1. Introduction
I'll argue for two theses. One is that NAS is a live possibility, where NAS says that the bodies of typical adult humans contain non-agential subjects. A non-agential subject (1) exists within an organism and (2) has phenomenally conscious experiences in a morally significant way, but (3) is not morally responsible for (some or all of) the organism's voluntary actions. I'll clarify NAS later, but for now, an example. Call the human organism sitting in this chair “Organism.” Call me “Author.” Perhaps Author is Organism, or perhaps Author is a soul, or cerebrum, or whatever; that doesn't matter here. I, Author, have conscious experiences in a morally significant way, and I'm responsible for what Organism does: either I am Organism, or I somehow control it. So I fulfill conditions (1) and (2) but not (3). I'm not a non-agential subject. But suppose panpsychism is true. Consider Ellie the Electron, who is conscious in a morally significant way and can be found somewhere in my brain. Ellie is within Organism, but isn't responsible for Organism's behavior. She doesn't know about Organism's actions or surroundings, isn't reasons-responsive, can't control what Organism does, etc. So Ellie is a non-agential subject. I won't argue that NAS is likely to be true. I'll just argue that it's worth taking seriously. This is because there are multiple ways it could turn out to be true and the disjunction of these possibilities is worth taking seriously.
The second thesis I defend is that NAS might have very important ethical implications. Many of these are in areas where agency is often thought relevant: the fact that I voluntarily do something means I am liable to, immune from, deserving of, obligated to do, etc. something else. For example, people often think my making an imprudent or immoral decision is a reason to let something bad happen to me. If NAS is true, the bad thing may directly affect, not only me, but many other innocent moral patients. Allowing the bad thing might then be harder to justify. Other implications of NAS don't involve agency. For example, if NAS is true, some organisms probably contain more subjects within themselves than others. If this is right, we might have reason to favor the interests of those organisms, since more subjects might thereby benefit. And if we're even just unsure whether NAS is true, that might have important implications for what we subjectively ought to do, given the information available to us. For example, perhaps we ought to avoid punishing people too harshly just in case we're also harming innocent non-agential subjects.
If the two theses are true, there's an important practical upshot. Ethicists should try to determine whether NAS is true, and they should try to determine exactly what the implications of NAS are. (Some might suggest the implications of NAS are so counterintuitive or repugnant that they constitute good reason to reject NAS, along with any views which imply it. We should also try to determine whether this is true.) My aim here is to show that this project is worthwhile. In section 2, I clarify NAS. In section 3, I lay out some non-goofy scenarios in which NAS is true, and argue that NAS is true in them. In section 4, I describe an assumption, Correlated Well-Being, which I'll use in evaluating the ethical implications of NAS, and present a few reasons one might accept it. Correlated Well-Being says there's a correlation between which events are good or bad for us and which are good or bad for the non-agential subjects with whom we share our bodies. If this is false, the ethical implications of NAS might be quite different, and I'll gesture at the ways in which this might be so. In section 5, I explore some potential ethical implications of NAS.
2. NAS
NAS is the proposition that the bodies of typical adult humans contain non-agential subjects, where a non-agential subject (1) exists within an organism and (2) has phenomenally conscious experiences in a morally significant way, but (3) is not morally responsible for (some or all of) the organism's voluntary actions. Some notes. First, understand talk about subjects existing “within” organisms loosely. If I just am Organism, still say I'm “within” Organism. If I'm a non-spatial soul causally connected to Organism, also say I'm “within” Organism. Etc.
Second, non-agential subjects have phenomenally conscious experiences “in a morally significant way.” I say this partly to exclude putative cases where an individual's experiences are not morally significant because the individual is only derivatively conscious. Suppose organisms have experiences. Suppose I'm not Organism, but an “embodied mind” (McMahan Reference McMahan2002, ch. 1). I have experiences. So two individuals have experiences. But maybe there aren't two morally important subjects here because Organism has experiences only in a derivative and morally unimportant sense, namely, by having a conscious part (McMahan Reference McMahan2002, 88–93). Then Organism wouldn't be a non-agential subject. (Compare: Organism digests food; Organism's digestive system digests food; Organism is not its digestive system; yet there aren't two things digesting food in the same sense.) None of the putative non-agential subjects I discuss are derivatively conscious in that way. Nonetheless, it may be possible to argue that the subjects in sections 3.1–3.3 are having numerically identical experiences, or that some are having experiences which are “nothing over and above” the experiences of others.Footnote 1 If subjects are mere “value receptacles” (cf. Yetter-Chappell Reference Yetter-Chappell2015), unimportant in themselves but significant only because of the valuable experiences they hold, one might hold that it's morally as if there's a single subject in these cases. (Thinking otherwise would be like thinking your gold is more valuable if you put it in a nesting doll, so that it's held by many different containers at once.) These questions are important, but for reasons of space, I'll mostly follow certain other authors (e.g., Johnston Reference Johnston2017; Simon Reference Simon2017) in assuming the different subjects are non-redundantly morally significant. Nothing like this worry arises for sections 3.4 or 3.5. Whether the subjects are conscious “in a morally significant way” may also depend on their cognitive capacities and which theory of well-being we accept; I'll discuss this in section 4.
Third, we should distinguish between full non-agential subjects, which aren't responsible for anything an organism does, and partial non-agential subjects, which are responsible for some and not other things. Usually, by “non-agential subjects” I'll mean full non-agential subjects. But in section 3.3 I'll discuss some partial non-agential subjects.
Fourth, some organisms (animals, babies, etc.) include non-agential subjects because they're conscious but contain no moral agents at all. But NAS is about typical adult humans, and it is very unpopular to believe that these organisms contain non-agential subjects. Notice, though, that the motivations for accepting NAS may imply that animals, etc. contain multiple non-agential subjects. That's also an unpopular view. And it would also have important ethical implications, for reasons discussed in section 5.4.
Fifth, it's conceptually possible that our bodies contain multiple conscious individuals, but no non-agential subjects. Suppose there are multiple subjects of experience within Organism, but they all are morally responsible for what Organism does. Then none would be non-agential subjects. I'll mention why we should think the relevant subjects are non-agential subjects in section 3.
3. Ways NAS might be true
Here, I lay out five ways NAS might be true. I claim these show that NAS is a serious proposal. There are also other ways NAS might be true. For example: I suggested above that perhaps I'm conscious and Organism is conscious, but Organism is only derivatively conscious, and derivative consciousness is not morally significant. But suppose derivative consciousness is morally significant (Stratton-Lake Reference Stratton-Lake2021), or that Organism is non-derivatively conscious. If only I am morally responsible – say, because I control Organism – Organism is a non-agential subject. But I focus on just five ways for reasons of space.Footnote 2
3.1 Panpsychism
Panpsychism says the fundamental physical entities are conscious (cf. Bruntrup and Jaskolla Reference Bruntrup and Jaskolla2017). If that's so, our constituent particles are conscious. Like Ellie the Electron, they are too ignorant, impotent, and stupid to be responsible for our actions. So we contain very many non-agential subjects. Of course, if panpsychism is true, everything contains conscious subjects. Whether panpsychism has particular implications for how we should treat human beings (and other sentient animals) depends on whether there is any interesting connection between how things go for macro-level experiencers and how things go for their constituents. I discuss this in section 4.
3.2 Spatial Overlap
The intrinsicality of consciousness may imply NAS. (I'll rely on Mørch (Reference Mørch2019), though other philosophers have explored similar themes.) Here's the rough idea. An intrinsic property “does not constitutively depend (though it may well causally depend) on properties of other things, or its external surroundings”; an extrinsic property “does constitutively depend on properties of other things, or its external surroundings” (Mørch Reference Mørch2019, 134). My laptop's having a certain mass is an intrinsic property; its existing simultaneously with Paris is an extrinsic one. Consciousness seems intrinsic: intuitively, it constituently depends on how I am on my own, not on how other things are. Obviously, my conscious states causally depend on my environment: raising the temperature changes my experience. However, it does so by causing a change in me. As Hedda Hassel Mørch (Reference Mørch2019, 134) notes, “Intuitively, one cannot change whether someone is conscious merely by changing the properties of other things (assuming these changes do not causally affect the conscious area of their brain).”
Whatever property it is in virtue of which my brain produces conscious experiences, it would still possess that property if Ellie the Electron were missing. But then it looks like my-brain-minus-Ellie should itself be conscious.Footnote 3 By the intrinsicality of consciousness, whether it's conscious depends on it, not on something extrinsic to it, like whether Ellie exists. And of course, there is nothing special about Ellie: the same applies for every other electron. So the bodies of conscious creatures contain many subjects.Footnote 4
More precisely: Mørch (Reference Mørch2019, 135) suggests that most theories of consciousness are committed to rejecting one of the following three theses:
INTRINSICALITY: Consciousness is intrinsic.
NON-OVERLAP: Conscious systems do not overlap with other conscious systems …
REDUCTIONISM: Consciousness is constituted by more fundamental properties (as per standard versions of physicalism and Russellian monism).
Rejecting INTRINSICALITY probably means accepting some sort of exclusion or maximality claim, according to which whether a system is conscious can depend upon whether it's part of a larger system. However, this might not even help. On some understandings (Sider Reference Sider2003), to say that consciousness is subject to a maximality condition is to say that something is conscious iff (i) it possesses some intrinsic property, which we can call consciousness*, and (ii) it's not a proper part of some larger thing which possesses consciousness*. But it seems plausible that consciousness*, which is intrinsically identical to consciousness, is just as morally relevant as it. My brain-minus-Ellie might then be, not a non-agential subject (because not conscious), but morally equivalent to a non-agential subject (because possessing the morally equivalent intrinsic component of consciousness).
Another option is to reject REDUCTIONISM. Reductionism does not here just mean physicalism: it also applies to views, like Russellian monism, according to which macrophysical consciousness is constituted by more fundamental properties. REDUCTIONISM can be avoided by adopting a view according to which our consciousness causally, rather than constitutively, depends upon these more fundamental properties (Mørch Reference Mørch2019, 156–59). Perhaps consciousness is an irreducible property of Cartesian souls, with our brains merely causing our souls to have various experiences. Then my brain would be conscious if Ellie the electron were removed, since it isn't conscious to begin with. So we get the result that there is just one conscious experience associated with my brain by positing that there is just one soul causally related to it. Peter Unger (Reference Unger2006, ch. 7) has actually provided something like this as an argument for substance dualism. However, many philosophers are committed to REDUCTIONISM.
Finally, we might reject NON-OVERLAP and grant that we contain many overlapping subjects of experience. But for two reasons, it's plausible that many of these will be non-agential subjects. First, some may be composed of just the parts of my brain responsible for conscious experience, and not other parts necessary for decision-making, etc. Second, there will be very many, largely overlapping subjects who are about equally good candidates for being responsible for my decisions. Since they're all roughly equally good candidates for being in control of my actions, perhaps we should hold that either all of these subjects are responsible for my actions or none of them are. Or perhaps, since all of their decisions are linked – if any one does something, all of them do it – at most one of them is free (cf. Hudson Reference Hudson2001, 39–44). If none of them, or just one of them, are responsible for my actions, then all or all but one of them are non-agential subjects. I'm not arguing for any specific solution to Mørch's trilemma; rather, I'm noting that there are credible motivations for adopting a solution which probably implies NAS.
3.3 Temporal overlap
Mark Johnston (Reference Johnston2017; cf. also Zimmerman Reference Zimmerman, Loux and Zimmerman2003; Olson Reference Olson2010; Taylor Reference Taylor2013; Kaiserman Reference Kaiserman2019; Longenecker Reference Longenecker2021) discusses the ethical implications of “personites,” “shorter-lived very person-like things extending across part, but not the whole, of a person's life” (Johnston Reference Johnston2017, 617). Suppose perdurantism is true and I have temporal parts spread throughout spacetime: perhaps I'm the maximal fusion of all the parts that stand in a certain kind of psychological or bodily continuity with one another. Suppose mereological universalism is true.Footnote 5 Then there are also shorter-lived fusions consisting of just some of the parts of my life: for example, the parts between my fifth and seventh birthdays. Further, if time is continuous, there are continuum many fusions beginning at, say, some point or another between my fifth birthday and the day after my fifth birthday, and ending on my seventh birthday. These are all personites. They are not persons only for extrinsic reasons: the segment between my fifth and seventh birthdays would itself be a person, except that my life extends before and after it, rendering it non-maximal. Further, the assumption of perdurantism may be unnecessary; Johnston argues that personites also exist:
on any view that treats us as cross-time sums of temporal parts, and on any view that treats us as entia successiva, i.e. as successions of shorter-lived entities … More surprisingly, there will be personites on any view that treats personal identity as consisting in the holding of relations of bodily and/or psychological continuity … there will also be personites on any view which allows that there are a plurality of available essences for self-conscious beings of our spatial extent and make-up to have. (2017, 617–18)
For space, I'll focus on the ordinary perdurantist version.
Since personites intrinsically possess what's necessary for personhood, Johnston thinks their moral status equals that of persons. And many personites will be non-agential subjects. There is the question, mentioned before, of how many of us can be free, given the way our actions are counterfactually linked. But even setting that aside, some may exist when I'm not performing any actions, and will therefore be full non-agential subjects. Others will be only partial non-agential subjects, responsible for some of my actions and not others, because they're only around for some of them. A personite beginning at my nineteenth birthday isn't responsible for anything I did before then. Johnston recognizes the radical ethical implications: he writes that “Our central practices of prudential reasoning, binding ourselves through promising, consenting, recognizing desert, fairly distributing surplus goods, rewarding and punishing, are all hard to reconcile with the existence of personites” (2017, 618). I'll pick up some of these themes in section 5.
Johnston raises another issue (2017, 635–41) that I won't pick up. He thinks personites undermine “the compelling idea of trying to make things better overall – of aiming to maximize interest-satisfaction” (2017, 618). Here's the worry. Suppose time is continuous. Then there are continuous many personites overlapping me at any moment. So if you benefit anyone, you have benefited infinitely many personites, doing an infinite amount of good. Doing more beneficent things won't, given the resources of standard transfinite arithmetic, increase the amount of good you've done. So we've all already done as much good as we can. The same point applies to doing harm. So we've all already done as much evil as we can. And positive infinity minus negative infinity is (in standard transfinite arithmetic) undefined, so maybe there's not even any fact about how much good anyone has done. We thus face “infinitarian paralysis” (Bostrom Reference Bostrom2011): none of our actions affects how well things go overall.
Suppose instead that time is discrete: between any two points in time there are only finitely many points. Then there are only finitely many personites overlapping me, assuming my life is finitely long. But this raises two issues. First, presumably there is some finitely large chance that time is continuous. So our actions will still be infinitely valuable or disvaluable in expectation, and the same problems arise for calculating the expected values of our actions. Second, longer lives will contain more temporal points, and so more personites (since there are more places for personites to begin and end). Accordingly, it looks like we have reason to discriminate in favor of individuals with longer lives. But that's “repellent, perhaps even absurd” (2017, 641) because it seems to require that we discriminate in favor of those who are already (ceteris paribus) better-off, and more or less because of the thing that makes them better-off: for instance, that we discriminate in favor of the rich, since they have greater life expectancies.
I'll discuss the discrimination issue later, in section 5.4. But I'll set aside the infinitarian paralysis worry and assume it makes sense to care about making the world better, trying to favor the interests of more rather than fewer individuals, etc. There are two reasons. First, we might question whether Johnston is aggregating interests in the right way (cf. Longenecker Reference Longenecker2021). Second, I don't see that personites contribute much to infinitarian paralysis. I think there's already some finitely large chance that the universe contains infinite value and disvalue (say, because it's infinitely large). And I think there's already some finitely large chance that any of our actions causes something infinitely good or bad to happen (say, by affecting whether someone goes to Heaven or Hell). So I think we need some way of incorporating infinite values into decision theory without leading to paralysis anyway. And I think there is one: it probably involves using surreal numbers (Chen and Rubio Reference Chen and Rubio2020).
3.4 Neuronal consciousness
The scientists Steven Sevush (Reference Sevush2006; 2016) and Jonathan C. W. Edwards (Reference Edwards2005, Reference Edwards2006) have defended the “single-neuron theory of consciousness,” which says phenomenal consciousness is “a property of individual neurons whose outputs were coordinated by natural selection to give the appearance of a single macroscopic mind” (Sevush Reference Sevush2016, xii). Each member of some population of neurons – perhaps located within the pre-frontal cortex – is having more or less qualitatively identical experiences, and there is no larger conscious subject composed of multiple neurons. Rather, it seems that there is because our reports, behavior, etc. harmonize with the experiences of every individual neuron. Due to space, I won't go deeply into the motivations for the view, but it's put forward primarily as a solution to the “binding problem,” the problem of “explaining how a spatially dispersed physical substrate could coalesce into a single experiential flow” (Sevush Reference Sevush2016, 209). On the single-neuron theory, NAS is true: there are very many conscious subjects in a human organism, and, for reasons explained before, it seems plausible that at most one of them is responsible for the organism's behavior.
3.5 Hemispheric consciousness
Split-brain patients have had the corpus callosum, which joins the hemispheres of the brain, surgically severed. They exhibit odd behaviors which seem to suggest that the patient now possesses two distinct centers of consciousness, one in each hemisphere. Suppose we ask the patient to retrieve an object. The word “keyring” flashes in such a way that only “key” appears to the left eye, and only “ring” appears before the right eye. The left side of the body is associated with the right hemisphere, while the right side of the body is associated with the left hemisphere. The left hemisphere controls verbal behavior in most people. The patient may then report that they see only the word “ring,” even as their left hand reaches for a key (Bayne Reference Bayne2010, 190–91). Sometimes, the two hemispheres appear to struggle with one another in attempts to complete what each understands to be the task requested, or aid one another in a phenomenon known as “cross-cuing.” In one case, a patient's left eye was shown a picture of Hitler, and he made a “thumbs-down” with his left hand. However – his left eye being associated with his right hemisphere – he couldn't explain why. He realized he'd seen Hitler after his left hand began tracing letters on his right hand (Puccetti Reference Puccetti1981, 96–97).
In these cases, maybe only the “verbal” hemisphere is conscious, with the other operating as a kind of philosophical zombie. But this view faces problems. First, certain split-brain patients produce speech with both hemispheres (Bayne Reference Bayne2010, 192). And second, as the above cases illustrate, the right hemisphere appears capable of complex reasoning and intentional action, so it's unclear what the grounds for denying it consciousness are (Bayne Reference Bayne2010, 193). Another model, defended by Tim Bayne (Reference Bayne2010, ch. 9), is that both hemispheres are capable of supporting consciousness, but that only one hemisphere is conscious at any given time, with consciousness quickly switching between hemispheres during experiments like those described. However, “The dominant view within both neuropsychology and philosophy” is that “there are times at which the split-brain patient has phenomenally disunified conscious states,” with “the received account” arguably being that split-brain patients possess two completely separate streams of consciousness (Bayne Reference Bayne2010, 193).
I'll assume that, if there are two different streams of consciousness, there are two different subjects.Footnote 6 In that case, there are two possibilities. One is that the second subject exists only in split-brain patients, coming into being with the surgery. The other is that a second subject exists in normal humans, too, and that this only becomes manifest in split-brain patients because autonomous action in the non-verbal hemisphere is typically “suppressed by the unconscious inhibitory influence of the speech hemisphere, acting through the intact corpus callosum” (Puccetti Reference Puccetti1989, 144). The first is often alleged to be more intuitive (Puccetti Reference Puccetti1981, 93), but it's also odd if severing the corpus callosum creates a new conscious subject, and some authors argue that positing two minds in ordinary subjects provides the best explanation of the data overall (e.g., Puccetti Reference Puccetti1981; Puccetti Reference Puccetti1989; French Reference French1989). I won't adjudicate; my point, as before, is just that the relevant view isn't crazy.
If there really are two subjects in intact brains, how should we think about their agency? A natural model is one on which the verbal hemisphere “calls the shots,” with the other able to dissent only under special conditions allowed by the split-brain procedure. Evan Williams (Reference Williams2015, 976, fn. 13) speculates that perhaps:
the function of the corpus callosum in a human brain is not to unite the two hemispheres in the production of a single consciousness, but rather to allow a dominant consciousness situated in one hemisphere to issue orders to, and receive information from, a subordinate consciousness situated in the other hemisphere.
If something like this is correct, it seems right to think of the non-dominant hemisphere as ordinarily being a non-agential subject.
4. Correlated well-being?
Before exploring the ethical consequences of NAS, I'll make an assumption called Correlated Well-Being. Correlated Well-Being says that there's at least a rough positive correlation between what's good or bad for us and what's good or bad for our associated non-agential subjects. So, I'll assume that if continued life is good for us, it tends to be good for these other subjects, and vice versa; that if health is good for us, it tends to be good for these other subjects; that if incarceration is bad for us, etc. There are a few reasons one might think this, with the exact reasons depending on which of the scenarios laid out above is correct. They fit most naturally with the popular view that pain is intrinsically bad for us, and pleasure intrinsically good for us – a view accepted by hedonists, as well as by many objective list theorists. (The story for desire theorists will be somewhat more complicated, for reasons I explain below.)
The neuronal consciousness scenario makes the case for Correlated Well-Being easy, since the relevant neurons are all having experiences which are identical to mine. The spatial overlap scenario keeps it easy: at least very many of the non-agential subjects who overlap me will have conscious experiences more or less qualitatively identical to my own, since the brain states associated with conscious experience which they possess will be identical, or nearly identical, to those associated with me. One wrinkle is that the number and identity of non-agential subjects will constantly change as matter enters and leaves my brain. So if I prudently sacrifice some of my well-being now for more later, this will on balance harm some of the non-agential subjects who are part of me now but won't exist later. But this doesn't undermine the idea of a correlation, since many or most of the non-agential subjects currently in me might benefit, and there will be other, new ones who benefit when I flourish later. A similar point can be made about the temporal scenario (their experiences are identical to mine at any given time, and they are constantly coming and going), with the added wrinkle that, if time is continuous, an infinite number of personites begin and end in any interval of time. I already said what I have to say about that. In the hemispheric consciousness scenario, the experiences of the two halves of my brain may not be identical, but it seems prima facie reasonable that they're similar enough for Correlated Well-Being to hold. (But see the discussion of Evan Williams’ suggestion below.)
The case for Correlated Well-Being is hardest in the panpsychist scenario. But consider that many panpsychists are so-called “constitutive panpsychists”: they believe that our experiences are in some way constituted by the experiences of the fundamental physical entities. One important motivation for this is that it allows us to avoid, on the one hand, the claim that conscious experiences are physically reducible, while also avoiding, on the other hand, the claim that conscious experiences are “strongly emergent,” arising with certain macrophysical configurations but without being constituted by anything more fundamental. (The view avoids strong emergence because the experiences which arise with macrophysical configurations, like ours, are grounded in something more fundamental – the experiences of microphysical objects.) The existence of strong emergence is thought implausible by many, and it may also lead to problems with mental causation. (For discussions of these issues, see Chalmers Reference Chalmers, Bruntrup and Jaskolla2017 and Bruntrup Reference Bruntrup, Bruntrup and Jaskolla2017.) On one view (perhaps popular among Russellian monists), the experiences of the microphysical particles remain constant, with differences in macro-level experiences resulting purely from changing configurations of micro-particles. However, it's hard to see how this could work: a phenomenally red square and a phenomenally blue one, say, don't seem to differ so much in their structural properties as in whatever “building block” experiences might constitute them. On another possible view, particles’ experiences change with their relational properties, and this accounts for the non-structural differences in our experiences. In that case, we might expect there to be some kind of qualitative similarity between our experiences and at least some of theirs. For instance, when I'm in pain, we might expect that at least some of them will be in at least some vaguely pain-like state. If this is right, we might have grounds for expecting some kind of rough correlation between my experiences and theirs.
There are three ways in which Correlated Well-Being could be false. First, there could be no interesting relationship between what's good or bad for us and what's good or bad for the non-agential subjects within us. Maybe, while things can be good or bad for them, there's no interesting relationship between those and what's good or bad for us. Or maybe things aren't good or bad for our non-agential subjects at all. Maybe they have conscious experiences, but they're all purely neutral. Or maybe they have pleasures and pains, but pleasures and pains are not intrinsically bad for us. People who hold desire satisfaction theories of well-being think that what's intrinsically good for us is desire satisfaction, and what's intrinsically bad is desire frustration. Pleasure and pain do tend to be good and bad for us, but this is because we generally desire the experience of pleasure and want to avoid the experience of pain. At least in certain scenarios (e.g., the panpsychist one), the non-agential subjects within us may not be sophisticated enough to have desires, so that pleasures and pains are not good or bad for them.
If there is no interesting correlation between what's good or bad for us and what's good or bad for the non-agential subjects we contain, the ethical implications of NAS will be lessened. If punishing someone doesn't affect the associated non-agential subjects, that removes an important reason for not inflicting the punishment. However, it might still be that NAS would have some important ethical implications. For example, perhaps it's prima facie wrong to destroy an innocent conscious individual for some deontological reason, apart from whether or not continued existence would be good for that individual. Suppose that killing someone would destroy some non-agential subjects. Then this would provide a reason not to kill them, even if the person in question is morally liable to be killed (as punishment, in self-defense, etc.). So even on the no-correlation view, NAS might have some important ethical consequences.
There is a second way of rejecting Correlated Well-Being. It might be that there is a negative correlation between what's good or bad for us and what's good or bad for the non-agential subjects associated with us. It's not clear why this would be the case, but if it was, the implications might be extremely troubling. If their aggregate condition is bad enough and cannot be ameliorated, perhaps we'd have obligations to commit suicide, or at least to not reproduce. If members of other species similarly contain tortured non-agential subjects, we might have an obligation to take them down with us. In any event, this would be important, if depressing, information.
Finally, it may be that, while there are important relationships between our well-being and that of our associated non-agential subjects, whether the correlation is positive or negative depends on the issue. For instance, suppose that non-agential subjects tend to be miserable. Evan Williams (Reference Williams2015, 976, fn. 13) suggests that if our non-verbal hemispheres really are non-agential subjects, subservient to the whims of the other hemisphere, then “The worst-case scenario … is that there are 300 million human slaves in America, whose frustration, boredom, and oppression we have not made any effort to ameliorate.” In that case, perhaps their lives are worse than death, so that our continuing on is bad for them. However, suppose that, given that we exist, it is better for them that we be better off than worse off: they avoid pain when we do, and so on. This would have interesting consequences. For instance, the existence of non-agential subjects might provide reason against inflicting most unpleasant punishments (since those make the non-agential subjects worse off) but for inflicting the death penalty (since that puts them out of their misery). Or consider activities, such as drinking alcohol or participating in extreme sports, which carry a risk of damaging brain cells and the connections between them. These might give me pleasure. But on the neuronal consciousness scenario, perhaps they risk degrading the future experiences of some of my non-agential subjects. Or in the spatial overlap scenario, they might cause some to cease existing (e.g., one that's differentiated from another only by including material in a synapse that's destroyed). In any event, if this possibility is right, the implications for ethical issues will have to be determined on a case-by-case basis: they will depend on exactly which factors bear which relationships to each other. But as these examples show, this would still have important implications for ethics.
5. Implications
In this section, I discuss four potential implications of NAS, given Correlated Well-Being. The first three concern issues to which responsibility has been thought relevant, while the last does not. There are other potential implications: we already saw Johnston's claim that “Our central practices of prudential reasoning, binding ourselves through promising, consenting, recognizing desert, fairly distributing surplus goods, rewarding and punishing, are all hard to reconcile with the existence of personites” (2017, 618). These are just some major implications to help illustrate NAS’ significance.
5.1 Paternalism
Paternalism is interfering with someone's freedom for their own good. Opponents of paternalism think we can interfere with agents’ actions in order to protect others, but deny that we can do so to protect (competent) agents themselves. However, given NAS and Correlated Well-Being, many supposedly paternalistic policies will also be supported by other-regarding considerations – namely, concern for the agent's non-agential subjects. If I ride a motorcycle without a helmet, I endanger, not only myself, but also the non-agential subjects I've brought along for the ride. A law banning this might be, in a way, more similar to laws banning reckless driving or driving without buckling in my child's car seat. This would not be an argument against the anti-paternalist view in principle: it would have no effect on whether preventing harm to an agent justifies interfering with them. But it might make anti-paternalism practically unimportant, because supposedly paternalistic policies would instead be supported by other-regarding considerations.
5.2 Criminal justice
Virtually everyone agrees that, under normal conditions, only criminals should be punished by the law. NAS may make it impossible for us to punish criminals without also harming their non-agential subjects. It would be almost as if every criminal had a hostage at hand whom we could spare only by letting the criminal go. Exactly what this means in practical terms depends on which theory of punishment we accept.
The deterrence theory states that criminal punishment is justified because it deters people from doing bad things, and so prevents more harm than it causes (e.g., Ellis Reference Ellis2003). The importance of punishing only the morally responsible is instrumental on this view: deterrence is (usually) not served by punishing innocent or insane people. Straightforward forms of the deterrence theory will likely sanction the same punishments that they would have anyway, even if NAS is true. It will be regrettable that we must harm non-agential subjects in order to punish a criminal. However, deterring criminal acts will be even more important than it otherwise would have been, since they harm, not only the people we ordinarily think they harm, but also the victims’ non-agential subjects. It seems likely that, often, the two effects will be roughly proportional to one another (but see section 5.4), so that the presence of non-agential subjects will not change what we should do.
However, many people reject the deterrence theory. They worry, say, that in certain circumstances it has difficulty accounting for the importance of punishing the guilty, rather than an innocent scapegoat (Pereboom Reference Pereboom2014, 164). So there are various alternatives: for example, retributivism, the view that punishment is justified by the need to make criminals pay for their offenses as a matter of justice (e.g., White Reference White2011), and the expressive theory of punishment, which asserts that punishment serves to express our condemnation of the criminal's bad acts (e.g., Feinberg Reference Feinberg1965). But NAS and Correlated Well-Being call into question these theories. If punishment also directly harms other, innocent subjects, it seems doubtful that it's a good way to express our disapproval, or balance the scales of justice.
So those who reject the deterrence theory might instead be pushed into what Derk Pereboom (Reference Pereboom2014, ch. 7) calls the “quarantine” theory of criminal justice. As a hard determinist, Pereboom must answer the question of how we should handle criminals, given that they're not culpable for their actions. His answer is that, just as we might forcibly quarantine an innocent person with a deadly disease to prevent their harming others, we might forcibly incarcerate dangerous criminals to prevent their harming others. The criminals would not exactly be punished any more than infected individuals who must quarantine are punished: any unpleasantness is the unintended side-effect of defensive measures. Rather, such a system would focus on rehabilitation and humane treatment. Of course, the justification for the quarantine theory provided by NAS would be quite different from that provided by hard determinism: the claim would be, not that no one is culpable, but rather that we cannot punish the culpable without also harming the innocent.
5.3 Distributive justice
I'll now discuss two relations between NAS and questions of distributive justice and interpersonal fairness. Here's one. Many people think responsibility is relevant to distributive justice. Economic or other inequalities between two people seem easier to justify if they result from one's person being more prudent than if they result from their getting a bigger inheritance. While this view is often associated with the political right, many left-wing philosophers have also attempted to honor judgments like this. So-called “luck egalitarians,” such as G. A. Cohen (Reference Cohen2008), make the elimination of inequalities caused by factors beyond anyone's control the centerpiece of their theory. Meanwhile, Ronald Dworkin (Reference Dworkin2011) suggests that balancing equal treatment with respect for personal responsibility is the central problem for a theory of distributive justice.
Even some philosophers who reject this role for responsibility often employ it elsewhere in their theory. Rawls (Reference Rawls1999, 64) rejects merit-based justifications for inequality partly because of the role chance plays in whether we make “responsible” decisions: “Even the willingness to make an effort, to try, and so to be deserving in the ordinary sense is itself dependent upon happy family and social circumstances.” However, Rawls measures how “well-off” one is by considering one's possession of primary goods, “things that every rational man is presumed to want” (1999, 54), such as rights or money, rather than considering how much utility one gets from these goods. This might seem unfair to those with “expensive tastes,” who get less utility from the same goods. Rawls’ response is that those with expensive tastes can be appropriately held responsible for their tastes, and so are not entitled to compensation:
The reply is that as moral persons citizens have some part in forming and cultivating their final ends and preferences. It is not by itself an objection to the use of primary goods that it does not accommodate those with expensive tastes. One must argue in addition that it is unreasonable, if not unjust, to hold such persons responsible for their preferences and to require them to make out as best they can … The use of primary goods, however, relies on a capacity to assume responsibility for our ends. (Rawls Reference Rawls, Sen and Williams1982, 168–69)
So Rawls, too, relies on judgments about responsibility. And of course, even utilitarians will, for purely instrumental reasons, sometimes want to reward responsible behavior so as to incentivize it.
The same point made about responsibility in earlier sections can be applied here. Purely instrumental, incentive-based justifications will probably wind up having about the same force, proportionally, as they otherwise would have. But any reasons to assign responsibility non-instrumental significance will, to some extent, be counterbalanced by other reasons. Allowing someone to suffer due to imprudent decisions will impose suffering on non-agential subjects who are not responsible for those decisions. It would be like allowing children to live in poverty because of the poor choices of their parents. On the other hand, rewarding someone for making prudent decisions will also reward non-agential subjects who played no role in making the decision. It represents the kind of undeserved good fortune manifest when a child of wealthy parents receives special opportunities. Our inability to implement policies which are not over-inclusive with regard to responsibility – letting those who haven't been irresponsible suffer, and rewarding those who have not been responsible – will weaken whatever overall force the case for assigning responsibility a non-instrumental role in our distributive policies had.
The second implication arises merely from considering non-agential subjects as among those to whom we are distributing goods, quite apart from worries about responsibility. Apparently equitable distributions may be inequitable when we consider their impact on non-agential subjects. Suppose that, during a pandemic, locking down will cause a short period of suffering to the whole population, whereas refusing to lock down will cause the same amount of suffering over a longer period but concentrated among vulnerable populations. It may seem that locking down is more equitable, since it spreads the burden among a larger group. However, on the temporal overlap scenario, locking down also concentrates suffering among personites who exist now (but not later) while sparing those who exist later (but not now). This may provide a distributive reason of some force against locking down. Similar considerations may apply to some individual actions. The exact relationship between societal-level distributive justice and individual action is controversial: am I obligated to pursue an equitable distribution myself, or merely to uphold and play by the rule of institutions which pursue such a distribution (cf. Cohen Reference Cohen2001)? But it seems clear that distributive considerations of some sort play some role in individual action. For instance, if I'm buying a set of books for the family, I should, ceteris paribus, buy a set both children will enjoy somewhat rather than a set only one child will enjoy a lot. But suppose what's really happening is that the left hemispheres of both children will enjoy the books, while their right hemispheres, having only meager verbal abilities, will be bored. It might be fairer to forego the books altogether and try to find something all our hemispheres can enjoy.
5.4 Differential treatment
Many people believe that speciesism is wrong – that it's wrong to favor one individual's interests over those of another just because of their species (e.g., Singer Reference Singer2002). We may nonetheless have reasons to usually favor actions which benefit individuals of certain species over those which benefit others, not because their equally weighty interests are more important, but because the members of certain species usually have stronger interests at stake. So maybe humans usually have a stronger interest in continued life than a dog, so that we should usually save a human's life over a dog's. Or maybe a dog usually feels pain more intensely than a salamander, so that we should usually prevent pain to a dog rather than a salamander. This is compatible with the general anti-speciesist principle that we should not discriminate according to species in cases which violate these trends – for instance, with the judgment that if I can prevent boredom in a human baby or worse boredom in an adult dog, I should favor the dog.
Some authors call into question the extent to which such reasoning really does support favoring humans over other animals, especially when it comes to preventing pain. Sahar Akhtar argues that pain may actually tend to be worse for non-human animals than for humans, both because “the intensity of pain for us can often be mitigated through expectations, memories, and the consideration of and attention to other interests” and because “we are able to engage in inter-temporal calculations with respect to our interests and frequently discount pain in order to achieve other, higher order or longer term interests” (Akhtar Reference Akhtar, Beauchamp and Frey2011, 496). Richard Dawkins (Chiorando Reference Chiorando2017) suggests that, because pain serves a biological role in teaching animals to avoid harmful behavior, it may be more intense in non-human animals: “An animal that is a slow learner or less intelligent might need more intense pain … than a human who is intelligent enough to learn quickly.”
NAS may provide a different reason for usually favoring humans over other animals, and tending more broadly to favor more cognitively sophisticated creatures over less sophisticated ones, or longer-lived ones over shorter-lived ones. If NAS is true, creatures with larger and more complex nervous systems may generally contain more subjects who could be affected by our actions. For instance, if panpsychism is true and how our actions affect the subjective experiences of particles depends on whether the particle is bound up in a consciousness-producing part of a brain, then animals that are configured such that these parts of their brains are larger will contain more non-agential subjects whom we could affect. Similarly, if each brain actually contains many slightly overlapping subjects, then larger brains will contain more. In the discrete personite scenario, longer-lived creatures contain more personites. If individual neurons are conscious, individuals with more of the relevant neurons contain more subjects. If individual hemispheres are conscious, then creatures with two brain hemispheres may contain an additional subject. In these cases, there might be a reason to prefer actions which benefit humans, not because human interests are inherently more important and not because they are stronger, but because benefiting a human will actually benefit more conscious individuals overall.
However, these considerations are unlikely to always tilt in the convenient pro-human direction. Some animals have more massive brains or nervous systems than humans, and similar reasoning may imply that they have more non-agential subjects. Elephant brains are about three times as massive as human brains, with about three times as many neurons (the cerebral cortex, specifically, is more massive but contains fewer neurons) (Herculano-Houzel et al., Reference Herculano-Houzel2014). Or consider octopi. While they possess centralized brains, most of their neurons are actually located in their arms, and much of the behavior of their arms is controlled locally by distributed nerve clusters known as ganglia. If we accept the hemispheric consciousness scenario, perhaps the same considerations which motivate it should also lead us to think that these ganglia, or certain sets of them, are themselves conscious. And of course some animals, such as certain tortoises, live longer than humans, and therefore contain more personites – or the non-person animal equivalent – at least on the discrete time scenario. Of course, none of this immediately implies that I should prevent elephant or octopus or tortoise pain over human pain, even granting the rejection of speciesism. Perhaps apparently similar instances are worse for each subject in a human, and ultimately worse overall, despite there being fewer subjects in a human. The point is just that the presence of the additional non-agential subjects makes the case for preventing the elephant or octopus or tortoise pain, as compared to the human pain, stronger than it would otherwise be.
Similar reasoning might also lead us to favor certain individuals over other members of the same species. We already saw Johnston's worry about discrimination between humans based on life expectancy. Relevant to the spatial overlap scenario, Simon (Reference Simon2017) notes that if every candidate experiencer in our “experiencer-clouds” (“a maximal set of all candidates for being an experiencer that significantly overlap one another” (451)) is an actual experiencer, then certain medical conditions will lead to certain experiencer-clouds containing more experiencers than others. (Simon suggests a hyperactive hippocampal subgranular zone – the part of your brain that oversees creating new neurons.) If Jane has such a condition and John doesn't, then Simon argues that we'd have more reason to alleviate Jane's pain than John's (section 1). He thinks this would violate a plausible moral principle he calls “Equal Treatment,” which states that “Fully distinct (person-shaped) experiencer-clouds deserve equal moral treatment, other things equal” (452).
Both Johnston and Simon view these consequences as unacceptable. Are they right? I doubt John is being denied “equal moral treatment” in the sense we should care about. We are counting John and Jane's interests equally; it's just that there are additional subjects at stake in Jane's case. So perhaps this is not a denial of “equal moral treatment” any more than rescuing the life raft with more people on it would be. But Johnston's worry was specifically about the thought that we should benefit individuals who are already better off than others, and more or less for the reasons that make them better-off. There is no particular reason to think that Jane is better-off than John. But if cognitive sophistication tends to make one better-off in general and is associated with containing more non-agential subjects, then the rationale under discussion might have a similar implication in some other cases.
One immediate thing to note is that maximizing welfare already requires that we benefit the better-off in some cases. As I write, there have been many discussions about how to distribute coronavirus vaccines. Many worry that, for some time, wealthy nations will monopolize limited vaccine supplies. But the life expectancy of an at-risk person in a wealthy country who gets the vaccine is probably higher than that of someone in a poorer country. Perhaps the person in the wealthy nation has a higher quality of life, too. So it seems, prima facie, that the better prospects of the person in the wealthier nation mean that we should give them even better prospects, since that will produce the most good.
Many of us would find that “repellent, perhaps even absurd.” There are various ways to reject the conclusion, but the obvious one involves fairness or equality: benefiting the better-off person seems ceteris paribus less important, since they're better-off. We might then have welfare-maximizing reasons to help the person in the wealthy nation, and fairness-based reasons to help the person in the poorer nation. That many of us would find it “repellent” to help the person in the wealthy nation on the very grounds that, since they're better-off, they have more to lose from death, suggests that we think fairness-based reasons are quite weighty here. Perhaps we can say something similar about non-agential subjects: while we may do more good by helping the additional non-agential subjects found in, say, a longer-lived person, this is usually balanced by fairness-based reasons to help those who are worse-off. Admittedly, this will be a blunt instrument: the life of a well-off person may contain many badly-off personites, and they may suffer under such a policy. But we're doing the best we can.
However, there are two points to make. First, the reasoning discussed in this section may not always support favoring the better-off individual. It seems plausible to think that elephants, octopi, and tortoises tend to be overall worse-off than humans, even if they contain more non-agential subjects. Perhaps we'd then have two reasons to favor their equally-weighty interests over those of humans: welfare-maximization and fairness. Or at least, this might be the case unless we accept the (speciesist?) idea that equality and fairness do not apply to animals in the way they apply to humans (cf. Kagan Reference Kagan2019).
Second, in the discrete time personite scenario, the personites in a life increase geometrically, rather than linearly, with time. If I've lived x of the smallest temporal units, adding another unit to my life increases the number of personites by x + 1.Footnote 7 A similar point holds for the spatial overlap scenario: adding a particle to a larger brain means there are more sets of particles for it to fuse with. In at least some cases, this means the considerations of fairness or equality will need to be much more important than they would otherwise need to be. Suppose we can perform a quality-of-life enhancing surgery on Baby A or Baby B. Both will experience an equal quality of life improvement, but Baby A, who is better-off anyway, will live to 81 whereas Baby B will live only to 80. Here, we might think favoring Baby B means sacrificing a comparatively small marginal amount of good (improved quality of life for an extra year) in the name of fairness. But on the personite scenario, it may mean sacrificing a tremendous amount of additional good, by depriving a tremendous, disproportionately large number of additional personites of the benefit. However, similar implications don't seem to hold for the other scenarios. Accordingly, the exact implications here depend heavily on how NAS obtains.Footnote 8
