So all right, we’ve clarified the importance of distinguishing between beings that exist and beings that don’t exist, and what the implications are for population ethics. But what about those cases that are right on the borderline? What about edge cases between beings that don’t exist and beings that do, like fetuses developing in the womb? Does this framework have anything useful to say about moral dilemmas pertaining to these edge cases, like abortion?
I think it does. To be sure, this is a topic with a lot of gray areas; but I think the approach I’ve been describing provides us with a way of navigating these gray areas in a way that’s more logical and systematic than just working off our intuitions alone.
As usual, it all comes back to preference satisfaction. Obviously, in the cases where a new person has already been born and has developed their own preferences (even if those preferences are still relatively rudimentary), the morality is clear-cut; there’s no question that they’re a real person whose preferences have to be recognized. Killing babies is unmistakably bad; I don’t think any ethical person would dispute that.
Where things get a little more controversial, of course, is when we start talking about the period between conception and birth – especially the earliest part of this period. If you’re someone who believes that the fertilized egg is endowed with a soul at the moment of conception, and that killing anything with a soul is inherently immoral, then you might think it’s just as bad to destroy a five-day-old embryo as it is to kill a five-year-old child (although if you believed that the soul would persist even after the body was destroyed, it does raise the question of why destroying the body would be so bad (more on that topic in my previous post)). But under the system I’ve been describing here, an action can only be wrong if it violates someone’s preferences – and based on everything science has shown us about embryology, there’s no reason to believe that a pre-sentient embryo is capable of having preferences. During that initial stage of its development (before it develops into a fetus), the embryo is just a clump of cells; it hasn’t formed a brain yet, so it lacks the ability to think or feel or experience anything. In other words, there’s no actual person there yet, much less one capable of holding preferences – so that means there’s no inherent harm in destroying the embryo, aside from whatever effects it might have on the parents and so on. Benefits and harms can only accrue to people who actually exist at some point; if the embryo never develops into such a person, then there’s no basis for including it in the moral calculus.
(There’s an argument often made that early-term abortion is immoral because the embryo is a potential person, and it’s wrong to destroy something that would have become a person without your intervention; but this argument doesn’t really hold up in any other context. Imagine, for instance, if you had a futuristic machine that would allow you to construct a new human being from scratch, from the ground up, in about six hours. If you started running this machine one day, but then changed your mind a moment later and pressed the cancel button before the machine had finished constructing the first sliver of a toenail, would this act be equivalent to outright murder? In fact, would there be anything wrong with it at all, aside from the trivial waste of having to throw away a toenail clipping? It doesn’t seem to me that merely preventing the existence of someone who otherwise would have existed is enough to constitute an immoral act; in order for it to be wrong, there has to actually be a sentient being there whose preferences you’re violating.)
Now, there is an important corollary to this point, which is that if the embryo actually were expected to develop into a person – i.e. if no abortion were performed – then the embryo would merit moral consideration, just like any other expected future person would. For the same reason that it would be immoral to, say, pollute the air for future generations – since it’d be violating their preference for good health – it would be immoral for a pregnant woman to smoke or take hard drugs, even during the earliest stages of her pregnancy, if she expected to actually have her child – since it’d be violating the future child’s preference for good health. It might seem odd to think that it could be morally worse to merely damage a pre-sentient embryo by smoking than it would be to destroy it outright. But remember, preventing new preferences (and new beings) from coming into existence isn’t the same thing as violating preferences that will exist; only the latter can be considered inherently immoral. So while the act of aborting a pre-sentient embryo can’t be considered immoral because it doesn’t violate any sentient beings’ preferences (and there can be no obligation to satisfy a preference that never exists), the act of smoking while pregnant does violate the future child’s preferences, so it can be considered immoral. If there’s one possible outcome in which a child is born, and another possible outcome in which no child is born, then (all else being equal) that fact alone isn’t enough to make one outcome worse than the other; but if there’s one possible outcome in which a child is born with a higher utility level, and another possible outcome in which a child is born with a lower utility level, then that is enough to make the latter outcome morally worse.
(Readers of Parfit might be tempted to bring up the Non-Identity Problem at this point: What if some particular child could only be born under circumstances that would cause it to suffer more harm than a child born under better circumstances? For instance, what if Child X was only conceived because its mother decided not to quit smoking before getting pregnant – whereas if she’d decided to quit first, she would have delayed her pregnancy for a few months and wound up conceiving a totally different child, Child Y? Sure, Child Y’s quality of life would have been better than Child X’s (because of its better health resulting from having a non-smoking mother), but how can we say that Child X is worse off because its mother smoked, when it never would have existed in the first place if not for that fact? The thing is, though, the utility calculus doesn’t make any such distinctions between the specific identities of the people whose preferences it quantifies; all it counts is the preferences themselves. As far as the calculus is concerned, the two alternatives in this scenario are either (1) a child is born with a preference for good health and a preference to continue existing, and both of those preferences are fully satisfied, or (2) a child is born with a preference for good health and a preference to continue existing, and only the latter preference is fully satisfied. (Remember, a hypothetical child’s initial preference to be brought into existence in the first place can’t be included in the calculus, because it wouldn’t be possible for such a preference to exist prior to the child itself existing.) According to the utility calculus, then, the first outcome is morally better, and there’s no need to even bring the preference holders’ specific identities into the equation at all. The Non-Identity Problem, under this model, never even comes up in the first place.)
At any rate, this might all seem fairly straightforward when it comes to pre-sentient embryos in the earliest stage of development, before they ever develop brains or the ability to hold preferences. But at what point, exactly, does the transition occur between an unthinking clump of cells with no inherent right to exist and a thinking, feeling person whose preferences have to be fully respected? When is that critical threshold of sentience crossed, and how exactly does the transition work in terms of the utilitarian calculus?
Well, when it comes to the biology of it, there really isn’t a critical threshold at which a person’s sentience instantaneously “switches on;” as Michael S. Gazzaniga explains, it’s much more of a gradual process:
As soon as sperm meets egg, the embryo begins its mission: divide and differentiate, divide and differentiate, divide and differentiate. The embryo starts out as the melding of these two cells and must eventually become the approximately 50 trillion cells that make up the human organism. There is no time to lose – after only a few hours, three distinct areas of the embryo are apparent. These areas become the endoderm, mesoderm, and ectoderm, the initial three layers of cells that will differentiate to become all the organs and components of the human body. The layer of the ectoderm gives rise to the nervous system.
As the embryo continues to grow in the coming weeks, the base of the portion of the embryo called the neural tube eventually gives rise to neurons and other cells of the central nervous system, while an adjacent portion of the embryo called the neural crest eventually becomes cells of the peripheral nervous system (the nerves outside the brain and spinal cord). The cavity of the neural tube gives rise to the ventricles of the brain and the central canal of the spinal cord, and in week 4 the neural tube develops three distinct bulges that correspond to the areas that will become the three major divisions of the brain: forebrain, midbrain, and hindbrain. The early signs of a brain have begun to form.
Even though the fetus is now developing areas that will become specific sections of the brain, not until the end of week 5 and into week 6 (usually around forty to forty-three days) does the first electrical brain activity begin to occur. This activity, however, is not coherent activity of the kind that underlies human consciousness, or even the coherent activity seen in a shrimp’s nervous system. Just as neural activity is present in clinically brain-dead patients, early neural activity consists of unorganized neuron firing of a primitive kind. Neuronal activity by itself does not represent integrated behavior.
During weeks 8 to 10, the cerebrum begins its development in earnest. Neurons proliferate and begin their migration throughout the brain. The anterior commissure, which is the first interhemispheric connection (a small one), also develops. Reflexes appear for the first time during this period.
The frontal and temporal poles of the brain are apparent during weeks 12 to 16, and the frontal pole (which becomes the neocortex) grows disproportionately fast when compared with the rest of the cortex. The surface of the cortex appears flat through the third month, but by the end of the fourth month indentations, or sulci, appear. (These develop into the familiar folds of the cerebrum.) The different lobes of the brain also become apparent, and neurons continue to proliferate and migrate throughout the cortex. By week 13 the fetus has begun to move. Around this time the corpus callosum, the massive collection of fibers (the axons of neurons) that allow for communication between the hemispheres, begins to develop, forming the infrastructure for the major part of the cross talk between the two sides of the brain. Yet the fetus is not a sentient, self-aware organism at this point; it is more like a sea slug, a writhing, reflex-bound hunk of sensory-motor processes that does not respond to anything in a directed, purposeful way. Laying down the infrastructure for a mature brain and possessing a mature brain are two very different states of being.
Synapses – the points where two neurons, the basic building blocks of the nervous system, come together to interact – form in large numbers during the seventeenth and following weeks, allowing for communication between individual neurons. Synaptic activity underlies all brain functions. Synaptic growth does not skyrocket until around postconception day 200 (week 28). Nonetheless, at around week 23 the fetus can survive outside the womb, with medical support; also around this time the fetus can respond to aversive stimuli. Major synaptic growth continues until the third or fourth postnatal month. Sulci continue to develop as the cortex starts folding to create a larger surface area and to accommodate the growing neurons and their supporting glial cells. During this period, neurons begin to myelinate (a process of insulation that speeds their electrical communication). By the thirty-second week, the fetal brain is in control of breathing and body temperature.
By the time a child is born, the brain largely resembles that of an adult but is far from finished with development. The cortex will continue to increase in complexity for years, and synapse formation will continue for a lifetime.
That is the quick and easy neurobiology of fetal brain development. The embryonic stage reveals that the fertilized egg is a clump of cells with no brain; the processes that begin to generate a nervous system do not begin until after the fourteenth day. No sustainable or complex nervous system is in place until approximately six months of gestation.
If you had to pick the exact point at which the fetus becomes a person entitled to protection, then – if only for legal purposes – the most reasonable choice would probably be somewhere around the end of the second trimester (which, conveniently enough, is pretty much in line with current US laws as of 2021). Legal necessities aside, though, it seems clear that the actual level of sentience for a developing fetus at any given moment can’t be reduced to a simple black-and-white binary – it exists along a continuum. So what does this mean for the fetus’s moral status? Well, it means that it shouldn’t be reduced to a simple black-and-white binary either. Just like its level of sentience, the fetus’s moral status should also be regarded as something that emerges gradually. As Parfit writes:
[Under the binary view,] there must be a sharp borderline. It is implausible to claim that this borderline is birth; nor can any line be plausibly drawn during pregnancy. We may thus be led to the view that I started to exist at the moment of conception. We may claim that this is the moment when my life began.
[But according to a more gradualist view,] we can […] deny that a fertilized ovum is a person or a human being. This is like the plausible denial that an acorn is an oak-tree. Given the right conditions, an acorn slowly becomes an oak-tree. This transition takes time, and is a matter of degree. There is no sharp borderline. We should claim the same about person, and human beings. We can then plausibly take a different view about the morality of abortion. We can believe that there is nothing wrong in an early abortion, but that it would be seriously wrong to abort a child near the end of the pregnancy. Such a child, if unwanted, should be born and adopted. The cases in between we can treat as matters of degree. The fertilized ovum is not a first, but slowly becomes, a human being, and a person. In the same way, the destruction of this organism is not at first but slowly becomes seriously wrong.
In this sense, it’s actually a lot like what we were talking about before with the continuum of sentience across animal species, ranging from minimally-sentient animals like insects at the lower end all the way up to humans at the higher end – only instead of this continuum stretching across all the different species of the animal kingdom, here it occurs entirely within one person, stretching across the duration of their gestation period from conception to birth. Just as killing a (barely-sentient) mosquito would be less egregious than killing a (more sentient) mouse, which in turn would be less egregious than killing a (still more sentient) human, aborting an (entirely non-sentient) embryo would be less egregious than aborting a (somewhat sentient) fetus, which in turn would be less egregious than killing a (still more sentient) newborn baby. There’s no specific point at which the act goes from being totally acceptable to totally immoral – it’s not black-and-white like that – there’s a whole spectrum of shades of gray in between. Like Parfit says, it’s all a matter of degree.
For the most part, I think this is pretty intuitive. Most of us would probably agree that if there were, say, an emergency situation in which we could only save the life of either a pregnant woman or her fetus, the mother’s life would take priority over the fetus’s, because she’d be more sentient and her death would therefore be worse. We’d probably also agree that if someone performed a late-term abortion, that would be worse than if they’d preformed an early-term abortion, for the same reason. Having said that, though, we can’t just take it for granted that sentience level must be the one and only determinative factor in these kinds of cases. After all, I mentioned an example earlier saying that it would be better to save an infant’s life than a chimp’s life, despite the chimp’s greater level of sentience, because the infant would derive more utility from being granted a longer and fuller life. So if we accept that premise, then clearly the sentience level of the individual(s) in question can’t be the be-all end-all; their expected future lifespan and quality of life must also play a role somehow. This also explains why, to return to another example I gave a moment ago, it would be better to let a turkey live out its full lifespan than to kill it prematurely; the turkey would derive more utility from being granted longer lifespan, so therefore the most moral thing to do would be to give it the longest lifespan possible.
But if the thing determining the morality of cases like these is expected future utility, rather than present level of sentience, then would that imply that saving the life of someone who was younger would always have to be better than saving the life of someone who was older, regardless of the immediate difference in sentience levels between them? And if so, then would that mean that saving the life of a pregnant woman would actually be worse than saving the life of a fetus, and that early-term abortion would actually be worse than late-term abortion (and for that matter, worse than infanticide)? What’s going on here? Which standard really determines what’s moral – expected future utility, or present level of sentience?
Actually, I think the only sensible explanation here is one that incorporates both. How does that work, exactly? Well, think of it this way: It’s true that in a vacuum, a person’s preference to (say) live for 65 more years will have more expected utility associated with it than an equally sentient person’s preference to live for five more years (assuming roughly comparable quality of life). So for that reason, it would be worse to kill a 15-year-old than to kill a 100-year-old, all else being equal. But crucially, the reason it would be worse wouldn’t necessarily be because of all the future utility you’d be preventing from coming into existence by killing the 15-year-old; remember, after all, preventing new preferences from coming into existence can’t be counted as a negative in the same way that violating existing preferences counts as a negative. The only preferences that can be counted in the utility calculus for your decision are those that are actually satisfied or violated by that decision. And that means that the only real reason why it’s worse to kill the 15-year-old must be the fact that their existing preference structure ascribes more value to the outcome of being given 65 more expected years of life than the 100-year-old’s preference structure ascribes to the outcome of being given five more expected years of life. In other words, the utility calculus itself isn’t directly counting all the possible future utility that may or may not come into existence for each individual; rather, the individuals are incorporating that expected future utility into their existing preferences, and those preferences are what the utility calculus is counting. It’s basically just like what we’ve been discussing about the morality of bringing a new person into existence: Such a decision can’t be justified just on the basis of all the hypothetical preferences a new person would be able to satisfy if they existed; it has be justified on the basis of whether bringing them into existence would satisfy existing preferences. And likewise, the decision to allow a living person to continue to live can’t be justified just on the basis of all the preferences they’d be able to satisfy if they were allowed to continue to live; it has to be justified on the basis of whether allowing them to continue living would satisfy their existing preference to do so. The moral calculation in both of these cases is functionally the same; it’s just that in the latter case, the “new person” whose preferences are being created is “an existing person’s future self.” (And this makes sense, considering that the utilitarian calculus doesn’t recognize the particular identities of the beings whose preferences it’s quantifying. It’s only concerned with the preferences themselves – so for its purposes, there’s no difference between a preference belonging to someone who was just created and a preference belonging to someone who’s existed for 50 years.)
In short, then, the fact that the 15-year-old would derive more utility from an additional 65 years of life than a 100-year-old would derive from an additional five years of life isn’t the determinative factor here; what matters is that the 15-year-old prefers this outcome more than the 100-year-old prefers the alternative. It’s the strength of those held preferences, not the amount of expected future utility that causes them to be held as strongly as they are, that determines whose life should be prioritized. (The expected future utility does still matter, but only in this second-hand way; it isn’t directly counted in its own right.)
And granted, when we’re comparing people whose sentience levels are roughly the same, this distinction is usually a moot point. Usually, the outcome that produces the greatest expected future utility (in this case, the 15-year-old being spared) is also the one that’s most strongly preferred. In fact, even in cases where it wouldn’t seem to be the most strongly preferred outcome – like if the 15-year-old were overly pessimistic about their future and didn’t value it very highly at the object level – their Master Preference would typically supersede whatever mistaken object-level judgments they might make, so (given that their future prospects were actually better than they assumed) their extrapolated preference to continue living would still be stronger than that of the 100-year-old. Of course, that doesn’t mean it would be impossible for a 100-year-old’s preference to ever be greater than a 15-year-old’s; if the 100-year-old’s remaining five years of life were expected to be unbelievably amazing, for instance, while the 15-year-old’s remaining 65 years of life were expected to be barely worth living, then it really might be possible that sparing the 100-year-old’s life would be better (since it would produce greater utility, and the 100-year-old’s extrapolated preference to experience it would therefore be greater than the 15-year-old’s extrapolated preference to live out their own remaining lifespan). But the vast majority of the time, this won’t be the case – the younger person’s expected future utility will be greater than the older person’s – so the fact that the two individuals are equally sentient (and are accordingly capable of desiring future utility equally strongly) will mean that the younger person’s interests will win out.
That being said, though, the reason why it’s so important to draw this distinction between a person’s expected future utility on the one hand, and their preference to continue living so that they can experience that future utility on the other hand, is that when two individuals aren’t equally sentient, this disparity might mean that the more sentient individual’s preference to continue living can outweigh that of the less sentient individual, even if their expected future utility is lower, simply because they’re able to hold their preferences so much more strongly. For example, if you took a 15-year-old with a remaining life expectancy of 65 years, and compared them to a barely-sentient fetus with a life expectancy of 80 years, then (assuming comparable quality of life) the fetus would clearly get more future utility from being allowed to live than the 15-year-old would – but because the fetus was so much less sentient, it wouldn’t actually prefer that outcome as much as the 15-year-old would prefer the alternative, because it wouldn’t be capable of holding any preference very strongly (not even its Master Preference). It’d be like comparing an adult human’s preferences to those of a mosquito; no matter how much the mosquito might prefer some particular outcome, it could never prefer it as strongly as the human would be capable of preferring something – it just wouldn’t be sentient enough – so ultimately, its interests wouldn’t count for as much in the moral calculus.
Here’s another way of thinking about it: Imagine if you had two audio recordings – one of someone talking normally, and another of them yelling loudly. As far as the recordings themselves went, the second one would obviously be louder than the first one. But if you were playing the second recording on a tiny, weak speaker, while playing the first one on a giant, powerful speaker, then it might be possible for the first one to be louder, simply because the smaller speaker wouldn’t be able to reach as high a volume level as the larger one. In the same way, a preference for a particular outcome which, in a vacuum, might be associated with a higher level of expected future utility (the equivalent of the audio in the recording itself being louder) might turn out to carry less weight in the moral calculus if the individual holding it doesn’t actually prefer it that strongly due to their lower level of sentience (the equivalent of having a smaller, weaker speaker that can only reach a limited volume level).
It’s for this reason, then, that our aforementioned intuitions about the different levels of moral consideration merited by different types of individuals – adults, children, fetuses, non-humans, etc. – can be justified. We can say with total consistency that (all else being equal) aborting a barely-sentient fetus would be less egregious than killing a full-grown human – or even a non-human individual like a chimp – while at the same time acknowledging that as that fetus grew and developed into a child, its increasing level of sentience would allow it to prefer continued survival more strongly than before, so that at some point its preference for continued survival would come to outweigh that of those older individuals with their shorter remaining life expectancies. When exactly these intersections would occur is debatable, of course, but my own personal intuition is that the version of this model that says it would be better to spare a chimp than a barely-sentient fetus, better to spare a newborn than a chimp, and better to spare all of the above (except maybe the fetus, depending on how barely-sentient it was) than a 110-year-old on their death bed, seems right overall.
Now, having said all that, it’s worth noting that up to this point I’ve only been considering things from the perspective of the actual individuals whose lives are at stake – and while that’s often the biggest factor in these kinds of decisions, it isn’t always. When it comes to issues like abortion, there are always other individuals whose interests are involved – most notably the would-be parents – and their preferences have to be included in the moral calculus as well. The mere fact that the fetus would prefer to live is not, in itself, sufficient grounds to justify keeping it alive in every case; that preference has to actually outweigh whatever other preferences might be opposed to it. If the parents, for instance, have a strong enough preference not to carry the pregnancy to term, and the fetus’s level of sentience is still low enough that its preference to continue existing isn’t that strong, then it’s entirely possible that the parents’ preferences might outweigh those of the fetus, making it morally justifiable for them to abort the pregnancy. Needless to say, of course, once the fetus grows enough to become more sentient and develops stronger preferences, at some point its preference to continue existing will reach such a high level that (in almost all cases) it will outweigh any preference its parents might have to abort it. Nevertheless, it might still be possible for such extreme circumstances to arise that its preference to continue living could still be outweighed by the preferences of others. In fact, this could even occur after its birth; for instance, if the parents were forced to choose whether to save the life of their newborn on the one hand, or the life of their eight-year-old on the other hand (with whom they’d built an extremely close relationship and to whom they’d grown extremely attached), it’s possible to imagine a scenario in which the trauma of the parents losing their eight-year-old would be so much greater than the trauma of losing their newborn that it would outweigh whatever edge the newborn’s greater future life expectancy might have given it (despite its lower level of sentience) over its eight-year-old sibling.
Obviously this scenario is completely hypothetical, and it should go without saying that people’s preferences can differ dramatically, so the outcome described above wouldn’t necessarily be the right one for everyone who found themselves in such a dire situation. Thankfully, most of us have no reason to expect that we’ll ever have to make such a terrible decision ourselves, so for the most part it’s a moot point anyway. Still, based on people’s responses when these kinds of hypothetical scenarios are presented to them, it does seem that popular intuitions mostly correspond with the outcomes described above (which, I admit, came as a slight surprise to me when I first learned it). As Adam Benforado writes (summarizing research from Geoffrey P. Goodwin and Justin F. Landy):
Imagine that there was a ten-year-old girl [lying on the sidewalk next to a 63-year-old man] and that both she and [he] were in such critical condition that only one could be saved. If you were one of the [emergency responders], would you just flip a coin? Or would you choose to help one victim over the other?
Despite what we say about giving equal respect to all victims, most of us would save the girl. A ten-year-old is at the peak of her perceived value, and participants in experiments involving tragic tradeoffs tend to choose saving her over both older individuals (like [the 63-year-old]) and babies. Privileging the young seems to be grounded in an understanding that old people have had a chance to live a fuller life and that young people have more years left to contribute to the world. But clearly there is a limit: infants and toddlers have not had as much invested in them as their preteen counterparts, and their social relationships are not as significant, so people don’t perceive their deaths to be as costly.
Again, it’s worth stressing that these valuations are highly subjective, and that just because people find something intuitive doesn’t mean that their intuitions are well-founded. Our history is unfortunately replete with examples of people doing tremendous harm because their misguided moral intuitions seemed obvious to them. So it might not in fact be the case that saving a ten-year-old would have to be better than saving a toddler. (I’m not too sure about the answer here myself.) But that’s why it’s so important to figure out exactly how the logic underlying our moral decision-making works, and whether our intuitions actually correspond with what’s morally justifiable. Operating under the mistaken assumption that, say, sentience levels are all that matter, or that expected future utility levels are all that matter, is a recipe for trouble. We have to make sure that we’re correctly accounting for both – because based on all the factors we’ve considered here, it seems clear that both are absolutely fundamental.