Marginal Cases, On Trial

Author’s Note: I was unsure whether to post this piece on the forum, as it is more about the moral philosophy of an issue generally than Effective Altruism specifically. If I get reasonably good feedback on it, or at least reasonably good feedback on the compatibility of this type of post with the forum, then I may post more of my pieces with a broad philosophical focus here in the future.

A note on formatting, the footnotes that start with “Ed. Note” are commentary Nick added while editing. It occurred to me I should mention this as it isn’t obvious, I am not just randomly referring to myself in the third person in some of these notes.

One of the most common pieces of advice I’ve gotten on other posts to this forum is that I should break up my piece and include a brief tldr summary at the top. I’m going to try that for this piece, hopefully it will be helpful:

Part I: I describe what the argument from marginal cases is, as well as some of the rough shape of why it is unappealing to some people but to others, like me, worthwhile.

Part II: I describe Christine Korsgaard’s objection to the argument, roughly that it is too reductionist, and defend against her charge.

Part III: I describe Matthew Liao’s response to the argument, roughly suggesting a plausible criterion for human rights that meets its challenge, and point out why I find his response unsatisfying, before conceding some generalized reasons why arguments like his show a way in which the argument is lacking.

Part IV: I describe an objection that I don’t attribute to any specific person, roughly that the argument could be used to undermine human rights, and defend against this concern.

Part V: I describe some alternatives to the argument, and the motivation for emphasizing other arguments more like these, alongside the argument from marginal cases at least.

I.

I consider myself part of the animal welfare, or animal rights, or animal liberation, or animal justice movement^[1], whatever you want to call it, and have engaged in many arguments with people on this topic. Easily the most notorious argument on my team’s side, the “one to beat” philosophically, is the so called “argument from marginal cases”.

The very general form of this argument goes something like this:

-There is no attribute that necessarily applies to only and all human beings other than species membership itself.

-Mere species membership is morally irrelevant, therefore any plausible standard of ethics must treat some subset of the human population the way we treat non-humans, or a chunk of the non-human population how we treat humans.

-Out of these options, the better option is almost always to treat many non-humans how we treat humans.

The way this argument is usually structured in practice is to take some candidate justification for treating non-humans differently from humans, and choose an example of a human who, either because of age or disorder or some more hypothetical condition, fails to satisfy this criterion to a degree necessary to include them but exclude the animals being factory farmed by the tens of billions at any given moment. Something like,

“You think it’s alright to factory farm pigs because they aren’t smart enough? Interesting, so I take it you also believe we can factory farm infants then? No, they will one day become smart enough? Don’t worry, we kill them before their second birthday, they never will be that smart. No good because if it wasn’t for our mistreatment they would grow to become that smart? Fine, we’ll genetically engineer them before birth with a terminal disease that kills them before their second birthdays…” etc etc.

The argument functions as a bit of a game of whack-a-mole in practice, and perhaps understandably, it appears very silly to many people. Can we really settle the question of animal rights by discussing torturing terminally ill babies? I am, personally, a fan of the argument from marginal cases. I think that it basically holds up. That said, I understand people who look at it, and feel like the whole exercise looks rather silly and like a trick. For me and people more sympathetic to the argument, the silliness of the argument can be accounted for by the silliness of the types of arguments it is responding to.

It is silly to consider farming babies precisely because, when we take it seriously as a basis for morality, it turns out that the argument that intelligence should matter to moral patienthood seems quite silly and even ad hoc. This was made especially clear in some of the earliest influential statements of this argument, like Singer’s in “Animal Liberation”. He not only discusses marginal cases, he points out how crazy it would be if we determined duties like whether or not to harm someone for our minor benefit by looking to intelligence in any other context. Would racists actually be right about how we ought to treat people based on their race if they were right about differences in intelligence between different races? Or would we be much better off if we replaced racists with IQists who treated people with IQ 70-85 differently from 85-100 differently from 100-115? If this is wrong, is that really just because IQ isn’t a good enough measurement of intelligence? The stronger version of this argument made through the argument from marginal cases is just meant to cut off retreat to some threshold-based version of this same view.

At the end of the day, to marginal casists like myself, two things are central to our understanding of this argument. First, if we look at something that has all of the intrinsic features we usually use to define harm, this, and not some totally independent fact about the subject of the harm, should usually determine how much we care about the harm, and arguments to the contrary inherently have the burden of proof. Second, throughout the history of philosophy, our consensus of how we ought to treat humans has been developed with much more serious discussion and introspection than our consensus on how we ought to treat non-humans. If we must combine the two ethical projects into one coherent ethics, it is our earlier consensus on how we treat non-humans that it makes sense to throw out, not our consensus on how we treat other humans.

That said, the argument is not enough to convince many people, and there is something of a backlash against it that has existed ever since it started being used, and has only gotten more sophisticated over time. The argument sets up a puzzle to solve about how to minimize the overlap the argument draws on. If this overlap between humans and non-humans did not exist, and otherwise the questions were identical, the argument would not work at all. In my opinion, this would not really change that the arguments it deflates (like the role of intelligence) are bad standards to begin with, but that badness is merely a necessary condition for the argument from marginal cases to do its work, it is not a sufficient condition. Consequently there have been various attempts, of varying adequacy, to simply beat the challenge or dissolve it.

II.

One of the most interesting opponents of the argument is Christine Korsgaard, herself a hardcore animal rights activist, who has written in her book “Fellow Creatures” about why she thinks this argument simply misses the point.

“I have been suggesting that the argument from marginal cases wrongly treats a way of functioning as if it were just a property you could have or lack without other changes. In general, it treats a living thing as a heap of properties rather than a functional unity. The picture it suggests is this:

Properties of a dog: alive, sentient, emotional, intelligent.

Properties of a human being: alive, sentient, emotional, intelligent, rational.

Subtract rationality from the human being, and you are back to something essentially like the dog.” (Korsgaard, 84-85)

Roughly, to her, a dog is not simply a human but with some modification, the removal of some specific feature like level of intellect or moral agency or biographical sense of self or something. Humans and dogs simply have totally different ways of life, and conditions for flourishing, that make the comparison between a dog, and a human that has one or two differences from the average human, unworkable. At first glance, this is a quite compelling point. I think nearly all of this appeal, however, ultimately comes from its vagueness^[2].

If the difference in ways of life or flourishing are relevant only to how we can respect the interests of different beings, this is entirely consistent with the point of the argument from marginal cases. Whatever it means for things to go well for a being, defeating the argument from marginal cases means showing that we have much less reason to fulfill this for beings with some way of life versus others. A possible counter to this response is to say that the challenge here to the argument from marginal cases does not come from arguing that we should give less weight to different animals. Instead by highlighting the vastly different conditions for flourishing, it is a way of suggesting that perhaps we are already treating different species equally based on their very different conditions of flourishing, and so, properly understood, the argument is irrelevant. I would find this version of the argument more credible if it were remotely believable that the conditions on factory farms were consistent with any plausible account of the “flourishing” of the animals farmed there. It seems to me that if you want to defend the status quo, it is not on the table for you to say that we ought to treat different species with equal moral consideration.

I think that something like this interpretation is closest to what Korsgaard actually means, but I don’t think it does much to actually challenge the argument from marginal cases in practice, so I also want to explore some other close-by interpretations she may mean, or other people with similar objections might mean.

The other version of this objection not only looks at this bundle of features as ways of determining different types of flourishing, but also says that this different bundle of features or resultant basis for flourishing can determine how much your flourishing matters in itself. Either you could line these up along a narrow dimension in some way (like how much being part of a moral community matters to your flourishing), or else you could try to take seriously the idea that no single simple aspect of a being that could be modified determines or undermines moral importance.

The first type of interpretation lends itself to the usual marginal cases tactics quite easily. Take an otherwise perfectly ordinary person, and change the minimum about their psychology and physiology to make what causes them to flourish to be closer to a non-human animal. An example my classmate Isabella Ingrao brought up is the case of a feral child. A human who grew up in the wilderness, and flourishes in similar conditions to wild animals. It is possible to deny any such cases of feral children who would actually flourish best in the wild exist, but it seems unlikely to me that if anyone could be convinced that a feral child would flourish best in the wild, that would simultaneously convince them that it was, say, alright to hunt them.

The other possibility is that it is closeness to particular central categories of life that defines one as more “human” or “non-human” in our moral web, so that any being that has just one aspect of life or mind modified from the human average, no matter which aspect that is, should still count as morally human. This presents similar ambiguities on the margins, but pushes those margins out so that this ambiguity might never exist in practice. It exchanges this problem with a laundered version of the problem of speciesism however. If what justifies our treatment of beings is not any specific morally relevant feature, but the closeness of their cluster of features to the cluster of features the average human has, we must give some reason why this cluster around the average human being is the important one, and since we have discarded the idea of highlighting any specific feature of this cluster as what ultimately matters, like intelligence or agency, it looks an awful lot like this will just come down to mere undefended speciesism once again.

I am pretty unsympathetic to this style of argument, but I want to be clear that Korsgaard herself agrees that our treatment of animals today is inexcusable, but merely isn’t a fan of this particular argument against it. I also want to leave some room for the possibility that if I was more sympathetic to the types of Kantian, Aristotelian, and contractualist traditions that seem to inform the writing of Korsgaard and people with concerns like hers most, I would be less harsh, and find some more generous and nuanced interpretation of this style of argument. I am interested in better understanding reasoning of this style, but am relatively unmoved by this limitation on this issue in particular, just because I think the stakes of getting it wrong in their direction are so high.

III.

That said, perhaps an even more interesting and persuasive style of response is to take on the challenge, and to beat it. My professor, Matthew Liao, has done an impressive job of trying to do just that^[3]. Noticing that nearly all humans who fit the marginal cases categories, whether infants or the severely intellectually disabled, have similar genetics to average adult humans, perhaps with some minor difference, he has suggested that a criterion that meets the challenge more or less successfully is that we should prioritize beings not based on whether they are agents^[4], but whether they have the genetic basis for moral agency.

This bears some resemblance to other challenges of this sort. Korsgaard to some extent, but even more closely Shelley Kagan’s account of so called “modal personism”. While Kagan is left with a rather unpalatable bullet to bite (given a more specific interpretation of his account, he admits it could justify us in treating people with serious mental disorders differently if they resulted from genetics rather than trauma in the womb), Liao is left with a much smaller space of exceptions, one many more people might be willing to bite the bullet on. Even if a disorder has a genetic basis, so long as the person with this disorder still has the genetic basis for agency as well, they fit his criterion. Also impressively, his account was published years before either Korsgaard’s or Kagan’s.

To me, this answer is unpersuasive. It is true that, as a matter of fact, pretty much every real human has the “genetic basis for agency”, especially given a little flexibility in the definitions, but it might not have been that way. Imagine if when we were studying cells, we found that inside of cell nuclei, rather than DNA, there was homogeneous goo. This goo, like DNA, told us all of the details of our biological development, but unlike DNA, it was reducible to a single core trait – say that the very specific color of the goo determined how one developed. Plants perhaps have red goo, while animals have goo that is some shade of blue, while individual humans have yet more specific shades of blue distinguishing them from one another. In such a world, there are humans who have agency, and humans who, due to severe disabilities, lack this same level of moral agency. In this world, there is no shared “genetic basis of agency”, the basis of agency, like the basis of every other feature, is contained in the single specific shade of goo found in one’s nuclei. One has the basis of agency in this world, only if one has agency.

In such a world, it seems obvious to me, a society that went through the same moral developments and movements as us would view these disabled humans, who I am positing would not live subjectively different lives or behave differently from their counterparts in our world, as just as worthy of respect. This is not merely pointing to the marginal marginal cases for Liao’s account, like the possibility of incredibly rare, possibly currently non-existent humans who lack the genetic basis for agency, or genetically engineered humans like this in the future. I am making what I think is a stronger point. I am saying that in this goo nucleus world, goo Liao would not have come up with a comparable account that picked out physical components for potential agency as the morally relevant feature, and concluded that, in this world, moral agents all deserve more rights than non-agents. Goo Liao would I suspect instead have proposed some entirely different clever counterexample to the argument for marginal cases, and this contingency, I think, strongly undermines the persuasiveness of the principle itself.

Although I think this contingency argument undermines Liao’s point, I don’t think that it does as much to rescue the argument from marginal cases itself. Once you open the door to arguments like these, it is equally clear that the argument from marginal cases also only works contingently. We might have lived in a world in which no human was a “marginal case” in the relevant way, in which there was no overlap between humans and non-humans on things like intelligence or moral agency or anything else. In such a world, most of the same humans and same non-humans we are arguing about would exist, with the same features, but the removal of this area of overlap would suddenly defeat the logic of the argument. I might make another hypothetical appeal in such a world, like I do with the goo nuclei in this one, but without some cultural familiarity with the idea of real marginal cases, my appeals in such a world would look hopelessly academic. I suspect most people in this world would be agnostic about how they would treat such overlapping people if they existed.

The strength of the argument from marginal cases is that it is very general, it can be used to provide a strong dilemma for almost any candidate for the moral difference between humans and non-humans. The weakness is that it is not a positive project. It does not itself give us a reason to believe non-humans deserve moral regard independent of this specific dilemma, so it is beholden to the lucky coincidence that it can usually buy this moral regard through the process of elimination. This structural feature of the argument, its lack of an explicit positive point, has also contributed to perhaps its most obvious problem, and the one that I think causes the very most backlash against it.

IV.

Since the form the dilemma takes is making comparisons between some group of humans and animals who we treat with less regard than humans, there are two ways one could solve the dilemma consistently in theory. You could give many non-humans more regard, or you could give many humans less regard. More specifically I think some critics of the argument are especially concerned by the example of Peter Singer, who was both instrumental in popularizing this argument, which often compares non-human animals to infants and the severely disabled, and who has permissive views on infanticide that seem pretty ableist.

This example is probably the most striking for many people, that the most influential philosopher in the animal rights space is also very controversial on the very topics this argument is most worrying on, but I think more investigation shows that this example does not quite work. Most straightforwardly, Singer himself does have a positive, fairly specific, moral theory that he is working from, utilitarianism^[5]. The argument from marginal cases is consistent with utilitarianism, but ultimately his own views on animal rights and infanticide are rooted most directly in a specific, independent theory.

Furthermore, Singer’s views aren’t as extreme or repugnant as often implied. Utilitarian ethics has some inherent tensions with disability activism, but for his own part Singer’s controversial views seem to derive from a combination of a certain amount of empirical ableism from underrating the testimony of the disabled about their own lives, something he has come to regret in more recent years, and a non-deontological view of the ethics of euthanasia decisions in cases when the patient is unable to weight in.

It may be more worrying if there was a case like this from someone with flexible moral views. Someone who either subscribes to commonsense morality, or on a somewhat more formalized level, what William MacAskill calls “no-theory deontology”. However I’m not aware of anyone who seems to fit the bill here. Perhaps examples I have given of people like Kagan and Liao, who have bitten bullets after whittling down the scope of the marginal cases arguments as much as possible, are counterexamples, but I don’t think so. Not so much because I don’t think a simplistic version of their views would commit them to treating some humans with disabilities differently based on things like their genes, but more because I personally don’t buy that they would actually advocate acting on these implications in practice. It isn’t so much that I think they are disingenuous as that their views seem incomplete or uncertain from my readings.

The generalized reason I am not as worried about the argument from marginal cases on the human end is just the point I made earlier. The history of ethics has been largely about taking human value seriously, not of taking non-human value seriously. It would not merely be repugnant to treat humans more like we currently treat non-humans, it would ignore the asymmetry in where our ethics has and has not actually made substantial progress.

I want to end this particular discussion by noting what I think might be one of the more common counters vegans bring up against this type of objection, one Singer himself has. That is, there is some research showing that having higher regard for non-human justice is correlated with having higher regard for human justice. I bring this up, mostly to note that it seems to me pretty unhelpful.

The simplified logic of people who are concerned about the human side of the argument from marginal cases is that while some people will come away with the conclusion that they should treat non-humans better, others will come away with the impression that they should treat humans worse. This worry is not only consistent with, but actually predicts the result touted by people like Singer. The people strongly attached to the idea of human rights will be more likely to come away from this argument on the side of treating non-humans better, those less attached to human rights or more attached to our current treatment of non-humans will be more likely to bite the bullet in the other direction. The end result is a sorting in which one group is fairly unconcerned with either non-human or human rights, and the other is concerned with both.

What is needed to tell whether the argument has negatively impacted regard for human rights overall, is to see how sympathetic people are to human rights before and after exposure to the argument. This is much trickier. It is easier to crudely, at scale draw a line of association for vegans – measure the attitudes of people towards human rights before and after they go vegan – but this concern predicts that vegans will be the group with the least changed view of human rights. The real change is expected to occur among people who don’t become vegans after exposure to this argument, and there it is harder to find a way to study this at scale. Perhaps the best you can do is expose people to the argument for the experiment and then measure their attitudes, but it seems hard to do this in a way that will immediately produce interesting results either. I am unaware of any study that I would expect to settle this question empirically, so it seems to me that it is mostly a matter of crude speculation at this point.

V.

Looking back on these problems with the argument from marginal cases, I remain unconvinced that it is a bad or obsolete argument, but I think it is a tricky one. Without the right framing, it looks like a cheap or dangerous gotcha, and it is not a good basis on its own for any view of ethics. There are views in ethics that are almost exclusively popular because of dilemmas that favor them by process of elimination, including some which I favor more or less for just this reason as well, such as pure aggregation of interests, and totalist population axiology. Views of this sort are always somewhat fragile and disappointing. Stuck, for the most part, forever on the defensive. Animal rights/liberation/welfare, whatever you want to call it, is not like that. For those of us who care about this issue and buy its implications, we mostly do so viscerally, with personal feeling and motivation. It is something that seems actually, positively worth caring about.

The argument from marginal cases does not give this sense to people, at best it starts them on the road to arriving at their on their own version of this sense. This on its own seems to me like a good reason to find arguments that focus somewhat more on the actual positive moral weight of this issue, in particular thought experiments that focus on empathy, even if some of them are less structurally watertight that the argument from marginal cases. I want to close this piece by suggesting a few arguments of this sort that I am fond of for this reason.

The simplest one works mostly as a counterpoint to some commonsense, I think often uncritical intuitions about human values. Imagine a person living a very standard human life. They have dreams which they pursue and often achieve, they form rich relationships, have interesting and beautiful thoughts, contribute to community goods... and one day they get kidnapped. While kidnapped, they are tortured almost nonstop for days. During this time, they do not think anything sophisticated about the injustice of what is happening, they do not think about their life plans or relationships, they do not contextualize their relationships in their heads. They are too distracted. Their experience is simple, it is the experience of being tortured. It is unmitigated, long lasting, and absolutely horrible. Then they are released, and by some psychological peculiarity, they are not traumatized by this. Indeed, it does not impact their remaining life very much at all. Their plans and relationships continue unimpacted, and they die old and mostly satisfied. Perhaps their memories were even removed of this torture.

I believe the most morally significant thing that happened during this person’s life was being kidnapped and tortured. I furthermore believe that this experience of being tortured has very few morally meaningful differences from the types of harm non-humans can be, and often are, subjected to. A similar argument with somewhat different framing and emphasis has been made by Ozy Brennan.

Another, somewhat more elaborate argument is one that appeals to reincarnation, my own preferred framing for utilitarian ethics. Imagine you were told that, after you die, you will be reincarnated as every human alive, and every factory farmed animal alive, one after another, or if this is too abstract, that you will be reincarnated as someone from the pool of living humans and living factory farmed animals at random. How concerned are you about the experiences the humans have versus those of the factory farmed animals in this situation? Given that, though it in fact won’t be you in the real world, someone has to be every one of these beings, should the fact that this will not happen change how much you worry about each different experience in the real world?

And a final framing, perhaps one of the more abstract and difficult, but to me one of the most persuasive ones. Say that you go through life neglecting, or even contributing to the suffering of factory farmed animals. One day, you meet someone, who tells you that she used to be a battery cage hen. She is, understandably, not pleased with how she was treated before magically transforming into a conversant agent who could confront you about it. How would you justify yourself to her?

This, I think, is importantly different from a closely related case, in which a rock you once kicked around, and which suffered from this, transforms and confronts you. In such a case, you could honestly say that you didn’t think you were hurting the rock at all, because you didn’t think the rock could be hurt. If this rock person was reasonable, and you could convince the rock that your extremely low credence in a scenario like this was reasonable, then it seems as though this would be a perfectly adequate excuse. There is no parallel between this reason and what you might say to the humanized hen, unless you were mistaken about the fact that as a hen she was suffering in her conditions. Perhaps you could instead say that you had, quite reasonable, very very low credence that she would ever be in a position to confront you about this treatment. Do you think she would accept this answer? Do you think she should? What differs between this case and the real world, in terms of what is right or wrong in your behavior, if we agree that your lack of credence that she would transform would be reasonable, but not a good enough answer? It is generally accepted that one should be held as blameworthy or blameless based on their actual beliefs. If these lead you astray in some act, it is a forgivable accident. Given that you are in the same subjective position in this world as you are in the real world, in terms of your credence that you actually will be confronted by a humanized hen, then it seems as though if you have adequate justification in the real world, then there is also something you could give as an adequate justification to this hen. Working backwards, if you have no adequate excuse you can tell the hen, you have no adequate excuse in the real world either.

This question feels like a trick, but try out the possible counters for yourself. I wonder if you will find anything satisfactory that should make it irrelevant. I cannot. The fact that you did not believe you would be confronted about this harm simply doesn’t seem like a good excuse, and I think the humanized hen would be well within her rights to ignore any argument that appealed to this, rather than the actual harm you knew you were causing to a subject of a life, a being there was something it was like to be. A sentient creature.

A final point you could make is to say that in fact you do think the hen is like a rock, and you could say the same things in this situation you would say to the rock. Doing this, I think, requires not merely the belief that the hen does not have this harmful experience, but a negligible credence that she does. If you are in the second to top floor of an apartment building you are 60% sure is empty, and fire a machine gun into your ceiling for fun, the fact that you didn’t believe there was probably someone above you is no excuse when someone gets shot. Likewise the hen would have quite a reasonable complaint with you if you said that you merely had a 30% credence that she was being seriously harmed. The question is comparable, in that just as we are asking whether there is someone in the apartment above you in one case, we are asking if there is someone inside of this chicken in my thought experiment. You should feel the fear of god in you as you act on the assumption that there is not, in each case, and getting to a negligible credence of this is swimming against the philosophical current to a worrying degree.

These arguments are far from watertight. They can all be countered by someone willing to say that the harms and differences in each case come only from the fact that they occurred to a being that at some point did or would have more complex thoughts and experiences. I think this argument is unsatisfying in each case, but there is no gotcha for denying this tactic in the way the argument from marginal cases excels at gotcha-ing. The strength arguments like this have is that they give you something positive, empathy, and with it a sense of why it is worth caring beyond the process of elimination.

An added bonus is that the argument from marginal cases is often taken to directly entail the “equal consideration of equal interests” conclusion for animal rights. I think this is the correct conclusion, and I think the arguments I have given, and others in their vicinity, are best interpreted in this way as well, but it is a very controversial conclusion that by its nature draws a target on the argument’s back. This is not necessary.

I recently read about the way many snakes are killed for their skins,

“Larger captured snakes are often first starved to loosen their skin and then stretched by being forcibly pumped with water. Snakes are routinely nailed to a tree and skinned alive, their bodies thrown on to heaps where they can take two days to die.”^[6]

This is one of the most monstrous things I have ever heard of in this space. There is an unwashable moral stain on the history of any world where this happens to even one conscious being. Snakes are not treated like this in huge numbers compared to other animals. Chickens are often boiled alive during the defeathering process after a botched slaughter through electrocution and having their throats slit. The characteristic fishy flavor and short shelf life of fish is not a feature of the animal, but due to the lactic acid their bodies are flush with in their final minutes of life from the stress as they struggle to escape, and are suffocated and crushed under other fish. Both of these happen on an astronomically bigger scale.

This is just a taste of some of the most dramatic moments in the agonizing, preventable lives of the animals we have outnumbered ourselves with for our use. I would like to stop speciesism full stop, but the truth is I don’t need equal consideration of equal interests. I need a world where this stuff doesn’t happen. All I am asking for is to burn our current animal agriculture system to the fucking ground and salt the earth behind it. I will happily stand shoulder to shoulder with anyone who I can get to agree with that modest goal.

I think the argument from marginal cases is still worth using, but I would be even more interested, at this point, in people developing more arguments and thought experiments along these lines. Ideally tighter and more unassailable ones, but barring that, just ones that will help people feel what it is like to sincerely care about this issue.