identity argument, which is simply a sound argument for an admittedly surprising conclusion” [33]. I, however, think we can still reject this argument’s conclusion, and by now you might already be able to anticipate how counterfactual utilitarianism shows us we can.

Think back once more to the case of my brother deciding whether or not to punch me. There are two counterfactuals to be compared: one which contains Milo and one which contains Milo (Punched). These are two different sentient experiences, just like Pebbles and Rocks are. So, could we technically say (in parallel fashion to the language of P1) that my brother’s act of punching me does not make Milo (Punched) worse off than he otherwise would have been? I suppose, technically, yes. But that is not a helpful way to understand the situation. Beginning our ethical analysis this way is going to end up implying that (following the non-identity argument) if my brother chooses to punch me, he has done nothing wrong because he has not harmed Milo (Punched) or anyone else. This line of reasoning could even be used to show that literally no actions are morally wrong, since they only ever bring about sentient experiences that would not have existed otherwise and therefore cannot be said to have been harmed by being made worse off than they otherwise would have been. So, clearly something has gone wrong right from the start in the way we set up our moral reasoning.

Counterfactual utilitarianism reveals that it is a red herring that Milo (Punched) is not made worse off by my brother choosing to punch me. Milo and Milo (Punched) are mutually exclusive sentient experiences, and therefore they are the persons whose well-beings we compare in order to judge the morality of the action of my brother punching me. Pebbles and Rocks are also mutually exclusive sentient experiences—neither exists in the current, actual world, and only one of them exists in each counterfactual (the counterfactuals associated with Wilma conceiving now versus taking the pills first). So, they are the persons whose well-beings we compare in order to judge the morality of the action of Wilma conceiving without taking two months of pills. And if we agree that the well-being of the person in the counterfactual associated with waiting to conceive is higher than the well-being of the person in the counterfactual associated with conceiving immediately (i.e. if Rocks is better off than Pebbles), then the former is the the morally better action.

Part of the reason these two cases feel different might be because the two sentient experiences we’re comparing in my case have basically the same name (Milo vs Milo (Punched)), while the two sentient experiences we’re comparing in Boonin’s case have totally different names (Pebbles vs Rocks). This might make us psychologically more likely to accept the validity of the comparison in my case than in Boonin’s (similar to how McMahan and Savulescu would probably think we could do a counterfactual utilitarian comparative analysis for gene editing cases but not embryo selection cases). But, we see that the two cases are in fact not relevantly different when we align personal identity with sentient experience and use mutual exclusivity as the key relational feature to determine which persons can be taken as counterparts across counterfactuals in our moral analyses. Ultimately, it is wrong for my brother to punch me for the same reasons it is wrong for Wilma to conceive without taking her pills once a day for two months. Thus, the non-identity problem is solved.

In an attempt to defend his position, Boonin presents two further cases which he thinks are morally equivalent to Wilma’s case, but for which we would not think that the primary action taken in the case is morally wrong. The first is the case of Fred, who “is walking past a lake in which two young boys, Billy and Timmy, are drowning” [34]. Fred has only one life preserver and only enough time to use it to save one of the boys before the other drowns. Billy is closer to Fred, making the decision to save Timmy slightly more inconvenient for Fred. Fred also happens to know that Billy is blind whereas Timmy is not. For his own convenience, Fred chooses to save Billy.

Let’s perform a counterfactual utilitarian analysis of Fred’s options. He could have elected to walk down the shore a bit and save Timmy. This action would have resulted in a counterfactual in which Timmy survives but Billy does not—meaning Timmy still exists but Billy does not. Let’s call that Counterfactual-T. Fred’s other option (which he in fact ends up choosing) was to save Billy, resulting in a counterfactual in which Billy exists but not Timmy. Let’s call that Counterfactual-B. So, we have one counterfactual containing a sentient experience (a person) called Timmy, and another containing a sentient experience (a person) called Billy. Can we compare the well-being of these two persons in order to decide which counterfactual and therefore which action is preferable? Although only one of them exists in each of the counterfactuals, both Billy and Timmy exist in the current, actual world; there is a possible world in which both these sentient experiences exist, making them not mutually exclusive and therefore not counterfactual counterparts.

There is no counterpart of Billy in Counterfactual-T, which means that attempting to use Billy in a counterfactual utilitarian evaluation of this case would be to compare his well-being in Counterfactual-B to… no one in Counterfactual-T. This would be a comparison between existence and nonexistence, which we established early on is ethically incoherent. We would get the same incoherent comparison if we tried to use Timmy’s well-being in our evaluation of this case, because he has no counterpart in Counterfactual-B. The only person who has counterparts in both counterfactuals is Fred, and if the well-being of his counterpart in Counterfactual-B is slightly higher than the well-being of his counterpart in Counterfactual-T, due to the convenience of not having to walk down the shore to save Timmy, then he is justified in choosing to save Billy. This makes Fred’s case markedly different from Wilma’s.

The second case Boonin puts forward is in an “admittedly somewhat less straightforward” [35] adaptation of the Fred case. In this case the main character is Barney, who is “walking past a lake in which two young women are drowning” [36]. Like in Fred’s case, Barney can only save one person. Barney also somehow knows that each woman “has just had sexual intercourse and conception will occur in each of them in approximately an hour, if they do not die first. One of the women will conceive an incurably blind child. The other will conceive a child who is not blind. The woman who will shortly conceive a blind child is the one who is closer to Barney. Barney decides to save the woman who will shortly conceive a blind child because this is more convenient for him” [37]. In this case, neither of the possible future children currently exist, nor is it possible for both of them to exist since Barney can only save one mother, so unlike Billy and Timmy they are indeed mutually exclusive sentient experiences. As such, I agree with Boonin’s assessment that Barney’s case is relevantly like Wilma’s. I am therefore also willing to defend the position that the morally better action for Barney to take is for him to be slightly inconvenienced by walking down the shore in order to save the mother who is farther away from him.

I will admit that this conclusion does strike my intuition as a bit odd. I expect much of the reason for that, though, has to do with the fact that for this conclusion to go through, it requires that Barney somehow know intimate details not only about the women he happens upon in the lake (viz. that they are both just on the verge of conceiving) but also about their potential children (viz. that one of them will be blind) who do not exist even as embryos yet. To me, it feels like a strain on the imagination to think that Barney (or anyone, for that matter) could possibly know these things in a real-life scenario; therefore, in any real-life scenario approximating this hypothetical case, we would almost certainly not be able to say that saving one or the other of the women is morally preferable. If that’s true, then this doesn’t seem like a good case to use for double-checking our intuitions about cases like Wilma’s.

Additionally, any remaining hesitation my moral intuition has about accepting my conclusion to Barney’s case can be removed by imagining that the harmful trait the first mother’s child has is something much worse than blindness—say, Tay-Sachs disease or some other painful genetic disorder that unquestionably makes a person significantly worse off for having it. In this modified case, where there are two possible persons who could exist, and their existences are mutually exclusive, and one of their well-beings is clearly significantly worse than the other’s, and the action Barney takes will determine which will exist, then my intuition is restored that this case is indeed relevantly like Wilma’s and that both Barney and Wilma would be making the morally worse choice by taken the action which is slightly more convenient for them.

However, this particular modification to the case (swapping out blindness for Tay-Sachs disease) brings to the table one final philosophical issue worth addressing in this paper: the distinction between ‘a life worth living’ and ‘a life not worth living.’ These notions came up in our brief look at Parfit’s Repugnant Conclusion (the final stage of which is made up of an enormous number of people, all of whom “have lives that are barely worth living” [38]) in section 1.3. They also play a role in Boonin’s set-up of Wilma’s case, since P1 relies on the position that “the act of conceiving Pebbles does not cause her to live a life that is worse than no life at all (assuming that such a thing is possible),” [39] and therefore Boonin says that Pebbles’ condition will be something “non-terrible” [40]. In Barney’s case, some utilitarians might want to interject that changing the possible child’s future condition from blindness to Tay-Sachs disease does not change the ethics of the situation by degree, but by kind. They might argue that while being blind is worse than being able to see, it is not enough to make the child’s life not one worth living—i.e. worse than not existing at all—and therefore the child is not harmed by being brought into existence. But, having Tay-Sachs disease is enough to make their life one that is not worth living, and therefore it would be a harm to them to be brought into existence. So, our judgment of Barney’s action totally depends on this detail. In other words, changing the possible child’s future condition from blindness to Tay-Sachs disease makes a relevant moral difference, and thus I cannot use our intuitions about the morality of Barney’s action in the case where the child would have Tay-Sachs disease to support a conclusion about the morality of Barney’s action in the case where the child would be blind.

But counterfactual utilitarianism shows that this rebuttal holds no water. A ‘life not worth living’ is understood as one which is worse than no life at all; if a person would be better-off not existing, then they have a ‘life not worth living.’ But again, we established in sections 2.2 and 2.3 that such comparisons between existence and nonexistence of a certain sentient experience are totally incoherent. We can only compare the well-being of one sentient experience to the well-being of another sentient experience which is their counterpart. So, it does not matter how bad the trait is that the first possible child has in Barney’s case in order for our basic moral conclusion to remain the same. As long as the first possible child is worse off than the second possible child, then it is morally preferable for Barney to save the mother who will conceive the second child. This conclusion is made particularly clear when we imagine that the first possible child would have Tay-Sachs disease, and therefore we should have renewed confidence that the conclusion holds if the first possible child would be blind, too.

To summarize this section: the main take-away is that we can say that Wilma’s chosen action is morally worse than her other option, and therefore (in opposition to what Boonin argues) we do not have to accept the conclusion of the non-identity argument. Furthermore, after looking at Boonin’s two cases which he thinks are morally equivalent to Wilma’s case, we discovered that Fred’s case is in fact not morally equivalent to Wilma’s case. Of Barney’s case, we found that as long as we grant the strange details which would probably not be achieved in a real-world version of the case, we can say that it is indeed morally equivalent to Wilma’s case, and therefore we can say that Barney’s chosen action was morally worse than his other option.

4.    CONCLUSION

Counterfactual utilitarianism is a consequentialist framework designed to set up the form of the ethical evaluations we should utilize when considering the morality of an action. It takes as jumping-off points the baselines that 1) the morality of actions depend on the consequences they have for the well-being of sentient beings, and that 2) the better actions are ones which result in the highest well-being for the sentient beings involved, both of which identify it as a broadly utilitarian framework. However, because I think that many utilitarians go astray when trying to maximize well-being “from the point of view of the universe” without paying attention to the specific sentient beings whose well-beings are affected, counterfactual utilitarianism aims to reapply focus on individual sentient beings. In fact, it highlights that a morally relevant sense of personal identity is one which identifies persons with individual, unique sentient experiences. The framework functions basically like this: we imagine actions as being associated with counterfactuals they would bring about, and we evaluate which actions are morally better or worse by comparing those counterfactuals and evaluating them according to the well-beings of the persons which have counterparts in multiple of said counterfactuals.

Counterfactual utilitarianism carries the implication that if a person has no counterpart in some counterfactual, then that counterfactual (and therefore the action that would generate it) cannot be said to be better or worse for that person than any other counterfactual (and therefore the actions that would generate them). This is because we cannot compare the well-being of a person to the well-being of not existing; to not exist in a morally relevant sense is to have no sentient experience, which means there is no person and therefore no well-being available to make the ethical comparison with. This might seem like a bizarre conclusion with worrying upshots, but we saw that the complexities and uncertainties inherent to real-life cases make it not something to worry about in practice.

On the other hand, counterfactual utilitarianism provides us the reasoning behind ethical views on many practical cases that we are probably inclined to already believe are correct. For example, while Jeff McMahan and Julian Savulescu have recently tried to argue that gene editing can make moral differences while embryo selection never can, counterfactual utilitarianism reveals that gene editing and embryo selection are in fact essentially morally equivalent. We saw that much of the problem with McMahan and Savulescu’s argument boiled down to conflating multiple senses of personal identity, and once we focus on sentient experience and apply the counterfactual utilitarian method to both gene editing and embryo selection cases, we see that the basic moral evaluations we end up making in both cases are the same.

We also saw that counterfactual utilitarianism supports our intuition that the conclusion of the argument associated with the non-identity problem is mistaken. Using Boonin’s framing of the issue, we found that counterfactual utilitarianism gives us reason to argue that Wilma’s choice to conceive immediately is a morally worse choice waiting to conceive until after taking a pill once a day for two months. This is because the person she would conceive now would have a lower well-being than the person she would conceive after taking pills for a couple months, and since those persons are mutually exclusive sentient experiences, they are the counterparts whose well-beings we compare in order to judge the morality of Wilma’s action.

5.    REFERENCES

Arrhenius, Gustaf, Jesper Ryberg, and Torbjörn Tännsjö. “The Repugnant Conclusion.” In The Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta and Uri Nodelman, Winter 2022. Metaphysics Research Lab, Stanford University, 2022. https://plato.stanford.edu/archives/win2022/entries/repugnant-conclusion/.

Beauchamp, Tom L., and James F. Childress. Principles of Biomedical Ethics. Seventh Edition. New York and Oxford: Oxford University Press, 2013.

Boonin, David. “How to Solve the Non-Identity Problem.” Public Affairs Quarterly 22, no. 2 (2008): 129–59.

Hume, David. A Treatise of Human Nature. Edited by David Fate Norton and Mary J. Norton. Oxford and New York: Oxford University Press, 2000.

Jubien, Michael. Possibility. New York: Oxford University Press, 2009.

Lewis, David. On the Plurality of Worlds. Oxford and New York: Basil Blackwell, 1986.

Liang, Puping, et. al. “CRISPR/Cas9-Mediated Gene Editing in Human Tripronuclear Zygotes.” Protein & Cell 6, no. 5 (2015): 363–72. https://doi.org/10.1007/s13238-015-0153-5.

McMahan, Jeff, and Julian Savulescu. “Reasons and Reproduction: Gene Editing and Genetic Selection.” The American Journal of Bioethics, September 11, 2023, 1–11. https://doi.org/10.1080/15265161.2023.2250288.

Parfit, Derek. Reasons and Persons. New York: Oxford University Press, 1984.

Pradeu, Thomas. “Genidentity and Biological Processes.” In Everything Flows: Towards a Processual Philosophy of Biology, edited by Daniel J. Nicholson and John Dupré, 96–112, 2018.

Roberts, M. A. “The Nonidentity Problem.” In The Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta and Uri Nodelman, Winter 2022. Metaphysics Research Lab, Stanford University, 2022. https://plato.stanford.edu/archives/win2022/entries/nonidentity-problem/.

Sermon, Karen, André Van Steirteghem, and Inge Liebaers. “Preimplantation Genetic Diagnosis.” The Lancet 363, no. 9421 (2004): 1633–41. https://doi.org/10.1016/S0140-6736(04)16209-0.

Sidgwick, Henry. The Methods of Ethics. New York: Dover Publications, Inc., 1966.

Sinnott-Armstrong, Walter. “Consequentialism.” In The Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta and Uri Nodelman, Winter 2023. Metaphysics Research Lab, Stanford University, 2023. https://plato.stanford.edu/archives/win2023/entries/consequentialism/.

Notes