Monothematica

Killing Animals Thesis

Killing Animals Revisions

Death and Animals

Introduction

In this essay I will examine how different views on the badness of death should inform our beliefs about killing animals. I’ll do this in two main parts. In part one, I’ll first explain three general accounts of the badness of death. I will then consider their relative merits, as well as their relationships with broader views about identity, welfare, and population ethics. All three of these accounts have various advantages and disadvantages. The aim of part one is not to establish one account as uniquely correct, but to clarify the best versions of each account, and the reasons different people might have to favor one over the others.

Having done that, in part two, I will apply these three accounts to animal deaths, and see what each one implies about reasons for and against killing animals. First, I will apply each account to simpler cases of killing animals without replacing them with new animals. Then I will consider what each account (in conjunction with different views of welfare and population ethics) implies about killing animals with replacement. I will then conclude by summarizing some of the practical takeaways of this essay.

However, I will first make some clarifications about the nature and goals of this essay.

Clarifications

In asking about the value of death, I am asking how death can be good or bad for the individual that dies. By ‘death’, I don’t mean the process of dying, nor any pain that accompanies it, but the event of death itself. In focusing on death’s badness for the being that dies, I’m not addressing the indirect harms death causes.

When considering the wrongness (and rightness) of killing and letting die, I will only consider aspects of this wrongness that derive from the harmfulness of death for the individual that dies. On the assumption that this harm doesn’t vary between cases of killing and letting die, I will sometimes say ‘killing’ or ‘letting die’, when I really mean killing or letting die.

Though I focus on this aspect of the wrongness of killing, I don’t think considerations of harm are the only considerations relevant to the wrongness of killing animals1. However, I think that these harm-based considerations should be practically relevant regardless of one’s general ethical views, while rights or virtue-based considerations may not be, so I’ve decided to focus on them here. Due to limited space, I will assume that, while many things that harm humans may not harm animals, if something does harm a non-human animal, we have a pro-tanto reason not to do it.

Finally, I won’t discuss the implications of Epicureanism about death in this paper. Epicureanism is the view that death is never bad nor good for the individual that dies, (Luper, 2021). One reason I’m not discussing Epicureanism is its general implausibility and unpopularity among contemporary philosophers. But the main reason is that its implications about animal death are entirely straightforward. No matter what view of welfare or population ethics one holds, if one is an Epicurean about death then animal deaths are never good or bad for them.

Why does this matter?

The question of whether death is bad for animals is practically important for two reasons.

First, it may determine whether ‘animal-friendly’ systems of farming are acceptable. In these systems, animals have net-positive lives, and are killed painlessly or near-painlessly. If death isn’t bad for the relevant animals, then such systems may be neutral or morally good. But if death is bad for animals, then these systems may still be bad in one of two ways. If animals’ happy lives can’t offset the harms of their deaths, such systems may be inevitably bad. If their happy lives can offset the harms of their deaths, then animals may still need to live significantly better than neutral lives for such systems to be morally neutral.

Second, if death is worse for some animals than others, then our moral priorities should shift in a more fine grained way. If pig-pain is typically worse or more important than fish-pain, we should value preventing pig suffering proportionally more. In the same way, if death is typically worse for pigs than fish, we should value preventing pig deaths proportionally more.

On most of the views I will discuss, both these considerations will be partly determined by facts about animal psychology. Due to limited space, (and my own ignorance), the best I can offer in this essay will be statements of the form; Insofar as an animal has XYZ psychological property, death will be better/worse for it. In order to generate specific, actionable claims about the relative importance of pig vs. chicken vs. fish deaths, the arguments I make here will have to be supplemented by further research into the psychology, neuroscience, and development of different species.

1 - Death

Three contemporary accounts

There are two basic reasons one might think death is bad. The first is death’s ability to frustrate the desires of the one who dies. The second is its ability to deprive us of benefits we would have otherwise accrued. These two reasons generate two basic kinds of views on the badness of death.

The first kind is a Desire-Frustration Account of death’s badness, (Belshaw, 2008, 2012, 2015). These accounts claim that death must frustrate certain desires of an individual in order to be bad. These accounts diverge based on (a) what kinds of desires are relevant, and (b) whether their frustration is necessary and sufficient, or merely necessary, for death to be bad.

The second kind are deprivation views of death’s badness, which focus on death’s ability to deprive us of future benefits. The simplest kind of deprivation view is a Life-Comparative Account of death’s badness, (Bradley, 2009, 2015; Nagel, 1970). On this account, all that matters in evaluating the value of death is whether it raises or lowers that individual’s lifetime well-being.

A more complex kind of deprivation account is Jeff McMahan’s Time-Relative Interest Account, (McMahan, 1998, 2002, 2019a). On this account, the value of death is determined by two factors. As in the Life-Comparative Account, the first factor is the deprivation of future benefits and harms. But the second factor is the strength of the psychological connections that would have held between the individual at the time of death and the times when the benefits or harms would have accrued to them. The weaker these connections, the less this deprivation contributes to the value of death. This focus on individual psychology is a similarity shared with Desire-Frustration Accounts.

All these accounts have different advantages and disadvantages; none is vastly better than the others. I’m most sympathetic to the Time-Relative Interest Account, but your preferred view will depend as much on your broader philosophical views as on the arguments discussed here. The purpose of part one is to clarify and assess these advantages and disadvantages, as well as highlighting the relationships between each account and views of welfare, identity, and population ethics. Once I’ve done this I will consider the implications of each of these accounts for animal deaths in part two.

1.1 - The Life-Comparative Account

Explication of the Life-Comparative Account

The Life-Comparative Account (LCA) states that the value of any event, (including death) for an individual is just the difference it makes to the amount of welfare their life contains, (Bradley, 2009). Unlike the other two accounts, the psychological properties of an individual aren’t directly relevant to the badness of their death. Instead, the relevant fact is that the individual who dies would have been identical with the individual that would have accrued the benefits or harms.

In this respect, the value of death is much more binary than on the other two accounts. On both the Desire-Frustration (DFA) and Time-Relative Interest Accounts (TRIA), the value of death varies continuously with the strength of certain psychological properties of the being. In the case of the DFA it’s the strength of the desires for continued life, and in the case of the TRIA it’s the strength of the psychological connections between the individual at the time of death and the times when they would have accrued benefits or harms. But on the LCA the value of death is determined by comparing the value of two possible lives, the actual life of the dead individual, and the life they would have lived, had they not died.

Evaluation of the Life-Comparative Account

The most obvious downside to the LCA is its implication that, if your life would have gone well, then the worst possible time to die is immediately after beginning to exist, (Bradley, 2009). If we begin to exist as fetuses, (or even as newborns) this implies that the deaths of fetuses (who would have lived good lives), are the worst of all deaths. If we begin to exist as fetuses, this implies the abortion massively harms fetuses. Of course it’s possible that, despite the harm it causes, abortion may not be wrong. Maybe the pro-tanto wrongness of the harm is trumped, or maybe we have no duties not to cause (certain) harms to fetuses. Nevertheless, the LCA naturally lends support to some anti-abortion positions.

A related, and I think more serious downside, is the implication that death becomes bad very suddenly, (McMahan, 2002). Recall that personal identity with your possible future self is the relation the LCA uses to determine the value of deaths. If we hold the commonsense view that this relation is both transitive and doesn’t come in degrees, this implies that beings must come into existence instantaneously. But this means that, whether we start to exist as fetuses or newborns, the deaths of fetuses/newborns become extremely harmful2 at some unknown instant in time. Aside from this being unintuitive in itself, if this harm is a morally-important one, then this generates the further counterintuitive conclusion that discovering when we do in fact begin to exist may be an overwhelmingly important moral priority, since it would help us distinguish the potentially-tragic possible death of a newborn/fetus at t+1 from it’s harmless death at t-1.

A final weakness, which I’ll discuss in section 2.1, is the extra-unintuitive implications of combining the LCA with animalist views about personal identity. The implications of the DFA and TRIA do not vary based on one’s views of personal identity. As such, animalists about personal identity may prefer the DFA and TRIA over the LCA.

Despite these weaknesses, the LCA does have three main advantages. The first is its relative simplicity. Unlike the DFA and TRIA, the LCA doesn’t need to posit special kinds of desires or interests to explain the badness of death, it just appeals to the welfare a being would have obtained across its life.

The second advantage over the DFA is the fact that the LCA, (along with the TRIA) counts the loss of presently-undesired future goods as a harm. I find it intuitive that part of what makes the death of a child or teenager bad is the fact they never got to attain the goods they would have grown to desire in later life. Even if a teenager at the time of death had no desire to be a parent later in life, if it’s still the case that they would have grown to desire this, had they lived, then this seems bad for them. Similarly, the LCA, (and TRIA) can more easily accommodate the intuition that, if the years of life an individual lost out on would in fact have been deeply harmful, then their death wasn’t as significant a harm to them, even if they strongly desired to live. In the next section I’ll argue the DFA can be modified to accommodate this intuition, but only by sacrificing some theoretical simplicity.

Finally, the LCA (and TRIA) are compatible with hedonist theories of welfare, while the DFA is not. As such, hedonists will generally prefer the LCA or TRIA. Additionally, while the DFA and TRIA in general are compatible with any view about population ethics, I’ll argue the most plausible versions of each strongly favor non-person-affecting views in population ethics. But the LCA doesn’t favor any particular view about population ethics. As such, people who hold person-affecting views will be more attracted to the LCA.

1.2 - Desire-Frustration Accounts

Explication of two Desire-Frustration Accounts

Desire-Frustration Accounts are a family of views that claim death is only bad for an individual if it frustrates their desires to continue living. Variants of the view differ on (a) which kind of desires are the relevant ones, and (b) whether desire-frustration is necessary and sufficient, or merely necessary, for death to be bad.

I will first explain Christopher Belshaw’s version of the DFA, (Belshaw, 2012; 2015). Belshaw’s DFA claims that (a) future-directed categorical desires are the relevant kind of desire, and (b) their frustration is necessary, but not sufficient, for death to be bad.

Categorical desires are desires which are not merely canceled, but either satisfied or frustrated by death, (Belshaw, 2012, pg. 29, McMahan, 2002, Bradley and McDaniel, 2013). Assuming beings have reasons to avoid the frustration of their desires, beings therefore have (pro-tanto) reasons to keep living if death would frustrate their categorical desires.

Categorical desires are meant to contrast with conditional desires. Conditional desires are desires for something, (e.g. visiting Venice), that we only hold conditional on some other state of affairs obtaining (e.g. Venice not being flooded), (Belshaw, 2012). When the condition fails to obtain, the conditional desire is neither frustrated nor satisfied, but canceled. Thus, for Belshaw, a categorical desire is any desire an individual has that is not conditional on it continuing to live, (Belshaw, 2015)3.

To motivate the view that these desires are necessary for death to be bad, Belshaw considers the example of a suicidally depressed patient whose only desire is the desire to receive painkillers tomorrow, (Belshaw, 2015). The fact that they won’t receive painkillers tomorrow if they die today gives them no reason to keep living. Belshaw claims this is because their desire for painkillers is conditional on their being alive. This means the desire would be canceled, not frustrated, by death. Since desire-cancelation is not bad for an individual, the patient’s death would not be bad for them.

Belshaw claims categorical-desire-frustration is necessary, but not sufficient, for death to be bad. Instead, (roughly speaking4) a death is bad iff it frustrates an individual’s categorical desires, and the life they lost would have been good for them, (Belshaw, 2012). Belshaw adds this clause to avoid the conclusion that the death of a patient in incurable chronic pain, who nevertheless has an extreme, baseless phobia of death, would be bad for them.

A contrasting version of the DFA could instead claim that (non-misinformed, presently-held) intrinsic desires are the relevant kind of desire, and their frustration is both necessary and sufficient to make death bad. For ease of reference I’ll call this the Singer-DFA5. I’ll outline this account by showing how it can derive the same conclusions as Belshaw’s DFA in the two cases already mentioned.

Regarding the depressed patient’s desire for painkillers, the Singer-DFA can appeal to the fact that the desire for painkillers is extrinsic, not intrinsic. Extrinsic desires are desires that p, conditional on p raising the probability that some distinct state of affairs q obtains, (Bradley and McDaniel, 2013). The desire to scratch an itch is an example of an extrinsic desire; ‘p’ is ‘I scratch my nose‘ and ‘q’ is ‘My nose doesn’t itch’. The depressed patient’s desire can be analyzed in the same way; p is ‘I get painkillers’ and q is ‘I’m not in pain’6.

Whether an extrinsic desire is satisfied or canceled is irrelevant to the wellbeing of an individual. To take the example of the extrinsic desire to scratch your nose, if your nose stops itching of its own accord, (i.e. q obtains) the probability of q is 1, so p can’t raise the probability that q obtains. This means the condition is falsified and the desire is canceled. In the same way, death makes it the case that the depressed patient is not in pain, thereby canceling his extrinsic desire. Since the cancellation of extrinsic desires isn’t bad for an individual, neither is the depressed patient’s death.

The Singer-DFA can also explain why a baseless phobia of death is insufficient to make death bad without mentioning the quality of the life lost. It can do this by claiming only the frustration of non-misinformed intrinsic desires makes death bad. So long as the phobia can be dismissed as a misinformed desire, (perhaps it’s based on the belief that hell exists, when it actually doesn’t), the Singer-DFA can preserve the verdict that such a death would not be bad for them. More generally, the Singer-DFA can claim that the value of death for an individual is proportional to the balance of satisfaction and frustration of their non-misinformed intrinsic desires. Presuming the man also had non-misinformed intrinsic desires not to be in pain, the Singer-DFA can claim his death was good for him.

Evaluation of the two Desire-Frustration Accounts

I will first outline the shared merits of the two DFAs, then move on to the advantages the Singer-DFA has over Belshaw’s version.

Regarding merits, both DFAs avoid the counterintuitive implication of the LCA that, the earlier in life death occurs, the worse it is. Instead, on Belshaw’s DFA, fetuses and infants up to a certain age lack categorical desires, so death isn’t bad for them, (Belshaw, 2012). And on the Singer-DFA, intrinsic desires for future goods gradually strengthen and extend as we age, so the deaths of the very young aren’t very bad for them. Secondly, unlike the relation of identity with your future self, both kinds of desire emerge gradually as we age, so death doesn’t suddenly become extremely harmful, like the LCA implies.

Despite these merits, any version of the DFA will face two disadvantages.

Firstly, DFAs are incompatible with pure hedonism as an account of welfare, since any DFA implies that desire frustration is necessary (and maybe sufficient) for death to be bad. The LCA and TRIA don’t have this disadvantage.

Secondly, any version of the DFA will struggle to account for the role that the loss of (presently-undesired) future benefits and harms seems to play in explaining the badness of death, (McMahan, 1998). This is still the case given desire-satisfaction theories of welfare, where these future benefits and harms are just the satisfaction and frustration of desires the individual doesn’t yet hold. I’ll return to the question of how each version can try to account for this shortly, but first I’ll discuss the distinct disadvantages of the Belshaw-DFA.

Recall that Belshaw’s DFA claims the cancellation of conditional desires by death is insufficient for death to be bad. But this seems odd when we consider conditional desires in general, (Bradley and McDaniel, 2013). To return to the example of a desire to visit Venice, conditional on it not being flooded, if Venice floods, this still seems bad for the individual with that desire. Even if this cancellation isn’t intrinsically bad, like desire frustration is, losing the opportunity to satisfy the desire still seems bad in a comparative sense. To make this clear, suppose the individual could control whether Venice floods. It seems that they would have a reason to ensure Venice doesn’t flood, and thereby ensure their conditional desire isn’t canceled. But if (intrinsic) conditional desires can give reasons like this in other cases, why shouldn’t they give the same reasons in cases of desires conditional on continued life?

The second disadvantage of Belshaw’s DFA is its claim that desire-frustration isn’t sufficient to make death bad, since the life lost must also have been good overall, (Belshaw, 2008). Any such version of the DFA has a number of downsides. First, it seems unparsimonious to think the badness of death has two distinct sources. Second, it’s not clear how to flesh this out into an account of when (if ever) death is good for someone, (e.g. a suicidal person who would in fact recover and have a good life). Third, it’s not clear how to aggregate these two aspects to decide how bad a given death is.

There’s three ways to perform this aggregation, but none seem very promising. One could claim that, (so long as death frustrates some of the relevant desires), then it’s badness is directly proportional to all the lost welfare, (plus the badness of the frustrated desires themselves)7. But this just retains the disadvantages of the LCA, without the advantage of being compatible with hedonism. Alternatively, one could claim that only some subset of the future goods contribute to the badness of death. But any way of deciding the members of this subset seems ad-hoc. Finally, one could count all future goods as contributing to the badness of death, but discount their contribution somehow. But this is just a less parsimonious version of the TRIA, minus the compatibility with hedonism.

Given this, the most plausible version of the DFA is the Singer-DFA, which claims the value of death for an individual is proportional to the balance of satisfaction and frustration of the non-misinformed intrinsic desires held at the time of death.

It’s true that such an account won’t be able to claim that the loss of presently-undesired future goods can make a death worse, but it’s worth paying this cost in order to avoid the problems outlined in the last two paragraphs.

1.3 - The Time-Relative-Interest Account

Explication of the Time-Relative Interest Account

The third account of the badness of death is Jeff McMahan’s Time-Relative Interest Account, (TRIA).

For McMahan, the badness of death is largely composed of two factors, (McMahan, 2002). Firstly there’s the amount of welfare that death deprives the individual of. Secondly, there’s the strength of the psychological unity relations that would have obtained between the individual at the time of death and itself in the future, had it lived. McMahan summarizes this as follows; The degree of psychological unity within a life between times t1 and t2 is a function of the proportion of the mental life that is sustained over that period, the richness or density of that mental life, and the degree of internal reference among the various earlier and later mental states, (McMahan, 2002, pg. 75).

These two factors combine to determine an individual’s time-relative interests. An individual’s present time-relative interests are what it currently has egoistic reason to care about, (or what a third-party would currently have reason to care about for that being’s own sake), (McMahan, 2002, pg. 80). What is in an individual’s time-relative interest, (henceforth TRI) at t1 is whatever is in their interest at t1, plus would be in their (non-time-relative) interest at the later times t2, t3,… tn, with the interests of each future time discounted by the strength of prudential unity that would obtain between them at t1 and themselves at each later times, (McMahan, 2002, pg. 75-80). Talk of an individual’s (non-time-relative) interests at a given time just refers to whatever would be intrinsically good or bad for them at that time.

The TRIA can be illuminated by particular cases:

  1. A woman dies at 40. Had she not died at 40, she would have lived happily until dying at 80.

Her death is almost as bad for her as the deprivation of those 40 years of good life, since there is a high degree between herself at age 40 and herself at all later times.

  1. Instead of dying at 80, she would have developed dementia at age 80, lived happily until 90, then died.

Even though she loses 10 more years of good life, this would not have been much worse than the case 1. This is because the badness of frustrating the interests she would have held between age 80 and 90 is heavily discounted due to the lack of psychological unity between her at age 40 and ages 80+.

  1. A newborn dies. Had it not, it would have lived happily to age 80.

Despite death depriving the newborn of a huge amount of welfare, this death is the least harmful of the three. This is because there would have been very little psychological unity between the newborn and themselves at later times.

Evaluation of the Time-Relative Interest Account

I will begin by outlining general advantages of the TRIA over the LCA and DFA, then consider some challenges and refine the TRIA in light of them.

One advantage of the TRIA over the LCA and DFA is its fit with our intuitions. Unlike the LCA, the TRIA doesn’t always count earlier deaths as greater harms. Instead, the TRIA implies death in a continuously good life is worst somewhere between late childhood and early adulthood8. The TRIA also avoids the LCA’s implication that death becomes extremely bad instantly or near-instantly. Instead, dying during a good life becomes gradually worse, peaks between 10 and 25, then gradually declines.

Unlike the Belshaw-DFA, the TRIA implies the deaths of human infants and non-human animals are still bad for them, even if they aren’t typically as harmful as the deaths of adult humans, due to their weaker psychological unity. Unlike the Singer-DFA, the TRIA lets the loss of undesired future goods contribute to the badness of death. This implies the deaths of human infants are more harmful than the deaths of psychologically-similar dogs due to the greater magnitude of (presently-undesired) future goods the human infant loses in dying.

In addition, the TRIA is compatible with either hedonism or desire-based theories of wellbeing, since the interests of an individual at a time are just whatever would be intrinsically good for them, at that time.

Despite these advantages, the TRIA still faces a number of challenges.

The first challenge is the Problem of Overdetermination, (McMahan, 2002, pg. 120-4). Consider the following case:

Overdetermined Loss: A seemingly healthy young man has a sudden stroke, and is left irreversibly and severely disabled. A week later, he dies of a heart attack caused by a hidden congenital heart defect.

The basic problem is that, if we consider the deprivation of good life caused by either the stroke or the heart attack on their own, presuming the other event would still (or had already) occurred, we radically underestimate the harm the man suffered. This is a general issue for deprivationist accounts of harm, so the LCA faces the same problem, and should adopt the same solution.

The solution is to recognise that the loss of good life caused by the stroke and the heart attack is not the sum of the losses caused by each event on its own, (McMahan, 2002). Instead, we must evaluate the overall loss caused by conjunction of the two events. It is this overall loss that’s practically relevant. If we can prevent both the stroke and the heart attack, we should not dismiss the value of the option just because the value of preventing either event individually is quite low.

The same kind of reasoning explains why we shouldn’t refrain from raising general quality of life on the grounds that this makes recent deaths worse, (since they lose out on better lives than they otherwise would have). Even if posthumous harm is possible, it could only make recent deaths more harmful by shifting the same overall loss from the conjunction of S died and we refrained from raising the quality of life to S died. As such, neither the TRIA nor LCA imply we can harm the recently deceased by raising the current quality of life.

This might be sufficient as an account of the value of different deaths, but if the TRIA is to guide action, we must translate this talk of value into talk of choiceworthiness. To do this, we must determine which of an individual’s time-relative interests we ought to consider in making choices.

Consider the following case;

Choice Between Deaths: A doctor can either save a 2-year-old, or let him die today. If he’s saved today, he will live happily until dying at age 15. (Greaves, 2019).

Even though the child’s death at age 15 would be worse than their death at age two, (due to it frustrating stronger TRIs) it seems odd to think this gives the doctor even a pro-tanto reason to let him die today9. This implies that we shouldn’t consider all the possible TRIs of an individual when deciding whether to save them.

If we only consider an individual’s presently-existing TRIs, that would imply we ought to save the 2-year-old. But such a presentist account is undermined by the following case;

Doctor’s Choice: A baby (Alice) and a 30-year-old (Ben) will each die in 30 years unless they are treated now. The doctor can only save one; whoever is treated will live happily until age 80, (Greaves, 2019).

A presentist version of the TRIA would currently recommend treating Ben. But in 30 years, when Ben only has 20 years left to live, and Alice is about to die, a presentist version of the TRIA will imply you made the wrong choice, since now Alice’s TRIs are stronger. This kind of foreseeable regret should be avoided, (Broome, 1999).

Considering only the actual TRIs of individuals won’t work either. This is because considering only the actual TRIs of individuals leads to choice-relative evaluations of acts, (Holtug, 2011). In the Choice Between Deaths, if you let the two year old die, you’ve frustrated the only actual interests, and done the wrong thing. But if you let them die at 15 instead, then (presuming the TRIs of the 15-year-old-self are strong enough), you’ve also done the wrong thing. As such, an actualist account of TRIs can’t work either.

In light of these issues, McMahan now advocates an independentist view about which TRIs count in decision making (McMahan, 2019a, 2021). McMahan divides TRIs into two groups, dependent and independent. Independent TRIs are those TRIs that will exist regardless of which choice one makes in a particular situation. Dependent TRIs are those TRIs that will only exist if some options are chosen, but not others. On this view, we should only care about individuals’ independent TRIs in assessing the badness of death10.

Since the only independent TRIs in Choice Between Deaths are those of the two-year-old, this implies we should save the two-year-old. In Doctor’s Choice, the doctor can either frustrate the independent TRIs of Alice at 30, or Ben at 60. Since Alice’s TRIs at 30 are stronger than Ben’s at 60, the doctor should save Alice as a baby.

This independentist version of the TRIA is a close analogue to asymmetrical views in population ethics. Unsurprisingly, it faces analogous problems. The most serious of these problems are cyclical recommendations in sets of pairwise choices, and contraction inconsistency in choices with 3 or more options.

Timothy Campbell explains how an independentist TRIA implies these results, (Campbell, 2019). Regarding cyclical evaluations, Campbell notes that an independentist TRIA is committed to the following three claims11:

  1. Saving Newborns from Death: If we can save a newborn at no cost, we should do so, even if they’ll still die at age 30.
  2. Young Adults over Newborns: If we can extend the life of a young adult by 40 years, or a newborn by 80 years, we should extend the young adult’s life.
  3. Weak Life Extension: If we can save one newborn and let them live to 80, or a second newborn and let them live to 70, we should save the one who will live to 80.

Campbell considers the following three options, where X, Y, and Z refer to outcomes where Alice and Bob die at the corresponding ages. In each case, both Alice and Bob are newborns at the time the doctor must decide between these outcomes:

Outcome:XYZ
Alice age at death03070
Bob age at death80800

When only two options are available at once, the TRIA is committed to cyclical evaluations in the following way:

  • We ought to choose Y over X, via Saving Newborns from Death.
  • We ought to choose Z over Y, via Young Adults over Newborns.
  • We ought to choose X over Z, via Weak Life Extension.

As a result of this cyclic evaluation, whatever choice we think is optimal when all three options are available will generate contraction inconsistency. McMahan thinks option Y is best in the three-option case, (McMahan, 2019b). But this commits him to the claim; We ought to choose Y if our options are X, Y, and Z. However, if Y and Z are our only options, then, as Young Adults over Newborns says, we ought to choose Z rather than Y, (Campbell, 2019). Analogous examples of contraction inconsistency follow if one thinks X or Z is optimal instead.

McMahan bites the bullet on both counts, (McMahan, 2019b). The basic source of both problems is the fact that which TRIs are independent, (and hence how choiceworthy a given option is) depends on what other options are available. If only Y and Z are available, then Alice’s TRIs up to age 30 are independent, and so we have a strong reason not to frustrate them by picking option Y. But if option X is available, then these TRIs are merely dependent, and we have no reason (or a much weaker reason) not to frustrate them.

At first glance, this seems like a decisive mark against the TRIA, but I think there’s two reasons this isn’t the case. The first and strongest reason is that, in order to derive practical guidance from an account of animal death in the two most important cases, (deaths in animal agriculture and the deaths of wild animals), one must also take a stance on the question of whether the badness of animal death is offset by bringing new animals into existence, (i.e., whether animals are replaceable, (Singer, 2011, Kagan, 2015)). But in order to do this, one must settle on a view of population ethics. Insofar as various impossibility results hold, (e.g. Arrhenius, forthcoming), any such view will entail similarly unintuitive conclusions. In particular, if one is already sympathetic to asymmetrical views in population ethics generally, and if these really do imply contraction inconsistency and cyclical evaluations, (of either prospects or choices), then there’s no additional cost to accepting the TRIA as an account of the badness of death.

The second reason is the fact that the cyclic nature of the TRIA only needs to apply to the ‘ought to choose X over Y’ relation, not the ‘better-than’ relation, which more plausibly must be transitive. Additionally, insofar as some views of sequential decision-making make it easier to avoid value pumps arising due an to intransitive ‘ought to choose X over Y’ relation, than an intransitive ‘better-than’ relation, this further mitigates the damage of Campbell’s objections, (Thomas, 2019, (pre-print version)).

Part 2: Animals

In this section, I’ll begin by explaining what each of the three views on the value of death imply for simple cases of killing animals without replacement. Then I will consider what the three views (when combined with different views of welfare and theses in population ethics) imply for more complex cases where the animals killed are replaced by more animals with a similar quality of life.

2.1 - Killing Animals Without Replacement

Implications of the Life-Comparative Account

Applying the LCA to questions about killing animals is the most straightforward. One has a reason not to kill an animal whenever doing so would make its life worse, i.e. reduce the amount of welfare it’s life would contain. The strength of this reason is proportional to this reduction in welfare. The same applies mutatis mutandis, for reasons to kill animals, so you have a reason to kill an animal for its own sake if doing so will increase the amount of welfare it’s life would contain, (i.e. it’s life from that point on would have been net-negative).

Unlike the other two accounts, the strength of these reasons doesn’t directly depend on the psychological attributes of the animal being killed. All that matters is that the animal would have been identical with the subject of the future harm or welfare. This in turn means the LCA isn’t sensitive to how the harms and benefits within a life are psychologically related to each other, it’s just sensitive to their overall balance.

This can lead to some counterintuitive verdicts, as in the following case:

Early Pain, Later Pleasure: An animal that’s only weakly psychologically connected to itself across time faces the following choice. The animal can either die painlessly now, or it can have an operation which will leave it in extreme pain for a year (welfare of -9) followed by ten years of moderately good life (welfare of +1 each).

The LCA claims we ought to give the animal the operation, since letting it die would make it miss out on one unit of welfare, and hence be bad for it12. Remember, the LCA gives this verdict so long as the animal’s identity persists over time. If we hold to a psychological-continuity view of animal identity, this is moderately counterintuitive, since the animal’s identity could persist even if the strength of their psychological connections are reasonably weak. This might be especially counterintuitive if the animal is typically strongly connected over time, with the sole exception of a major (but identity-preserving) disconnection at the end of the year of pain. If we instead hold an animalist view of animal identity in combination with the LCA, then it seems like the animal’s identity could persist even in the complete absence of psychological connections, so long as it retains the ability to experience pain and pleasure. Death being bad for such an animal in Early Pain, Later Pleasure seems even stranger.

As in the case of human beings, the LCA retains the counterintuitive implication that the worst time for an animal to die (in a continuously net-positive life) is when it has just begun to exist. Furthermore, unlike humans, this can’t be explained by claiming, had the animal not died, they would judge this the worst time to have died in retrospect, since most adult animals lack the capacity for such sophisticated judgments.

Since it’s a deprivation-based view, advocates of the LCA will also have to deal with the same questions around counterfactual impact and realism that advocates of the TRIA face.

Despite these counterintuitive aspects, the LCA does have some advantages as an account of the badness of death for animals. For one thing, it’s compatible with both hedonism and desire-based accounts of welfare13. Given hedonism, the badness of a death is proportional to the net balance of pain and pleasure the animal’s life after the time of death would have contained. Given a desire-based theory, it’s the net balance of desire satisfaction and frustration the life after the time of death would have contained, combined with any of its desires which were actually frustrated or satisfied by death. As such, when combined with a desire-based theory of welfare, the LCA will provide a slightly stronger protection against death for animals who have intrinsic desires not to die than those which do not.

The LCA also deals with uncertainty quite easily. Since it treats all harms and benefits symmetrically, and doesn’t care about how they are psychologically related, the LCA can be modified to account for uncertainty by saying death harms/benefits an animal whenever it reduces/increases the amount of welfare it’s life would contain in expectation.

In summary, when applied to animal deaths, the LCA retains its counterintuitive implications about when death is bad for animals, (and how bad it is). When combined with an animalist account of animal identity, these implications are even stranger than on a psychological continuity account of animal identity. The two main advantages of the LCA as an account of animal deaths are its compatibility with both hedonism and desire-based theories of wellbeing, and the ease of dealing with choices under uncertainty.

Implications of the Desire-Frustration Account

As I explained in the earlier section on the DFA, I think the Singer-DFA is the best variant of the DFA, so I will largely focus on it here.

In uncertainty-free cases, this view states that the value of death for an animal is proportional to the balance of satisfaction and frustration of its non-misinformed future-directed desires. This can be easily extended to cases of action under uncertainty by saying the strength of our reasons to kill/not kill an animal for its own sake is proportional to the expected balance of satisfaction and frustration of its non-misinformed, future-directed desires.

Unlike the LCA, how this translates into practical recommendations doesn’t depend on views of animal identity, but on views of animal psychology and the nature of desires. Regarding animal psychology, the crucial point is how far forward in time their desires reach, and how strong these desires are. That means that, so long as an animal’s desire to avoid pain reaches forward in time far less than ten years, (and their desire to avoid death itself isn’t too strong), the Singer-DFA will endorse killing the animal in Early Pain, Later Pleasure. Generally speaking, unless an animal would live an almost neutral-welfare life, the Singer-DFA will claim that reasons for or against killing animals are weaker than the LCA implies. This is especially true regarding animals whose desires not to die or to attain future goods are relatively weak.

However, the Singer-DFA can lead to counterintuitive verdicts in other euthanasia cases. Consider;

Early Pleasure, Later Pain**:** An animal currently has no direct desire not to die, and its desires for future goods only reach one week into the future. Despite this, its other psychological connections, such as memory, are quite strong and extensive. Now is your only chance to kill the animal. If you don’t kill it, it will enjoy a mildly good life for a week, then experience extreme pain for a year before dying.

The Singer-DFA implies the death of this animal would be bad for it.

There are two ways to mitigate this implication. First, the Singer-DFA could be modified to count ideally-informed desires, not non-misinformed desires as the relevant kind. The view could then claim that, were the animal more informed, it would desire to avoid the future pain, and so its death would be good for it. This verdict might hold for a handful of animals, but most non-human animals either have the intrinsic capacity for such desires, and already have them, or lack the capacity entirely. To derive this result for the second class of animals, one would have to presume an implausibly strong notion of idealization, (involving enhancing cognitive abilities, and hence radically departing from actual desires) to get this result.

The second option would be to accept that the death would be bad for the animal, but good in a more absolute sense. If we have a reason to prevent the existence of negative welfare, regardless of whether doing so would currently benefit any beings, we could still have a reason to kill the animal, though doing so would be bad for its (current) self. This modification doesn’t quite make the view incompatible with person-affecting views in population ethics, since this could be viewed as preventing a posthumous harm to an actual individual, but it’s an awkward position to hold.

Additionally, unless the (presently-undesired) future welfare of an animal can at least offset the badness of future harms, this view will imply we should kill an animal whenever it would prevent the existence of a single instant of mildly net-negative welfare14, (Belshaw, 2015). As such the Singer-DFA is only plausibly compatible with (roughly) totalist or offsetting-asymmetry views about possible future harms. Still, I think accepting this incompatibility is better than appealing to idealized animal desires.

Aside from this downside, the other main flaw of the Singer-DFA as an account of animal death is its incompatibility with hedonism. Insofar as hedonism is especially plausible as an account of animal wellbeing, this makes the Singer-DFA especially ill-suited to an account of animal deaths.

Implications of the Time-Relative Interests Account

The TRIA implies we have a reason to kill or save an animal for its own sake when doing so would, on net, frustrate or satisfy its independent TRIs. The strength of this reason is proportional to the magnitude of this net frustration or satisfaction.

If this is the only relevant reason, the TRIA will generate similar verdicts to the Singer-DFA in cases similar to Early Pleasure, Later Pain. To see this, consider a variant of the case where the animal would have a very strong degree of psychological unity between itself now and itself during the time of mildly-positive welfare, but very a very weak degree of psychological unity between itself now and itself during the year of pain. In such a case, the TRIA implies the death of the animal is bad for it. Even if this is a plausible account of the animal’s interests, it certainly seems incomplete as an account of what we ought to do in this case. Instead, as I discussed in the last few paragraphs of the previous section, it seems like similar questions about future harms and benefits should be considered.

McMahan recognises this dilemma between ignoring future harms and implying the mandatory euthanasia of animals with little psychological unity over time, (McMahan, 2015, 2019a). As such, he introduces an element of offsetting into his account, but it’s both more plausible and more interesting than typical accounts of offsetting in population ethics.

For McMahan, insofar as the value of an individual’s future non-time-relative interest is not entirely ‘accounted for’ by their independent TRIs, it can still have offsetting value in the following way, (McMahan, 2019a)15.

Suppose, (under some option-set) Spot has independent TRIs up to and including time t1. Given a possible outcome O, he will exist for four time-periods, t1, t2, t3, and t4. His welfare levels in each period would be +60, +30, -60, and +100. The degree of psychological unity between himself at t1 and t2 would be 0.5; 0.5 between t2 and t3; 0.5 between t3 and t4; 0.25 between t1 and t3; and 0.125 between t1 and t4.

To determine the value of outcome O, first determine Spot’s independent TRI in this sequence of moments. This equals (+60 * 1) + (+30 * 0.5) + (-60 * 0.25) + (+100 * 0.125), or +72.5. This leaves ‘uncounted’ +15 welfare at t2, -45 welfare at t3, and +87.5 welfare at t4. Now try to offset moments of uncounted negative welfare by spending his excess positive welfare at nearby moments, subject to the discount rate provided by the degree of psychological unity between these moments16. After spending all 15 units of welfare from t2, and 75 units from t4 to entirely offset t3 at a discount rate of 0.5, we get a remaining offset-adjusted welfare of 0, 0, 0, and +12.5 for the series.

Now consider a similar outcome O*, which is identical to O, but without t4. This would give Spot a TRI of + 60, with offset-adjusted remaining welfare of 0, 0, -37.5.

We have two ways to evaluate O and O*. We can endorse a strongly-asymmetrical TRIA, and claim the value of an outcome equals the value of independent TRIs, plus any remaining offset-adjusted negative welfare. This would give O an overall value of +72.5, and O* a value of +22.5. Alternatively, we could endorse a weakly-asymmetrical TRIA, where the value of anoutcome equals the value of independent TRIs, plus any remaining offset-adjusted negative welfare, plus any remaining offset-adjusted positive welfare discounted by some factor r such that 0 < r < 1. If r = 0.2, O* would still be worth +22.5, but O* would be worth 75, not 72.5. On either version, the expected value of an uncertain prospect is just the likelihood of it occurring, times this overall value.

Either of these versions implies we should kill (weakly-connected) animals in cases like Early Pleasure, Later Pain. The strongly-asymmetrical TRIA implies we should kill weakly-connected animals in Early Pain, Later Pleasure. Unless we pick an extremely high value of r, the weakly-asymmetrical TRIA also implies this.

Aside from these intuitive results, either account of offsetting also has the interesting implication that, even after considering an animal’s TRIs, it’s not just the balance, but also the distribution of positive and negative welfare across an animal’s life that’s morally important. To see this, consider an outcome that frustrates a dependent interest that’s only weakly connected to the dependent interests this outcome would satisfy. If this frustrated interest isn’t accounted for by a very strong time-relative interest, its badness will largely fail to be offset, and so creating it will count strongly against the outcome. This contrasts with an outcome that frustrates a dependent interest which is more closely connected to satisfied dependent interests. Even if this interest is just as strong, its badness will be offset to a greater degree, and so count less strongly against picking this outcome.

More generally, when it comes to creating, then frustrating dependent interests, our reasons not to do so will typically be stronger when;

  1. The being that holds them will only have weak independent TRIs in avoiding their frustration.

And

  1. The frustrated interests will be only weakly psychologically connected to/continuous with any strong, satisfied interests.

In practice, (b) might favor making extremely painful experiences (like slaughter) less painful over making the day-to-day life of farmed animals less painful. Alternately, (and more strangely), it might favor making sure animals’ lives are especially good near the times they will have strongly negative experiences17. Both these imbalances in our priorities should be especially strong when it comes to animals with little psychological unity over time.

These kinds of implications are why I called this kind of offsetting interesting. I also think it’s more plausible than certain views of offsetting in population ethics. This is because it’s difficult to explain why good dependent lives should be able to offset bad dependent lives, but not bad independent lives, since they seem equally disconnected from both, (Thomas, 2019). But since offsetting frustrated interests with satisfied interests only works insofar as the two interests are psychologically connected to/continuous with each other, there’s a natural explanation of why this offsetting doesn’t occur across-lives.

2.2 - Killing Animals with Replacement

The replaceability thesis is the view that, even if death is bad for animals that die, killing animals is still permissible, so long as this badness is entirely compensated for by bringing new (happy) animals into existence, (Višak, 2015, Kagan, 2015). The logic of the larder argument is a stronger extension of the replaceability thesis that claims it’s not only permissible to kill and replace animals in this way, but positively good to do so, since the animals are benefited by their good lives, (Višak, 2015). Since the logic of the larder implies the replaceability thesis, if some combination of views implies the replaceability thesis is false, they also imply the logic of the larder is false.

Whether either of these views are true depends on one’s views about population ethics, animal wellbeing, and the badness of death. In this section, I will consider how different views about welfare and population ethics relate to the three accounts of death’s badness in cases of killing animals with replacement. However, I will not discuss how independently plausible each of these views in population ethics are, as doing so would take me beyond the scope of this essay. Instead, I will divide these views into three broad classes, based on how easily they can imply either replaceability or the logic of the larder.

First are views in population ethics which are mostly or entirely incompatible with the replaceability argument, (Kagan, 2015). These include person-affecting views, which only consider the effect of an outcome on those individuals who will inevitably exist, regardless of whether that outcome is chosen, (Višak, 2015). ‘Strongly-asymmetrical’ views, meaning any view that thinks creating good, dependent lives can offset the harm of creating bad dependent lives, but not count positively in favor of an outcome, also fall in this camp, (McMahan, 2019b)18.

The only way either view could imply replaceability is if we think death is not at all bad for animals, and killing them isn’t wrong for other reasons. The only way to claim this about all animals would be to accept something like the Epicurean view of death. The easiest way to get this result about some animals would be combining one of these views with a version of the DFA that places quite a high bar on which animals have the relevant interests. The animals that don’t reach this bar would then be replaceable.

But, as mentioned previously, the most plausible version of the DFA, (the Singer-DFA) is an awkward fit with person-affecting views in population ethics. So this leaves the Singer-DFA, plus a strong-asymmetry, as the easier way to imply that (some) animals are replaceable. Exactly which animals are replaceable will depend on how demanding one’s account of intrinsic desires is. Even if such a view is true, it won’t imply the logic of the larder, and it seems unlikely they can exclude animals like pigs from having any intrinsic desires.

A second set of views can more easily imply replaceability, but only under certain conditions.

One such view is a slight variant of the strong asymmetry which claims that the benefits to dependent individuals can also offset harms to independent individuals, including the harm of death, (Kagan, 2015). Whether such a view implies replaceability depends on (a) how bad death is for the independent animals, and (b) whether the benefits to the dependent animals are discounted before offsetting this harm. If death isn’t too bad, and the benefits aren’t too heavily discounted, this view could imply replaceability, but not the logic of the larder.

If we modify this view to claim that benefits to dependent individuals can also count positively in favor of some outcome at some discounted rate, we can imply both replaceability and the logic of the larder about certain animals. As before, if this discount is minor enough, and the relevant deaths aren’t too bad, this view can imply replaceability, and the logic of the larder.

However, either of these views, when combined with the LCA or the TRIA, face an extra difficulty in implying either replaceability or the logic of the larder.

This is because the LCA and TRIA regard the badness of death as either entirely or partly a function of the quality of the life lost. This means that, as the quality of life which animals would live, (were they not killed) increases, so does the badness of their deaths. But if, as we raise their quality of life, the badness of the animal’s deaths increases faster than the discounted contribution made by the benefits to dependent individuals, replaceability may never obtain.

Whether this is true given the LCA just depends on (1) how prematurely the animals are killed, (both since earlier deaths are worse, and the shorter lives of the dependents would be less good), and (2) the discount rate on benefits to dependent individuals. Given the TRIA, this also depends on a third factor; (3) the degree of psychological unity of the animals which are killed. Generally, given the LCA, this impossibility of permissible animal agriculture applies to all the cases in which it would apply under the TRIA, and the cases where animals have too little psychological unity for the TRIA to imply this result.

The third kind of view is totalism, which counts all the interests of both dependent and independent interests equally, (Kagan, 2015). On this view, an outcome is good to the extent that it satisfies these interests. Even if combined with the LCA or TRIA, totalism will not fall for the same double-bind as the previous views, since it doesn’t discount benefits to future people. The only way totalism won’t automatically imply replaceability and the logic of the larder is if we combine it with certain desire-satisfaction theories of welfare, (Kagan, 2015)19.

One such theory is Singer’s ‘debit model’ of preferences, (Singer, 2011). On this view, the frustration of preferences counts against an outcome, but the satisfaction of preferences doesn’t count in favor of it. Therefore, so long as animals have some preference not to die, the satisfaction of dependent future animals can’t offset this, and these animals aren’t replaceable.

Alternatively, Shelly Kagan proposes a simple model of preference-satisfaction where, at any given moment, individuals with preferences to exist at future moments have distinct preferences to exist at each of those future moments, (Kagan, 2015). The value of their lives is just the balance of all these preferences.

Consider an individual who wants to live from t1 to t4. At t1, they have 4 preferences, one to exist at each time t1-t4. At t2 they have 3 new preferences, one each for t2-t4. At t3 they have two new preferences, and at t4 they have one. If this individual lives until t4, the value of his life will be +10, since all ten of his preferences will be satisfied. But if he’s killed after t2, then the value of his life will have been -1, (3 + -4). This is because he never forms the three preferences at t3 and t4, and four of his earlier preferences are frustrated by dying.

Even if he’s replaced by a new, identical individual after t2, and the new individual lives a life of value +10 to t6, the value of the preferences he holds at t3 and t4 is still only +7. This gives a total value of t1-t4 of only +6 when the original individual is killed and replaced, compared to +10 when he is not.

If a model of preference-satisfaction like this is true, then, even given totalism, the lives of animals with desires to exist in the future will have to be significantly better than neutral to justify replaceability or the logic of the larder. The stronger and more far-ranging their preferences for future life, the better their lives will need to be to compensate their deaths.

Conclusion

I’ve now applied each of the three accounts of death to cases of animal death, and drawn some broad practical conclusions. Of course, these conclusions are tentative, and to be truly practical, they must be supplemented with further research into the psychological properties of different animals. Despite this, I’ll conclude by summarizing these practical conclusions.

Firstly, the LCA tends to imply death (in good lives) is worse for animals than the other two views. This is especially true for animals that lack strong desires about the future, or have little psychological unity across time. Unlike the other two views, its verdicts about the badness of animal deaths don’t depend on facts about animal psychology, but on the correct account of animal identity.

The DFA generally views animal deaths, especially the deaths of psychologically-simpler animals, as less bad (or good) than the other two accounts. In order to avoid implausible verdicts in Early Pleasure, Later Pain, the DFA should be supplemented by additional reasons for preventing animal’s future pain. In this form, it tends to regard preventing moments of negative-welfare as especially important.

Like the DFA, the practical implications of the TRIA depend on facts about animal psychology. The TRIA doesn’t imply avoiding negative welfare is quite as important as the DFA does, but it’s still especially important. Due to its unusual version of offsetting, the TRIA doesn’t just care about the balance of positive and negative welfare in an animal’s life, but also how they are related psychologically.

When it comes to cases of killing animals with replacement, certain views in population ethics more easily imply replaceability than others. However, whether they imply this depends on how they are combined with views about welfare and the badness of death, and facts about different kinds of animals. Even totalist views don’t automatically imply replaceability given certain views about death and welfare. The most interesting practical result here is the implication that, when combined with certain views in population ethics, the LCA and TRIA can imply a kind of double-bind against the permissibility of (certain kinds of) ‘animal-friendly’ agriculture.

Bibliography

Arrhenius, G. (forthcoming). Population Ethics: The Challenge of Future Generations. Oxford University Press.

Belshaw, Christopher. Annihilation : The Sense and Significance of Death, Taylor & Francis Group, 2008. ProQuest Ebook Central

Belshaw, Christopher, (2012). ‘Death, value and desire’, in: Feldman, Fred; Bradley, Ben and Johannsen, Jens (eds), The Oxford Handbook of Philosophy of Death. Oxford Handbooks in Philosophy. Oxford, UK: OUP

Belshaw, Christopher, (2015). ‘Death, Pain, and Animal Life’, in Tatjana Višak, and Robert Garner (eds), The Ethics of Killing Animals (New York, 2015; online edn, Oxford Academic, 22 Oct. 2015)

Bradley, Ben, 2009, Well-being and Death, Oxford: Oxford University Press.

Bradley, Ben, ‘Is Death Bad for a Cow?’, in Tatjana Višak, and Robert Garner (eds), The Ethics of Killing Animals (New York, 2015; online edn, Oxford Academic, 22 Oct. 2015)

Bradley, B., & McDaniel, K. (2013). Death and Desires. The Metaphysics and Ethics of Death, 118–133. https://doi.org/10.1093/ACPROF:OSO/9780199751136.003.0007

Broome, J. (1999). Ethics out of Economics. Cambridge: Cambridge University Press. doi:10.1017/CBO9780511605888

Broome, J. (2019). “The Badness of Dying Early.” In Saving People from the Harm of Death

Campbell, Timothy (2019). Healthcare rationing and the badness of death : should newborns count for less? In Espen Gamlund & Carl Tollef Solberg (eds.), Saving People from the Harm of Death. Oxford University Press.

Greaves, H.R. (2019). Against “the Badness of Death”. Saving People from the Harm of Death.

Holtug, Nils (2011). Killing and the Time-relative Interest Account. The Journal of Ethics 15 (3):169-189.

Kagan, Shelly, ‘Singer on Killing Animals’, in Tatjana Višak, and Robert Garner (eds), The Ethics of Killing Animals (New York, 2015; online edn, Oxford Academic, 22 Oct. 2015)

Korsgaard, Christine, ‘A Kantian Case for Animal Rights’, in Tatjana Višak, and Robert Garner (eds), The Ethics of Killing Animals (New York, 2015; online edn, Oxford Academic, 22 Oct. 2015)

Luper, S. (2021). Death. Fall 2021 ed. [online] Stanford Encyclopedia of Philosophy. Available at: https://plato.stanford.edu/entries/death [Accessed 2 Oct. 2023].

McMahan, Jeff. “Preferences, Death, and the Ethics of Killing”. Preferences, edited by Christoph Fehige and Ulla Wessels, Berlin, New York: De Gruyter, 1998, pp. 471-502.

McMahan, Jeff, The Ethics of Killing: Problems at the Margins of Life (New York, 2002; online edn, Oxford Academic, 1 Nov. 2003)

McMahan, Jeff, ‘The Comparative Badness for Animals of Suffering and Death’, in Tatjana Višak, and Robert Garner (eds), The Ethics of Killing Animals (New York, 2015; online edn, Oxford Academic, 22 Oct. 2015)

McMahan, Jeff, (2019a) ‘Early Death and Later Suffering’, in Espen Gamlund, and Carl Tollef Solberg (eds), Saving People from the Harm of Death, Population-Level Bioethics (New York, 2019; online edn, Oxford Academic, 21 Mar. 2019)

McMahan, Jeff, (2019b) ‘In Defense of the Time-Relative Interest Account: A Response to Campbell’, in Espen Gamlund, and Carl Tollef Solberg (eds), Saving People from the Harm of Death, Population-Level Bioethics (New York, 2019; online edn, Oxford Academic, 21 Mar. 2019)

McMahan, Jeff, ‘Suffering and Moral Status’, in Steve Clarke, Hazem Zohny, and Julian Savulescu (eds), Rethinking Moral Status (Oxford, 2021; online edn, Oxford Academic, 19 Aug. 2021)

Nagel, T., 1970. “Death,” Noûs 4(1):73–80

Parfit, Derek. 1986. Reasons and Persons. Oxford: Oxford University Press.

Rowlands, M. (2009). Virtue Ethics and Animals. In: Animal Rights. Palgrave Macmillan, London.

Singer, P. 2011. Practical Ethics. 3rd ed. Cambridge: Cambridge University Press.

Thomas, Teruji (2019). The Asymmetry, Uncertainty, and the Long Term. Philosophy and Phenomenological Research.

Velleman, D. (1993). Well-Being and Time. Pacific Philosophical Quarterly, 72(1).

Višak, Tatjana, Robert Garner, and Peter Singer, The Ethics of Killing Animals (New York, 2015; online edn, Oxford Academic, 22 Oct. 2015)

Višak, Tatjana, ‘Do Utilitarians Need to Accept the Replaceability Argument?’, in Tatjana Višak, and Robert Garner (eds), The Ethics of Killing Animals (New York, 2015; online edn, Oxford Academic, 22 Oct. 2015)

Williams, B. (1973). The Makropulos case: Reflections on the tedium of immortality. In Problems of the Self: Philosophical Papers 1956–1972 (pp. 82-100). Cambridge: Cambridge University Press.

Footnotes

  1. The virtue-ethical considerations presented in Rowlands 2009 and the rights-based considerations presented in Korsgaard 2015 are both relevant and interesting. ↩

  2. Or beneficial, if the life they’re deprived of would have been bad. ↩

  3. Belshaw’s usage of the term implies individuals can have desires that are both categorical and conditional, they just have to only be conditional on things other than one’s continuing to live. The Venice-desire is one example. ↩

  4. I’m neglecting some refinements about the temporary absence of categorical desires, but none of my objections turn on this aspect of Belshaw’s view. See Belshaw 2012, pg. 10 for details. ↩

  5. This account is similar to some of Peter Singer’s views, (e.g. Singer, 2011, pg. 71-94), but I’m mentioning it as an account of the badness of death, not the wrongness of killing, and avoiding Singer’s account of preferences. I’m just using his name as a memorable shorthand. ↩

  6. One might object that Belshaw really means to refer to extrinsic desires when he’s talking about conditional desires. But this is incompatible with his claim that infants and most animals only have conditional desires, (Belshaw, 2015), since a being can’t only have extrinsic desires on pain of regress or circularity. ↩

  7. This seems to be roughly Belshaw’s view, though he’s not very clear on the point, (Belshaw, 2008, pg. 117). ↩

  8. When exactly it’s worst depends on one’s exact account of psychological unity. ↩

  9. I mean it’s odd to consider the TRIs of the 15 year old self in this way. But if the 15-year-old self would have a very strong desire not to die, or their death would be painful, then considering these non-time-relative interests seems reasonable. ↩

  10. I’ll return to the question of how to deal with non-time-relative interests later. ↩

  11. In this discussion, assume all the years of life are net-positive. ↩

  12. If one is persuaded by totalist views in population ethics, this might seem innocuous. But the counterintuitive aspects of the LCA is that (a) it implies this death would be bad for the animal, and (b) it always implies this kind of ‘within-life-totalism’, even if one holds non-totalist views in population ethics. ↩

  13. Given the diversity of objective-list views of wellbeing, I won’t be able to cover their interactions with accounts of animal death. ↩

  14. Or even just reduce the probability of such an instant existing, given uncertain choices. ↩

  15. McMahan doesn’t give a precise account of this in his published work, but the following is a plausible, (though artificially-precise, oversimplified, and ambiguous) model of the view he hints at in McMahan 2019a. ↩

  16. More specifically, do this while retaining the largest possible sum of offset-adjusted welfare across all the moments. If there are two ways to do this, draw from moments of positive welfare symmetrically. ↩

  17. Though doing this may be less efficient than merely reducing pain, since the pleasant moments would only offset the pain at a discounted rate. ↩

  18. Variants of either view that also count the harms of dependent people as weighing against an outcome, but don’t think these harms can even be offset by benefits to dependent individuals, will never imply replaceability. ↩

  19. Or we appeal to reasons not to kill animals which aren’t derived from the harmfulness of their deaths, which I won’t do here. ↩

Graph View

Backlinks