Cross posted from my blog, but it makes more sense being here. This is quite long for the forum, at 5,500 words, so if it's too long I'd appreciate if someone told me what I'd need to cut it down to and I'll provide an edit or something.


I’ve been surprised to learn recently that so many people I know in the effective altruism community believe there is more total suffering than happiness in the lives of wild animals. Brian Tomasik appears to be the main current proponent of this view, developing the work done by economist Yew-Kwang Ng in the 90s. Tomasik has taken his concerns with wild animal suffering (WAS) to its logical limit, arguing we should consider destroying ecosytems so that fewer animals exist. This cuts against the popular intuition, frequently promoted by nature documentaries, that wild animals live enjoyable, if somewhat barbaric, lives and are best left to their own devices. As the number of animals in the wild is so vast, WAS is therefore potentially of huge moral importance. Concerned that I had overlooked the area, I investigated. After some consideration, I think the arguments in favour of there being net WAS are unconvincing. I decided to write this essay to explain why others should be similarly unconvinced. I’ve reconstructed Tomasik’s argument below and then set out five objections. For readers impatient for the punchline, the thrust of my argument is 1) the case for net wild animal is highly selective, relying on imagining obviously bad aspects of animals' lives, such as being eaten alive, rather than accounting for all their positive and negative experiences and 2) it requires immense speculation about how good or bad animals’ experiences are (e.g. how many hours of eating grass is equivalent to 5 minutes of being eaten alive? how many unhappy ants is worth one happy lion?). We should be wary of overstepping our epistemic limits and engaging in careless anthropomorphism. I, briefly, present my own moral response: certain, targeted interventions may reduce WAS but habitat destruction is unlikely to be justified. I note WAS is unlikely to be a particularly cost-effective way of reducing suffering. To be clear, I do not claim the lives of wild animals are net positive or negative overall. I do not think I (or anyone else) can show that. Nor do I deny there is wild animal suffering. There clearly is. My point is that Tomasik overstates the case for net wild animal suffering and understates the difficult of making interpersonal and interspecies comparisons of happiness.

Update: Brian Tomasik has written a reply on his website. I list my reply to his reply at the bottom of this document.

  1. Introduction

Animal welfare advocates have historically tended to focus on improving the lives of the (domesticated) animals humans interact with, such as household pets or pigs and chickens in factory farms. More recently, advocates have started to focus on the problem of wild animal suffering (WAS) and argue it merits at least as much attention. As Tomasik notes:

Wild animals are vastly more numerous than animals on factory farms, in laboratories, or kept as pets. Most of these animals endure intense suffering during their lives, such as from disease, hunger, cold, injury, and chronic fear of predators. Many wild animals give birth to tens or hundreds of offspring at a time, most of which die young, often in painful ways (Tomasik 2015)

Whilst it seems clear there is much suffering in nature, Tomasik makes a bolder empirical claim that wild animals experience more suffering than happiness and so nature plausibly contains net negative welfare. He uses those empirical claims to support moral arguments about how to reduce suffering, such as by causing animals not to exist. In this paper I set out why I don’t think strong belief is warranted.

The structure of the paper is as follows. Section 2 sketches Tomasik's empirical arguments for why wild animals experience more suffering than happiness. Section 3 discuss five objections I list in the section itself. Sections 4 identifies what someone who wants to show there is more suffering in nature would need to show and argue this is will be difficult to do. Section 5 closes with a few thoughts on the moral implications of my analysis. In this paper I assess the lives of animals from a hedonistic framework, thus assuming all that makes their lives go well is experiencing happiness and avoiding unhappiness. I assume happiness refers the positive conscious states - the ones that feel good to the experiencer - and unhappiness the converse. I use pleasure as a synonym for happiness and I take pain and suffering as synonyms for unhappiness. Whilst the process of dying can obviously be pain, I assume being dead is equivalent to being a neutral state of neither pleasure nor pain. I make no normative considerations until section 5: the main aim of this paper is just to assess the factual question of whether there is more suffering than happiness in nature before deciding what to do about it.

  1. The argument for why there is more suffering than happiness in nature

Tomasik's (2015, 135-142) argument has a number of components, so I will recreate the core of his argument in steps so as to motivate it:

Premise 1: Many animals experience painful deaths. For instance, as a zebra might be eaten alive by lions and die over the course of some minutes.

Premise 2: Even where they are not killed by other animals, death by disease, starvation or other means is also very unpleasant.

Premise 3: Animal life before their deaths is not necessarily pleasant overall. They may enjoy some pleasures such as food and sex. However, unlike humans, animals are often hungry, dehydrated, hot or cold and constantly living in fear of predators.

Conclusion 1: Even if animals lives contains some happiness, this is probably outweighed by their suffering at death.

Premise 4: Small animals, such as insects, are probably the worst off: they have short lives of days and weeks followed by death.

Premise 5: The majority of animals in nature are these small animals. These are the ones produced by ‘r-selection’, birthing many offspring in the hope some survive.

This contrasts with ‘k-selection’ animals such as humans, who produce few offspring and put lots of effort into stopping them dying.

Conclusion 2: Most animals in nature have short lives with early deaths.

Conclusion 3: Even if smaller animals experience less happiness than larger ones, the fact there are so many of them, and their lives so bad, suggests there is more suffering in nature than happiness.

Tomasik discusses and dismisses two objections to his argument. First, that humans are inclined to misjudge the happiness of animals. Tomasik claims we might be underestimating the suffering of animals because, from our air-conditioned rooms, we are no longer used to experiencing anything like it. Second, he considers why we don't see animals committing suicide if they are so unhappy. I’ll return to the first point later on, but I don't see why discussing suicides is relevant: it's presumably quite hard for animals to kill themselves, except by falling from a great height or drowning. Further, you'd expect animals to only attempt suicide once they are below the neutral level of happiness anyway, so that wouldn’t tell us anything whether non-suicidal animals have net positive or net negative lives.

  1. Objections to the argument there is more suffering than happiness in nature

In this section I present five objections, which I outline here. First, the analysis is selective. It focuses on the most vividly painful moments in nature, such as being killed, and gives very little analysis to the average, humdrum and potentially pleasant lives of animals. Second, judging how much animals suffer will be hard. Trying to assess how happy animals are by considering how much we, as humans, would like to be them is an irrelevant anthropomorphic mistake. It's not clear how we could make the necessary cardinal comparisons either, except by assumption. Third, the analysis is one-sided. Only the suffering of the prey is counted, not the pleasure of the predators. Fourth, it employs a 'bait and switch', treating relevant different types of animals as if they were the same. Having argued 'k-selector' animals (e.g. gazelles, which have few offspring, few of whom die immediately) have bad lives and painful deaths, he argues there is more suffering than happiness in nature because most animals are 'r-selectors' animals (e.g. fish, which have many offspring, most of which die immediately). Clearly, we should treat different animals differently. Fifth, it is incomplete. We need a common way of assessing the suffering of animals of different types to establish where suffering lies. Having presented these objections, I draw them together. I suggest, if we want to know the balance of happiness to suffering in nature, we should to give an evolutionary explanation of how much pleasure and pain the animals in a given ecosystem would experience on a daily basis. This analysis will be piecemeal, riddled with anthropomorphic assumptions, and likely different from animal to animal anyway.

I'll now discuss these objections in turn.

3.1. Selective analysis: relies heavily on assumptions about the suffering caused by death

The main argument for the idea that there is net WAS seems to be that most animals’ deaths are so bad that, even if they experience some happiness whilst alive, it outweighs the happiness they experience beforehand. In the paper, Tomasik quotes a vivid account of a zebra being killed by a lion (Tomasik 2015, 135).

Let's assess the zebra example provided. The life expectancy of a zebra in the wild is 25 years. In the account Tomasik gives the zebra dies painfully over 6 minutes. As a percentage of the zebra's life, that 6 minutes is 0.000046%. Let's assume the zebra is asleep for 8 hours of every day, so that 6 minutes is 0.000068% of its conscious experience. Even granting the death of the zebra is gruesome and those 6 minutes may feel like a long time, that is nevertheless a tiny fraction of the animal's entire life and should not be overstated.

For the sake of argument, let's assume that we can make cardinal intra- and interpersonal comparisons of utility of and across other species and do so lazily by intuition. That is to say, we can just guess how many times worse 6 minutes of being eaten is that 6 minutes of some other experience, such as eating grass. Let's also assume the zebra's life has been happy overall excluding the last 6 minutes of her death (I’ll come back to assessing the day-to-day lives on animals in the next section). Using the numbers above, for the zebra's death to cause so much suffering it outweighs the pleasure in the rest of its life, that means each minute of its death would need to be at least 2,190,000 times more painful than the average minute of the zebra's life. I provide no evidence for this, but intuitively it seems rather hard to believe moments of dying could really be that bad and so much worse than any other experience.

Some might object I've chosen an unrepresentative case: zebras can live 25 years, in comparison to which 6 minutes is rather short. The ratio of time-dying to time-before-death may vary tremendously. It would not advance the philosophical analysis to consider a range of cases, so I will end this section by pointing out that even if it takes some animal an hour to die, if the animal lives for 1.15 years and has had a positive life until that point, each minute of death would still need to be 10,000 times more unpleasant than its pre-dying life was pleasant. Even in this scenario death would still need to be implausibly bad to outweigh even a slightly pleasant life.

On the assumption that time spent dying is a small part of animals’ lives, the crucial question must be whether the ordinary life of an animal is pleasant or not. Tomasik notes that, unlike us, animals may not have access to food and water whenever they want it, shelter if it rains, and must be constantly wary of predators. However, isn't sufficient to prove his case that animals have some bad moments in their lives. What his arguments need to show is that there is more suffering than happiness overall. Presumably there are some positive moments in animals lives. I assume zebras enjoy eating grass, having sex, and enjoy time spent with the rest of the herd. Maybe a zebra's life looks something like Figure 1.


Suppose the lows at t5, t7, and t11 represent hunger, thirst and cold respectively. Nevertheless, the zebra's life is positive overall.

I'm not going to make a general claim about whether the lives of animals are net positive or negative. My point is simply that identifying that there is some suffering doesn't show there is more suffering than happiness in animal lives. Somewhat analogously, if some aliens looked at human life and noted that we are often ill, hungry, bored, stuck in traffic and grieving for the losses of our loved ones, then concluded our lives contained more suffering than happiness we’d think that was a mistake on their part and they’d conducted too partial an analysis.

3.2 Difficulties in judging whether there is net happiness or unhappiness

Once we have a firm grasp of how our zebra has lived her life, we need to know how good or bad those moments were. Simply knowing the zebra has had X happy moments and Y unhappy moments don’t tell me if it’s lived an overall happy life unless I know the magnitude of those moments: maybe there were more good moments, but they were less intense. But how could we figure this out?

One option is just to imagine whether we would want to swap places with the animal we’re concerned with. For instance, Tomasik writes:

There are reasons to think that it would be better not to exist than to find oneself born as an insect, struggle to navigate the world for a few weeks, and then die of dehydration or be caught in a spider’s web (Tomasik 2015, 138)

However, I think we need to be much more careful than this. We often don’t want to swap places with other people (or animals) even though we think they are happy: just because I don’t want to be a rockstar, I shouldn’t conclude rockstars are miserable. Hence, when we consider the lives animals lead, we should avoid carelessly straying into anthropomorphism and judging their lives by our standards. We need to try to understand what it’s life for an insect to be an insect, rather than for us to be an insect.

I suggest a more promising strategy is understanding pleasure and pain as reward mechanisms evolved to help animals survive and reproduce (e.g. see Ng 1995). Economists refer to this as a 'principal-agent' problem: ‘nature’, so to speak, wants animals to pass on their genes. Animals want to experience pleasure and avoid pain. Animals who experience too much or too little pleasure and pain from various activities would not successfully reproduce. To give a simple example, if eating a mouthful of grass caused a zebra to be in a state of distracted ecstasy for several days, they would starve or be easy prey. However, if that zebra was born with a comparatively low desire for food, and it gave them minimal pleasure, you would expect other zebras to out-compete her for resources, causing the low-pleasure zebra not to pass on her genes.

I’m not going to attempt to provide a worked-out version of such a theory, only to point out that that’s the place we’d need to look to get a firm handle of what causes happiness and unhappiness for animals.

What such an account would still fail to tell us, however, is the magnitudes of happiness animals experience. Making interpersonal cardinal comparisons of utility is a general problem: how do I know that this cup of coffee will give me more pleasure that it will give you? You might desire it more, but that wouldn’t tell me how much more pleasure you would experience by drinking it.

Let’s continue to suppose, as I did in the previous section, we can make cardinal comparisons just by imaginative guesswork. To do this for humans, we could point out we all share a neutral point, where we feel neither good nor bad, assume we all have roughly the same maximum level of happiness (based on having near-identical brains), and assume we can assign cardinal values between those two points. Granting this doesn’t help us with animals: we have no idea what maximum capacities of happiness for different species is compared to us (or to other species). Whilst I might be able to automatically generate intuitions about how much worse 6 minutes of being eaten is for the zebra than 6 minutes of eating grass is for the zebra, I don’t have any confidence in those intuitions because I’m not sure what, if anything, would ground their reliability.

One tempting option here would be to assess brain size or, perhaps more specifically, the size of the parts of the brain that seems to be associated with happiness, in different species (Kringelbach, Berridge 2009). But again the problem re-emerges. Let’s say I know zebra have a nucleus accumbens five times smaller than that of a human, I still have no idea if I should say it’s maximum happiness is five times less: I can’t experience what the zebra experiences to check what the link is between brain mass and intensity of experience.

It seems we’re left with armchair hypothesising. As we need to make choices, it’s important to try to be roughly right about animal welfare, rather give up because we can’t be precisely right. However, if someone looks at the entirety of zebra’s life and consider its pains as more intense than I consider them, and its pleasures as less intense than I imagine them, and so decides the zebra has a net negative life, whilst I think it has a net positive life, and both of us are armed with the last biological understanding of pleasure and pain, I’m not sure how we can resolve our differences. In contrast, suppose a friend and I are engaging in my favourite passtime of 'water-watching'. He and I look at two different-shaped water containers and then guess which one holds more. If we disagree, we can at least go and check who was right. We can weigh the containers when full and empty, find the answer and use that information to improve our judgements in future. As there is no way of seing who is right when we disagree about animal suffering, I can't see how we could ever improve our judgements or put faith in them. This is why I don’t see how it follows, if we don’t have much confidence in assigning cardinal values to experiences, that the correct response is to assume animals experience more suffering than happiness. There may well be some fact of the matter about the net balance of happiness over suffering, but it's not one we can know. We shouldn't pretend we can know what we can't, particularly when so much rests on it.

An alternative option would be drop the assumption that we can make cardinal interpersonal comparisons but assume we can make ordinal ones. Thus, we know being eaten is bad, and eating grass is good, but we can’t give a magnitude by which the former is worse than the latter. This does leave open with the possibility of counting up all the moments of an animal’s life and concluding whether it had more good ones than bad ones. Hence someone who was highly sceptical of cardinal comparisons could reach some indicative conclusions, at least for a given animal’s life. It’s not clear this is very helpful because there is still the problem of inter-species comparisons: if you think zebras can suffer more than wasps, but you’re not sure by what amount, there’s no meaningful way to trade them off against each other.

3.3. Ignoring the joy of killing

What’s also missing from the analysis of WAS is the joy predators get from killing prey. It follows, for each animal that suffers the intense pain from being eaten, there must be another animal (or animals) that gains at least some pleasure from eating the eaten animal. This must be part of our analysis too.

Having stated this, I want to make few clarifying remarks. First, I am not making any normative claims about the killing of prey by predators. We might object that even though lions enjoy killing zebra, and need to do so to survive, it is nevertheless wrong that they do it. We might go as far as to say it is so wrong that, if we cannot find a way to manage some ecosystem without using lions to keep the zebra numbers under control, we should wipe out both species. I am not interested in any such arguments. My point is that, if our analysis is just of the balance of happiness and unhappiness in nature, we need to take the pleasure of predators into account.

The second point is that, particularly in light of what I said in the previous section, is to not clear what the balance of happiness is between the suffering of prey and the happiness of predators. I would assume that the unhappiness of the zebra, whilst it dies over 6 minutes, is more intense than the happiness of the lion of the same period. But this misses several important facts. Possibly killing and eating their prey is highly pleasurable to predators in a way it is hard for us human omnivores, who have not evolved solely to eat by hunting, are likely to understand. Whilst the zebra will stop experiencing pain once it is dead, the lion will continue to enjoy the kill for some minutes or hours afterwards. There may be multiple lions who enjoy the kill, whereas there is only one zebra who dies. Obviously we could push this analysis, and wonder if the rest of the zebra herd are sad at the loss of their kin, or perhaps relieved at not having been eaten themselves, but such speculation doesn’t seem very helpful as it’s so unclear.

Third, not all deaths will be accompanied by a joyful predator. For example, a zebra that dies from old age, starvation or disease may not give any pleasure to predators (perhaps its corpse is not found).

If we’re going to count the suffering of prey, we need to count the happiness of predators. I take no position on whether, on average, this produces net happiness or net suffering. What is relevant is that, however much suffering you think the zebra feels, the pleasure of lions much counter-balance it to some degree, which correspondingly makes it harder to conclude there is more net suffering amongst wild animals just because being eaten is painful.

3.4 The ‘Bait and switch’; treating different types of animals as if they were the same

So far I’ve argued against what I earlier labelled Tomasik’s first conclusion. Simply identifying that animals suffer is not sufficient to show they experience net suffering, it’s far from clear how this could be shown and, in any case, he hasn’t accounted for the predators. Let’s assume though that we are satisfied with the first conclusion and turn to premise 4.

The claim here is that smaller animals, such as insects, probably have lives that have a higher ratio of unhappiness to happiness than larger animals. This is because smaller animals, such as fish and insects, tend to engage in ‘r-selection’ as an evolutionary strategy, having very many offspring and hoping some survive. As previously mentioned, this contrasts with ‘k-selectors’ such as humans and zebra who have few offspring and try to care from them. The reason we might expected r-selectors to have worse lives that k-selectors is because so many of them die so quickly. Unlike the zebra, which may at least get to grow up, eat grass, gallop around and mate before dying, the life of the average fly appears to contain few, if any, good moments; most of them get killed before they could experience any pleasures. Hence it seems much less likely they could lead lives with net happiness.

Leaving aside all of the previously raised concerns, the reason I call this a ‘bait and switch’ is because Tomasik’s use of ‘animals’ is ambiguous. The sort of animals he uses to motivate our intuitions that animals can suffer – the k-selecting zebra being eaten – are not obviously similar to the animals who live very short lives – the r-selecting fruit fly. To call both types ‘animals’ and assume the same rules would apply to both is either careless or misleading.

What’s probably required here is a neurological analysis of the different types of animals (again, see Ng 1995 for a plausible account I don’t go into here). As a crude way of making sense of this, we would want to know which sort of brains different types of animals have and judge whether they have the structures required to feel pleasure and pain. One reason to think rocks can’t feel pain is because they lack any sort of nervous system. Even if we’re unsure about how much more pain humans can feel than other animals, we might take the non-existence of certain bits of the brain in other animals as indicative that they can’t suffer, or at least suffer to a lesser extent, depending on what is missing. As Ng notes, it’s unclear if insects have the capacity to feel pain (Ng 1995). If they are not conscious, I can’t see how they could experience suffering, let alone more suffering than happiness.

More generally, one might expect k-selectors to have much greater capacity for happiness than r-selectors. If you produce fewer offspring, it’s more important that they can react to circumstances to can avoid being killed and eaten, hence reward and punishment systems are important. If you produce very many offspring, there is much less need for cognitive complexity and hedonic capacity: operating ‘on autopilot’ is sufficient for some of them to survive and reproduce.

We might assume there’s some natural tension here: we can grant premise 4, that small, short-lived animals have net negative lives, but note that these are likely to be the ones with the least capacities for consciousness. And this is offset by the fact animals that have few offspring, and plausibly have net positive lives, have greater capacities for happiness. Hence premise 5, that there are many more r-selectors in nature than k-selectors is rather uninteresting. And the final conclusion, that there is more suffering in nature, largely on account of the fact there are many more r-selecting entities than k-selecting entities, is unproven.

You might object that I’ve not been very charitable to Tomasik here. The use of a lion eating a zebra was simply used because it’s easier for us to imagine than a fly being eaten by a spider. However, these two seem very different in light of what we know about their neurological capacity. It’s easier to believe the zebra suffers greatly from death in the way a wasp does not.

3.5 An incomplete analysis

If you thought that the lives of all wild animals, large and small, contained more suffering than happiness, you could safely conclude that there is more suffering than happiness is a given ecosystem. Therefore, if you removed that ecosystem and there were no other effects, such as on the happiness of humans, that would reduce suffering.

However, it not at all clear that all wild animals do have lives that contain net suffering. Suppose for the sake of argument large animals, like lions and zebra, have positive lives whereas small animals, like insects, have negative lives. The problem of cardinal interpersonal (or rather, ‘interspecies’) comparisons re-emerges because you have to reach a conclusion about how much happiness the happy animals have compared to the unhappy ones in order to work out if destroying a given ecosystem would increase or reduce happiness. I’m not going to say anything about this, other than that it seems quite hard to do in a principled way.

  1. What would someone do if they wanted to conclude there was more suffering than happiness in nature need to show?

Drawing these criticisms together, I want to briefly comment on what someone who wanted to claim there is net WAS would need to do. To start with, you would need to know whether each species or type of animal had a net happy life or not. This would presumably involve observational work to understand what their ordinary daily experiences were like, rather than just focused on the ‘peak’ experiences like being eaten or eating other animals. Then you’d need to make speculative cardinal judgements about the magnitude of happiness and unhappiness in those moments. Asking yourself whether you would like to be one of those animals is a tempting but misguided approached. More than that, you’d need to make further speculative interspecies comparisons, perhaps based on brain analysis, to work out if the happiness of the happy animals was great than the misery of the unhappy animals. This is important and complicated work and I don’t think Tomasik’s arguments do nearly enough to show it.

As a comparison, it seems much easier to conduct an analysis of the happiness of animals in farms than it does of animals in nature. Factory farmed animals, such as chickens, spend their lives in the same narrow range of conditions that we expect cause them to experience net misery each day. Further, which reduces complication substantially, farmed chickens do not live in an ecosystem with other animals (if you exclude their ‘predators’, humans), so it’s easier to conduct the counterfactual analysis of what would happen if humans ate less chickens. If we sought to stop zebras getting eaten by lions by removing the lions, it’s less clear what effects the expanding zebra population would have on other animals and overall happiness.

  1. What should we do?

The argument over whether there is net suffering or net happiness in nature only seems relevant if you want to argue for the destruction of ecosystems, which Tomasik does want to argue for. There are plenty of reasons to object to habitat destruction, even if one believes there is net WAS, such as concerns about the detrimental impact it might have on human happiness. Someone who did want to advocate for ecosystem destruction wouldn’t necessarily need to show there was net suffering, however. If one adopted a view such as negative utilitarianism, which roughly holds a given amount of unhappiness has more disvalue than the same amount of happiness has value, might think there is net happiness is nature but it would still be better to remove it. Such a view doesn’t seem plausible, so I only note it and move on (see Ord 2013 for a discussion of negative utilitarianism).

If you don’t want to recommend habitat destruction, arguing there is net suffering seems irrelevant because most plausible ethical frameworks will anyway conclude we should aim to reduce suffering for others, at least if we can do so at a trivial cost to ourselves. To reach a pro tanto, rather than an all things considered conclusion, that we should intervene in nature, all that’s required is showing there is some suffering we can easily remove (leaving aside how ‘easily’ might be understood).

On the basis that habitat destruction isn’t a viable option for reducing WAS, it looks like animal welfare advocates will have to a piecemeal approach. This might include finding humane ways to kill animals at the end of life rather than having them being eaten, providing meat-substitutes to predators, and providing healthcare and pain relief to ill and suffering animals.

The relevant question here is cost-effectiveness. I don’t provide any such analysis, which is outside the scope of this paper due to its complexity and uncertainty. Speculatively, if our concern is restricted to animal suffering, wild animal suffering looks much less cost-effective than focusing on factory farming. Presumably, once consumer habits change, either to buying higher welfare animal products, or to becoming vegetarian or vegan, those habits will stay changed and potentially even spread. Whilst habitat destruction would be a one-time cost, it’s unclear if that would reduce suffering. In contrast, the ways in which we could intervene in nature that seem very likely reduce WAS, such as by putting down old animals and providing meat-substitutes to predators, would require on-going intervention in the way we look after animals in zoos.

It’s possible technological improvements will make radical interventions in wild animal welfare feasible. Artificially intelligent flying drones could replace and extend the activities park rangers have typically filled in animal reserves, for instance delivering medication via dart guns to animals that would cure misery-causing intestinal worms or provide vaccinations. Further, if we believe the zebra’s experience of being killed in truly terrible, we could also fill such zebras with amphetamines or opiates prior to letting the lions kill it. In the (far) future, once many other problems have been solved, this may be a cost-effective way to increase happiness compared to the alternatives.

  1. Conclusion

In conclusion, I’ve argued against the idea there is more suffering than happiness in wild animal populations. Whilst there is some suffering, which gives us one reason to intervene, this isn’t sufficient show there is net wild animal suffering. Those who want to argue there is net WAS need to conduct a more careful and principled approach, trying to understand the daily lives on different animals and their balance of happiness to unhappiness. Such analysis seems fraught with anthropomorphism and problematic subjectivity. I’ve suggested habitat destruction would not obviously increase happiness (although negative utilitarians may argue it would still be good) and that cost-effective interventions for wild animal suffering do not seem to be presently available.

7. Reply from Tomasik and my response to his reply.

Tomasik has written a reply to my essay. He makes one substantive point and then picks up a number of smaller issues. I'll quote a large chunk of Tomasik and then add my thoughts.

My position on the net balance of hedonic experience in nature can be confusing.

First, I'm an anti-realist both about moral value and even about consciousness itself. I don't think there's an objective fact of the matter about how intense a given experience is, or how to compare the magnitudes of pleasures vs. pains. There are many objective measures of these things relative to some precise specifications, but it's a moral question which specification(s) we want to use. For example, one could crudely measure happiness as the number of neuronal firings per second in hedonic-hotspot regions of the brain, and suffering as some comparable measure for negative experiences, but it's not clear that we should directly sum these quantities for hedonistic-utilitarian calculations. Of course, this example is a bit of a strawman, since few hedonistic utilitarians would endorse the procedure just described. But there are countless possible more complex and nuanced measures of wellbeing, and it's up to an individual hedonistic utilitarian to decide which measure(s) she wants to optimize for.

Given that happiness and suffering aren't quantities whose meanings everyone agrees upon, it becomes unclear what the difference is between negative and non-negative forms of utilitarianism. A person who is considered by others to be a negative utilitarian may just feel that he's a regular utilitarian who uses a definition of suffering according to which suffering is vastly more intense than most people assume. In light of this problem, I think the best way to define "negative utilitarian" is just "someone who gives (much) more weight to suffering than most other utilitarians do when evaluating situations". For instance, in a world where most people only gave moral weight to happiness, Jeremy Bentham would be a (weak) negative utilitarian.

I think this explanation clarifies where Tomasik is coming from and also ends the debate. I take it the question of whether there is net unhappiness or not to be a subjective judgement of fact rather than anything to do with morality. This was the point of my analogy where a friend and I stare at buckets of water and try to judge which is more full. We're not saying more water is good, we're making a subjective judgement of fact. We still need to define happiness, but once we've done that we can conduct an arm-chair empirical analysis of how much of it there is.

Tomasik's defence actually undermines his original argument. If there is no way fact of the matter about how to compare pleasure vs pain, how could it possibly make sense for him to say there is more suffering than happiness? He's denied himself the ability to make such claims. This seems to be a motte-and-bailey approach. Tomasik's original argument was written as if there was really a psychological state called 'suffering' of which there could be more less. When that was challenged, he claim he meant suffering in some other, I think moral, rather than psychological, sense whilst also remarking that he was an anti-realist (it would have been helpful if he'd said what he meant by that). So he's now moved the argument away from the empirical debate he started and I responded to.

I'm afraid I don't follow the second paragraph to know how to respond. Presumably a classical utilitarian says that 1 unit of happiness is as morally good as 1 unit of unhappiness is bad. What composes the units seems to be a different question. The negative utilitarian is, again presumably, one who says 1 unit of happiness is less morally good that 1 unit of unhappiness is bad.

The only small point I'd mention was regarding r-selectors. I agree it was probably too quick of me to lump them together: there are lots of different types of r-selectors. Accepting this as true only makes the factual cases, either for their being net happiness or net suffering in the wild, more complicated to show.


Thanks to Frank Arntzenius and Gary O'Brien for provided comments on the original draft. Thanks for Brian Tomasik and Jacy Reese for their subsequent comments which prompted some edits.

KRINGELBACH, M.L. and BERRIDGE, K.C., 2009. Towards a functional neuroanatomy of pleasure and happiness. Trends in cognitive sciences, 13(11), pp. 479-487.

NG, Y., 1995. Towards welfare biology: Evolutionary economics of animal consciousness and suffering. Biology and Philosophy, 10(3), pp. 255-285.

ORD, T., 2013-last update, Why I am not a negative utilitarian. Available: [25/11/, 2016].

TOMASIK, B., 2015. The Importance of Wild-Animal Suffering. Relations. Beyond Anthropocentrism, 3(2), pp. 133-152.


Sorted by Click to highlight new comments since:

Thanks for the response to my reply. :) It ends the debate in the immediate sense, but there remain many things to talk about and research further. For example, we can propose different concrete measures of welfare (such as neuronal firings in hedonic hotspots) and discuss how morally compelling they seem to us. We can also do further research to understand what kinds of behaviors and brain processes wild animals exhibit in various circumstances.

Presumably a classical utilitarian says that 1 unit of happiness is as morally good as 1 unit of unhappiness is bad.

Fair enough. :) My point would then be that we need to include a footnote to say how happiness and suffering are being defined. A classical utilitarian according to definition X might be a negative utilitarian according to definition Y.

whilst also remarking that he was an anti-realist (it would have been helpful if he’d said what he meant by that).

Sorry. :P Here is one explanation.

I think I agree. There's a sense in which your choice of definition of happiness might be motivated by thoughts on morality, but once you've picked a definition, then we're trying to make factual judgements. I don't recall if you offer a definition in your orignal article.

I guess I don't discuss how happiness is to measured, although I do say what it is. I think you need to get happiness from self reports, asking somehow how good-bad they feel right now. As happiness is ultimately a subjective state, this has to be the place to start. From there, you might able to correlate that with brain activity, although I'm somewhat uncertain of how much that would really tell us.

Cool. :)

I think you need to get happiness from self reports, asking somehow how good-bad they feel right now.

How do you handle the problem of inconsistent self-reported welfare over time, such as in this example?

And there's also the standard issue of interpersonal comparisons. For instance, how do you handle a person who would not accept 1 second of torture for any amount of happiness? If that person is tortured, what welfare score does she get? Perhaps she would want to say "negative infinity", although you could prevent that by setting a lower bound on how negative her self-reported welfare could go. That arbitrary lower bound would then strongly influence the degree of weight given to suffering vs. happiness in utilitarian calculations.

WTA determined like this per species averaged over many individuals with transparency about the variance of the samples seems at least better to me than our current state of knowledge.

Definitely. :) More such studies should be done. That said, they don't resolve the fundamental issues of interpersonal comparisons and intrapersonal comparisons over time.

One of the things I find most frustrating in this area is how there are such wildly differing views on 3.1: we should be able to get some measurements on this - are there EAs working on that? Just looking at how long offspring typically starve in different varieties of k-selected species?

The first time I read this I was perplexed at the lack of a clear conclusion, but now I realize that the argument seems to be against concluding that we have reason to believe net wild animal suffering is different from zero in a negative direction without making a claim about it being positive, is that correct? So: the conclusion is that we should focus on areas where we have more certainty, and wild animal suffering is likely not one of those areas. That seems like a conclusion many EAs share who still care about WAS prima facie.

This page includes an interactive calculator that lets you plug in your own numbers for happiness and suffering for various species to compute an estimate of total welfare that accounts for lifespan and infant mortality.

As far as the painfulness of starvation, I wrote a bit here.

The painfulness of other kinds of deaths and life experiences is something that Sentience Politics wild-animal researchers will write more about in the future, and Animal Ethics also writes about such things.

A related post by Carl Shulman, arguing that in the long-run, on the face of it, happiness and suffering ought to be about equally easy to produce. There are more reasons to produce the latter though.

For a number of evolutionary and biological reasons, we can tell that the suffering of dying as an animal is much greater than the pleasure of killing it. But considering the pleasure as an interesting variable in and of itself is wrong as well. There is no reason to regard the happiness a carnivore gets from eating meat as any different from the happiness that a herbivore gets from eating plants: for all intents and purposes, they are functionally equal activities. The happiness of a predator can be expected to be similar to the base rate for ordinary large animal lives in the wild. Talking about the ordinary happiness of animal lives, and then specifically adding happiness from predation, is double counting.

I don't doubt that the zebra has a worse time than the lion has a good time. In fact, I say "I would assume that the unhappiness of the zebra, whilst it dies over 6 minutes, is more intense than the happiness of the lion of the same period."

And I don't know what you mean by 'pleasure as an interesting variable in and of itself'. I take pleasure as a synonym for happiness as state that in my definition.

I would expect the pleasure a lion gets from eating zebra is greater than the pleasure a zebra gets from eating grass for the following reason: lions literally have to chase and kill their prey. It's a lot of effort and risks injury and death to the lion if it gets kicked, etc. So you'd expect lions would need stronger incentives to overcome their fear of pain, and so on. The zebra just needs to bend its neck when it's hungry.

I don't know what you mean by "The happiness of a predator can be expected to be similar to the base rate for ordinary large animal lives in the wild". Which ordinary animals are you thinking about?

And I don't think I'm double counting. Largely because I'm not counting. I do not come to a final estimate. In BT's original essay, he focuses on the suffering of prey and says nothing about the pleasure of predators, which I think he admits should be counted too.

I don't doubt that the zebra has a worse time than the lion has a good time.

Yes, I know. But it's much greater. As in, many times greater.

And I don't know what you mean by 'pleasure as an interesting variable in and of itself'. I take pleasure as a synonym for happiness as state that in my definition.

I mean the pleasure of eating another animal, as separated from the pleasure of eating simpliciter.

I would expect the pleasure a lion gets from eating zebra is greater than the pleasure a zebra gets from eating grass for the following reason: lions literally have to chase and kill their prey. It's a lot of effort and risks injury and death to the lion if it gets kicked, etc. So you'd expect lions would need stronger incentives to overcome their fear of pain, and so on. The zebra just needs to bend its neck when it's hungry.

If lions suffer from extra effort and injuries in the process of hunting, then the net happiness of predators is going to be less, ceteris paribus. You hypothesize that the pleasure they get is more, to counterbalance it, but that just takes you back to the original point that there is no reason to view their happiness as different from that of herbivores.

I don't know what you mean by "The happiness of a predator can be expected to be similar to the base rate for ordinary large animal lives in the wild". Which ordinary animals are you thinking about?

Other animals in general. Lions equal to zebras, perhaps. (On average.) I'm saying that introducing happiness-from-predation as a separate line of analysis is flawed because it implies double counting - imagine if one made two arguments; one that we need to take into consideration all kinds of ordinary happiness in daily life, and a second that we need to take into consideration the fuzzy warmth of sleeping in the sun. Clearly the second is just part of the first, because it's just another ordinary part of how some animals are motivated in their daily lives. It's the same thing with pleasure from killing.

As we probably all agree, the pleasure per unit time for eating a zebra is probably higher than the pleasure per unit time for eating grass. (Compare the human experience of eating steak with that of eating kale.) But zebras eat grass for much longer.

Eating grass (and any food) can also involve unpleasantness if you're eating low-quality food (as well as if you're nervous about predators or conspecifics, etc.).

Thanks for the post. I doubt the length is a problem. As long as you're willing to produce quality analysis, my guess is that most of the people on this forum would be happy to read it.

My thoughts are that destruction of ecosystems is not justifiable especially because many of its effects are probably irreversible (e.g. extinction of some species), and because there is huge uncertainty about its impact. The uncertainty arises because of the points you make, and because of the shakiness of even some of the assumptions you use such as the hedonistic framework. (For example, in humans the distinction between the "experiencing" and "remembering" selves diminishes the value of this framework, and we don't know the extent to which it applies to animals.) Additional uncertainty also exists because we do not know what technological capabilities we might have in the future to reduce wild animal suffering. So almost regardless of the specifics, I believe that it would certainly be better to wait at least until we know more about animal suffering and humanity's future capabilities, before seriously considering taking the irreversible and drastic measure of destroying habitats. This might be just a different point of emphasis rather than something you didn't cover.

Hello Michael,

Yeah, I totally agree. The scope of what I was talking about was more limited. If there were clearly net WAS, we'd have to weigh up the apparent benefits of ecosystem destruction (i.e. less animal misery) against the sort of costs you're taking about.

My aim was to challenge the argument about their being net WAS. Unless there is net WAS (or you're a negative utilitarian) the case for habitat destruction looks pretty thin anyway.

FWIW, I don't think their being experienced vs remembered selves is a problematic for a hedonic framework. In fact that distinctly requires the assumption people do feel things and can rate how bad it is, and those can then be compared to their memories. That stuff is a problem for our ability to make good affective forecasts (which I admit we suck at).

Ah, you're right about the hedonistic framework. On re-reading your intro I think I meant the idea of using pleasure as a synonym for happiness and taking pain and suffering as synonyms for unhappiness. This, combined with the idea of counting minutes of pleasure vs. pain, seems to focus on just the experiencing self.

Curated and popular this week
Relevant opportunities