Factory farming

Summary

History is littered with moral mistakes — things which once were common, but we now consider clearly morally wrong, for example: human sacrifice, gladiatorial combat, public executions, witch hunts, and (of course) slavery.

In my opinion, there’s one clear candidate for the biggest moral mistake that humanity is currently making: factory farming.

The rough argument is:

  • There are trillions of farmed animals, making the scale of the potential problem so large that it’s hard to intuitively grasp.
  • The vast majority (we estimate 97.5%1) of farmed animals are in factory farms. The conditions in these farms are far worse than most people realise.
  • Even if non-human animals don’t morally matter as much as humans, there’s good evidence that they are conscious, and that they feel pain — and as a result, the poor conditions in factory farms are likely causing animals to experience severe suffering.

What’s more, we think this problem is highly neglected, and that there are clear ways to make progress — which makes factory farming a highly pressing problem overall.

Scale  

We kill around 1.6 to 4.5 trillion animals a year – which amounts to around 50 billion when adjusting for the best guess we can find at their relative moral importance to humans. Before slaughter, these animals are almost always kept in factory farming conditions. We expect this problem to worsen in the short term but we would guess that working on it doesn’t have a substantial impact on the long-term future.

Solvability  

Making progress seems moderately tractable and there are some plausible ways we could make progress.

Profile depth

In-depth 

This is one of many profiles we've written to help people find the most pressing problems they can solve with their careers. Learn more about how we compare different problems, see how we try to score them numerically, and see how this problem compares to the others we've considered so far.


We’re going to start by investigating how important we should consider ending factory farming to be — that is, how much good would be done if we solved the entire problem. We’ll look at:

We’ll then look at whether there are ways to make progress, and how neglected factory farming is, before discussing how you can help solve this problem with your career.

How many animals are in farms?

Every year, we kill somewhere between 400 billion and 3 trillion vertebrates (e.g. cows, chickens, fish) — some are killed for sport and some are dissected for experiments, but the vast majority are either slaughtered for food or die in farms before they’re old enough to be purposefully slaughtered.2

That doesn’t mean there are trillions of animals in farms at any given time, as the life-span of these animals is often less than a year. There are probably around 120–210 billion vertebrates alive in farms at any given time.

If we include invertebrates (e.g. octopuses, insects, crabs, snails, shrimp) — which is more controversial because we have less evidence of their ability to feel pain — it’s more like 1.6 trillion to 4.5 trillion farmed animals killed per year, and around 350–700 billion animals alive in farms at any given time.3

It’s much harder to estimate how many of these animals are in factory farms. In part, this is because “factory farming” doesn’t have a clear definition. Generally, we’d say something is a factory farm if it has generally poor welfare conditions and a large number of animals in a small space (this is known as high stocking density).

Rather than drawing a clear distinction, we’ll go through how different animals tend to be treated on farms and note any key variations, such as in standards between countries or what it means for hens to be “cage-free.”

We’re not going to focus on animals that aren’t farmed — wild animal welfare and animals killed by fishing and hunting are whole separate issues.4

How do we treat these animals?

Let’s go through some of the most commonly farmed animals and look at how they’re treated.

Discussing how animals are actually treated on farms can be pretty disturbing. So as we go through each animal, we’ll use the slightly clinical, euphemistic language used by the farming industry. If you’d like to read the same thing described in clearer, more plain-spoken language, we’ll put that in a collapsed box, which you can choose to expand and read.

We think it’s important to be sceptical of many of the claims of animal advocates — like anyone advocating for a cause, they have incentives to exaggerate their claims. But, when we looked into it further, it seems there’s pretty strong evidence that standard industry practices are often extremely awful. We tried hard not to cherry-pick the worst cases of animal abuse — this is just how the vast majority of animals are treated.

Chickens and birds

There are around 24 billion chickens alive in farms at any moment.

  • Around 10–16 billion of these are “broilers”, i.e. chickens bred for slaughter. They live only a few weeks to months — we slaughter around 70 billion of these chickens per year.
  • Around 7–10 billion chickens are in the egg industry.
  • Additionally, each year we slaughter around 3–8 billion male chicks from the egg industry shortly after they’re born because they won’t lay eggs, and they’re not the right breed to be cost-effectively raised for meat.

We’ve genetically selected broiler chickens to grow very fast; they may live as little as five weeks before reaching market weight) This causes ascites (cardiovascular issues) and lameness. Even in free-range farms, each bird often has only around 0.25m² of space, restricting physical activity. Prolonged contact with litter causes hock burn and footpad dermatitis. Heat stress is common due to ambient temperatures of 30 to 40°C in housing, particularly in summer months and tropical regions. Breeder chickens can live for up to a year, but experience chronic hunger in that time period. In some systems, like in Europe, 90% of chickens are stunned before slaughter; even so, live scalding and injuries during shackling are still common during slaughter.

Caged laying hens tend to have around 0.07m² of space, which prevents standing, preening, turning and wing flapping, as well as inability to nest or forage. Osteoporosis leads to frequent keel bone fractures. Withdrawal of food is sometimes used to induce moulting (and increase egg production) in countries outside Europe. Egg peritonitis is quite common, causes substantial pain, and is sometimes fatal.

The lives of cage-free hens are marginally better. They are usually kept in indoor aviaries. They have a little more space per bird, and are able to move around somewhat. Keel bone fractures are still common. Injurious pecking including vent wounds and cannibalism, are also a problem. Most hens, both caged and cage-free, are beak-trimmed. Male chicks are culled by maceration.

We’ve genetically selected chickens that are farmed for their meat to grow very fast; they may live as little as five weeks before they are killed. They grow so fast that they have small hearts and lungs — over time, low blood pressure can cause fluid to build up in the abdomen, which compresses organs, causing pain and meaning they struggle to breathe. This build up of fluid is known as ascites. These chickens’ large sizes also cause difficulty moving. Much like injured humans, chickens change the way they move in response to their weight and injury — which then causes muscle and bone pain. Even in free-range farms, each bird often has an area of space less than a single A4 sheet of paper, making it hard to move around. Difficulty moving leads to prolonged contact with excrement left by other chickens, which can be so powerful it leads to chemical burns. Heat stress, with ambient temperatures of 30 to 40°C, is common. Breeder chickens can live for up to a year, but because they’re still broiler chickens, they’re bred to want to eat a lot.Feeding breeders isn’t that helpful for farmers, so they experience chronic hunger. In some systems, like in Europe, 90% of chickens are stunned before slaughter, mostly by having their heads dunked in water with an electric current. Even using pre-slaughter stunning, many chickens experience injuries during the live shackling (e.g. leg breaks) prior to stunning when they are still conscious, and some are boiled alive due to improper shackling leading to failed stunning. In other systems, birds are shackled upside down by their legs while still conscious and defeathered in scalding water. All stunning systems will fail sometimes, e.g. by birds moving their heads aside from the electrical bath, so even supposedly-stunned chickens will sometimes be conscious when their throats are cut.

Caged laying hens also tend to have less space than an A4 sheet of paper, which prevents them from moving in any real way, so they can’t do their natural nesting or foraging behaviour. Poor diets and a lack of movement often leads to broken bones. Sometimes, in countries outside Europe, birds are starved and exposed to 24-hour bright lights to get them to shed their feathers in order to lay more eggs. Eggs that break before they are laid leave bits of yolk and shell inside the chickens causing physical pain — or, worse, bacteria-infected fluid filling a hen’s body, causing severe pain and potentially death.

The lives of cage-free hens are marginally better. They are usually kept indoors. They have a little more space per bird (although usually still less than an A4 sheet of paper), and are able to move around somewhat. Broken bones and fractures are still common. However, because of stress and the inability to socialise naturally with such a large flock, hens still end up wounding each other by pecking and sometimes even resort to cannibalism. As a result, it’s common to remove the tip of both caged and cage-free hens’ beaks using a hot blade (with no painkillers), which are highly enervated and used by birds to interact with the world ( like we use our hands). Male chicks who can’t lay eggs are killed by being thrown into a grinder.

There are many non-chicken birds we farm. In particular:

  • Ducks (around 1.1 billion alive at any one time, three billion slaughtered per year)
  • Quail (around 400–550 million alive at any one time, 1.5 billion–2.5 billion slaughtered per year)
  • Turkeys (around 450 million alive at any one time, 650 million slaughtered per year)
  • Geese and guinea fowls (around 350 million alive at any one time, 700 million slaughtered per year)

These birds face similar conditions to chickens.

Pigs

There are around one billion pigs alive in farms at any one time. We slaughter around 1.5 billion pigs per year.

Most farmed pigs are kept indoors in high density farms. Heat stress is common (pigs do not have sweat glands; they keep cool in the wild through wallowing and rooting).

Piglets go through tooth cutting and tail docking without anaesthesia. Most male piglets are also castrated (other than in the UK, Ireland, and Switzerland) to prevent boar taint.

A small proportion of pigs are kept outdoors. However, these pigs usually undergo nose ringing, which deliberately causes pain when pigs go to root in order to prevent the ground from being upturned.

The worst conditions are experienced by breeding sows. In countries other than Sweden and the UK, pregnant sows are kept in gestation crates sufficiently narrow to prevent any form of movement, even turning around. And in countries other than Sweden, Norway and Switzerland, farrowing crates are used to prevent sows being able to move away from their piglets. Piglets are weaned at around three to four weeks, after which sows are inseminated again immediately.

At slaughter, pigs are gassed, undergo electronarcosis, or undergo live-shackle slaughter.

Most farmed pigs are kept indoors for their entire lives with very little space. The buildings can reach temperatures of 30 to 40°C — which is particularly bad as pigs do not sweat; instead they root with their noses, putting their heads into the ground, or wallow in mud or water in order to cool off, which they often won’t have access to.

Soon after they are born, the sharp ends of a piglet’s teeth and the ends of piglets’ tails are cut off without the use of painkillers. Most male piglets are also castrated (with a knife or by tying a tight band around the testes until they necrotise and fall off) — other than in the UK, Ireland and Switzerland — in order to preserve the flavour of the meat (preventing ‘boar taint’).

A small proportion of pigs are kept outdoors. However, these pigs usually have a piece of metal or wire inserted into their nose, which deliberately causes pain when pigs try to dig in the soil, in order to prevent the ground from being upturned.

The worst conditions are experienced by sows kept for breeding. In countries other than Sweden and the UK, pregnant sows are kept in gestation crates sufficiently narrow to prevent any form of movement, even turning around. Pigs will often injure themselves through chewing or headbutting the cages. And in countries other than Sweden, Norway, and Switzerland, ‘farrowing crates’ are used to stop sows from moving away from breastfeeding piglets. Sows not kept in farrowing crates will often crush or cannibalise their piglets (through accident or stress). Piglets stop drinking milk from their mothers at around three to four weeks, after which sows are inseminated by farmers again immediately.

When pigs are killed, they are usually gassed or shocked to make them unconscious (but it only works some of the time), after which their throats are cut. Gassing seems to be very painful. Often this occurs after a prolonged period of being shackled upside-down by one leg while they are conscious.

Cows and veal

There are around 1.5 billion cattle alive in farms at any one moment. We slaughter around 300 million per year.

Generally cattle are treated better than chickens and pigs, with 5 to 10 square metres of space per cow.

The main issues are:

  • Many cows are kept indoors for at least half of the year, and — for beef cattle — for 80–300 days for fattening before they are killed. Indoor systems are cramped, and slatted floors lead to injuries and lameness.
  • Both male and female cows are dehorned, often without anaesthesia.
  • Lameness and mastitis are common causes of pain in dairy cows.
  • Dairy cows give birth annually. Calves of dairy cows are separated from their mothers almost immediately, causing a negative emotional state.
  • Male calves from the dairy industry are often slaughtered at or shortly after birth. Veal calves are fed an iron-restricted diet to keep their flesh pale. Outside the EU, veal are raised in crates that prevent movement for their entire lives.
    • Few slaughterhouses deal with spent dairy calves, leading to long and stressful transportation to slaughter.
  • Stunning at slaughter sometimes goes wrong, and there are often poor conditions at slaughterhouses.

The main issues are:

  • Many cows are kept indoors for at least half of the year, and — for beef cattle — for fattening before slaughter. Indoor systems have around five square metres per cow. The floors have gaps in them to allow excrement to fall through, but cows get their feet stuck in the gaps and injure themselves.
  • Young cows have their growing horns (“buds”) burned off using a hot iron, often without anaesthesia.
  • Dairy cows often end up with diseased hoofs and injured legs, making it painful to move around. This is often left untreated. Also, excessive milk production causes cows’ udders to become infected and painful.
  • Dairy cows are kept pregnant to ensure constant milk production. This creates an issue: dealing with the unwanted calves. All male calves and almost all female calves are separated from their mothers almost immediately — both calves and mothers cry for each other for days to weeks afterwards.
  • Male calves from the dairy industry are often killed by being shot with a gun shortly after they’re born. This is because dairy cows and meat cows are different breeds bred for different traits. If they are kept for meat, it’s usually veal (i.e. young cow meat). Veal calves are fed an iron-restricted diet to keep their flesh pale. Outside the EU, veal calves are raised in single-occupancy crates no bigger than the calves: these prevent almost all socialisation and movement for their entire (short) lives.
    • Dairy cows are slaughtered after 4 to 6 years, and often travel long distances in cramped and hot conditions to slaughterhouses. Many die on the journey.
  • Many slaughterhouses employ electrical prods to keep live cattle moving, and — after stunning via captive bolt gun — hang cows upside down by one leg while their throats are cut and their trachea removed. A reasonable proportion of cattle are slaughtered without any stunning at all, and some of these slaughter cattle via inversion and bleeding them out.

Sheep and goats

There are around 2.2 billion sheep and goats alive in farms at any one moment. We slaughter around 1.05 billion per year.

Sheep and goats also tend to be treated better than chickens and pigs.

The main issues are:

  • Castration and tail-docking, usually done without anaesthesia
  • Lack of outdoor access for lambs (unless the label says ‘grass fed’ or ‘access to pasture’)
  • Separation of lambs and ewes
  • Transport in hot and overcrowded trucks without food or water
  • Failed stunning and poor conditions at slaughterhouses

Fish

There are around 100 billion to 180 billion fish alive in farms at any one moment. We slaughter around 100 billion farmed fish per year (many more fish are wild-caught).

We’re more uncertain what produces a poor welfare environment for fish, and there are so many different species of fish that it’s difficult to say much about their welfare in general. For example, it seems like some fish prefer high stocking-density environments. (There are a variety of reasons for this, including that high stocking density can prevent territorial and aggressive behaviour). Other fish species seem to prefer more turbid (murky, dirty-looking) water to clean water.

That said, there are almost no legal protections for fish welfare. Somewhere between 15 billion and one trillion fish die in farms before reaching slaughter age, and that extremely high pre-slaughter mortality rate is a bad sign for their general welfare.

Many issues occur at slaughter. Fish are rarely stunned before slaughter. Common slaughter methods include asphyxiation in air, carbon dioxide narcosis, live chilling or gill-cutting.

Common issues in fish farms include:

  • Poor water quality, including insufficient dissolved oxygen; inappropriate temperature, pH, and salinity; and build-ups of ammonia and nitrates from waste
  • Disease and parasites, such as sea lice
  • Social stresses which lead to injury, fin erosion, and cannibalism

Since fish are smaller and grow slower than most land-based farm animals, they live longer lives and more fish are killed for the same amount of food. What’s more, unlike herbivorous land-based farm animals, predatory fish such as tuna and salmon (and the lesser-known mandarin fish) must be fed other fish, often live, increasing the welfare concerns per animal sold.

Common slaughter methods include:

  • Leaving fish in air for several hours to slowly suffocate to death or be crushed by the weight of other fish being poured in on top.
  • Putting fish in baths of ice slurry where they gradually lose consciousness. Ideally, fish are left until they die from lack of oxygen, before they’re cut open and drained of blood. Sometimes, if fish aren’t left in the ice slurry long enough, or if there isn’t the right ratio of ice to water and fish, they can recover and regain brain function as they warm up, leaving them conscious when they’re cut open.
  • Bubbling carbon dioxide into the fish’s water to make it gradually acidic, which eventually kills them. Fish will swim vigorously and try to escape the tank when this is done. Again, if fish are removed too early or the water doesn’t become acidic enough, they’ll be conscious when they are cut open.
  • Simply cutting off the gills of a fish, causing them to bleed out and suffocate to death.

Other vertebrates

People across the world eat a wide variety of other vertebrates, but a few deserve particular mention because we farm more than 100 million of them at any one time:

  • Frogs (around 120 million to 2.6 billion alive at any one time, 300 million to one billion slaughtered per year)
  • Chinese softshell turtles (around 250 million to 2.1 billion alive at any one time, 180 million to 900 million slaughtered per year)

Both frogs and turtles are kept in high-density indoor farms. They’re commonly slaughtered by being cut open without stunning.

  • Frogs’ legs (the main part of the frog that’s eaten) are cut off while the frogs are still alive.
  • Turtles are killed by ‘de-carapacing’ (removal of internal organs while conscious).

Shrimp and other decapod crustaceans

We kill around 255 to 605 billion farmed decapod crustaceans for food each year. That includes:

  • Crabs (5 to 16 billion slaughtered per year)
  • Crayfish and lobsters (37 to 60 billion slaughtered per year)
  • Shrimp (213 to 530 billion slaughtered per year)

The main welfare issues for shrimp (the vast majority of farmed decapod crustaceans) are poor water quality, eyestalk ablation (i.e. removal of one or both eyestalks), and slaughter by asphyxiation (or by being crushed by other shrimp).

Other invertebrates

As mentioned above, including invertebrates substantially increases the number of animals in farms. In particular, we farm:5

  • Snails (2.9 billion to 7.7 billion slaughtered per year)
  • Crickets (34 billion to 41 billion alive at any one time, 370 billion to 430 billion slaughtered, sold live, or die before slaughter per year)
  • Mealworms (25 billion to 31 billion alive at any one time, 290 billion to 310 billion slaughtered, sold live, or die before slaughter per year)
  • Black soldier flies (8.1 billion to 16 billion alive at any one time, 190 billion to 300 billion slaughtered, sold live, or die before slaughter per year)

We’re very uncertain about what produces a poor welfare environment for these animals. We’d guess the major issues are: black soldier flies aren’t fed once they become adults, causing death by starvation; larvae are photophobic but kept under lights; shipping causes vibrations causing frequent death from stress.

Other issues likely include: high-density farming, disease, high rates of death before slaughter and difficulty providing anaesthesia or stunning, and difficulty identifying and helping with specific injuries and diseases.

To what extent do animals deserve our moral consideration?

On reading the above, it’s pretty clear that if even some animals deserve our moral consideration, something pretty awful is happening. But to what extent some or all animals deserve our moral consideration is, in fact, a subject of substantial debate.

We’re going to look at:

  1. Whether animals are conscious and whether they feel pain.
  2. What various ethical theories say about who (and which beings) deserve our moral consideration — we’ll consider a wellbeing-based approach, as well as other moral theories.
  3. Based on these discussions, we’ll try to give some estimates of how many animals it’d be reasonable to count morally the same as one human.

Overall, we find that while there’s disagreement on these issues, and grappling with this moral uncertainty is tricky, we think there’s enough agreement on the science and among plausible moral theories that the correct approach is to give many species of non-human animals — including the ones we keep on factory farms — at least some non-negligible moral consideration.

Animal consciousness and pain

Many of the moral arguments we’ll consider below come down to whether animals are conscious and whether they have the capacity to feel pain.

The question of consciousness remains highly philosophical. By consciousness, we mean phenomenal consciousness (or capable of subjective experience) — that is, we mean that there is “something it is like” to be a certain animal in the same way there’s “something it’s like” for you to experience a sunset or a good mood.

This is a particularly difficult question for non-human animals. The main ways we get evidence about the consciousness of other humans is communicating via language and by analogy to our own minds. But we can’t clearly communicate with non-human animals, and their minds are less similar to our own, so it’s harder to argue by analogy.

But there is plausibly a way to make progress: by looking at the various philosophical theories of consciousness, we can try to identify “potentially conscious-indicating features”,6 and then look for these features in animals.

A 2017 report by researcher Luke Muehlhauser identified over 40 features which are indicative of consciousness.7 They fall into a few categories:

  1. Broad similarity with humans. There are, of course, many ways to measure this. Following Muehlhauser, we’ll look at the time since the closest common ancestor with humans. The other three categories in this list can also be seen as measures of similarity with humans.
  2. Neurobiological features, in particular having a large brain and a centralised, complex nervous system.
  3. Clear and complex nociception — the physical states and behaviours associated with feeling pain in humans. For example, having fibres that respond to dangerous stimuli (known as “nociceptors”), protective behaviour like wound guarding or limping, responding to painkillers, and learning to avoid things that cause nociceptors to fire.
  4. Other indicators of cognitive ability, like play behaviour8, grief behaviour, mirror self-recognition, tool use, language capabilities, or theory of mind. While the relationship between cognitive ability and phenomenal consciousness is not well understood, many philosophers and scientists think there’s likely an association between the two.9

So let’s look at some of the animals we keep in factory farms and assess them on some of these criteria.

We’ll pay particular attention to nociception, because it’s both an indicator of consciousness, and also — as we’ll see below — plausibly an indication of moral status in itself.10

Unfortunately, there’s still a lot we don’t know in this area (even things like how many neurons are in the brains of various animals). So there are lots of gaps in the table below — even among the small number of indicators we’ve decided to focus on.


Humans
Brain and nervous systemAvg. adult brain mass: 1300g36

Neurons in brain: 85 billion37

Have a neocortex and a central nervous system

Nociception38Neural nociceptors and reflexes to move away from dangerous stimuli

Nociceptors respond to painkillers

Self-administer painkillers if needed (and even pay a cost to do so)

Exhibit protective behaviour (e.g. wound guarding, limping)

Other indicators of cognitive ability38Play behaviours and grief behaviours

Clear tool use

Recognise themselves in mirrors39

Can move around obstacles to reach a goal (even when they temporarily can’t sense the goal)40

A range of other cognitive abilities not observed in other animals which people have argued indicate consciousness, such as abstract language capabilities and the ability to do mental time travel

Chickens
Time since closest ancestor with humans41310 million years ago
Brain and nervous systemAvg. adult brain mass: 3.5g42

Neurons in brain: 220 million43

May have a neocortex-like structure44 and have a central nervous system

Nociception38Neural nociceptors45 and reflexes to move away from dangerous stimuli46

Nociceptors respond to painkillers

Self-administer painkillers if needed (no studies were found on whether they would pay a cost to do so).47

Exhibit protective behaviour48

Other indicators of cognitive ability38Play behaviours49

Grief behaviours have been claimed by some backyard chicken owners,50 but we found no studies confirming this

No clear tool use51

Might recognise themselves in mirrors (although don’t pass the conventional mirror test)52

Can move around obstacles to reach a goal (even when they temporarily can’t sense the goal)53

Pigs
Time since closest ancestor with humans4195 million years ago
Brain and nervous systemAvg. adult brain mass: 135g54

Neurons in the brain: 430 million55

Have a neocortex56 and a central nervous system

Nociception38Have neural nociceptors57 and reflexes to move away from dangerous stimuli58

Nociceptors respond to painkillers59, but no studies were found on self-administration

Exhibit protective behaviour.60

Other indicators of cognitive ability38Play behaviours,61 and likely experience grief (but we’re not sure)62

Tool use.63

Do not pass the mirror test.64

Can move around obstacles to reach a goal (even when they temporarily can’t sense the goal)65

Fish
Time since closest ancestor with humans41400 million years ago (lungfish) to 530 million years ago (jawless fish)
Brain and nervous systemAvg. adult brain mass: 0.2g (rainbow trout), 0.3g (carp), 1g (salmon)66

Total number of neurons: 10 million (zebrafish)67

Don’t have a neocortex,68 but do have a central nervous system

Nociception38Have neural nociceptors and reflexes to move away from dangerous stimuli69

Nociceptors respond to painkillers. They may self-administer painkillers if needed (and pay a cost to do so), but the evidence is unclear (zebrafish)70

Exhibit protective behaviour (carp, zebrafish, rainbow trout)71

Other indicators of cognitive ability38Likely play behaviours,72 but no clear grief behaviours.

Tool use (wrasse, cichlids, whitetail majors).73

Cleaner wrasse pass the mirror test.74

Can move around Likely play behaviours,72 but no clear grief behaviours

Tool use (wrasse, cichlids, whitetail majors)73

Cleaner wrasse pass the mirror test74

Can move around obstacles to reach a goal (even when they temporarily can’t sense the goal) (zebrafish, goldfish) 75

Insects
Time since closest ancestor with humans41700 million years ago
Brain and nervous systemAvg. adult brain mass: 3mg (bees), 1.6mg (*Apoidea*), 0.6mg (black soldier flies), 0.4 to 0.06mg (butterflies)76

Total number of neurons: 600,000 (western honey bee), 300,000 (black soldier fly), 100,000 (fruit fly)77

Have a central nervous system but nothing clearly resembling a neocortex

Nociception38Have neural nociceptors, and reflexes to move away from dangerous stimuli

Nociceptors respond to painkillers78

Exhibit protective behaviour (bumblebees)79

Other indicators of cognitive ability38Most indicators have not been studied 

Some complex behaviours have been found, including: 
- moving around objects in complex ways (bees)80
- pessimism (bees),81
- learned helplessness and response to antidepressants  (fruit flies)82
- something a bit like tool use or play behaviour (bees)83
- anxious-avoidant behaviour which responds to anti-anxiety drugs (fruit flies)84

Each indicator that’s present for some species is clearly some evidence for its consciousness; similarly each indicator that’s missing is clearly some evidence against.

But it’s not really clear how much evidence each indicator is — without a decisive underlying theory, there’s no clear way to interpret the data. And this table is incomplete; so much so that evidence not included is probably more significant than the evidence presented! There have been two more comprehensive attempts that we know of to look at many more possible indicators, and a wider variety of species (in particular, we are missing decapod crustaceans and cows from the above table),11 but these are also still very incomplete.

And then there are other issues with these indicators:12

  • Many systems typically thought to be non-conscious (for example the spinal cord, plants, bacteria, and some computer programs) have at least some of the above features.
  • Many of the behaviours discussed can also occur unconsciously in humans, which, unless humans have conscious subsystems, suggests they’re not great evidence for consciousness.
    And lots of the information might be mistaken — much of this is based on single studies, many of which may fail to replicate.

That said, if you’re anything like me, you probably find the extent to which animals express complex behaviours pretty surprising — which suggests that we should be more willing to consider the idea that they could be conscious. In fact, it seems like we are generally finding, over time, that sophisticated behaviours are more common than we once thought.

There may be other ways of making progress here, but we haven’t found any particularly illuminating.13

Overall, there are three conclusions we can reach:

  • It seems very hard to rule out that the animals we keep on farms are conscious.
  • It seems even harder to rule out the existence of nociception in these species.
  • The evidence varies from species to species, and seems stronger for animals like pigs and chickens than for fish or insects.

It seems to us that, given the evidence we currently have and the huge uncertainty, it’s hard to justify thinking that the probability of consciousness for any factory farmed animal is less than, say, 5%.14 And for some animals (chickens, sheep, goats, cows, pigs), I think the chances of consciousness seem above 80%.15

What makes something deserving of moral consideration?

So, if some species of animals are conscious and have nociceptors (pain responses) or emotion, what does that imply about their moral status?

We’ve argued previously that a core piece of doing good is about promoting wellbeing, considered impartially. In short, this is because when it comes down to it, most moral theories agree that:

  • If you can make others better off — i.e. have more wellbeing — that’s a good thing to do.
  • It’s even better to make more individuals better off than fewer.
  • This is true equally, no matter who those individuals are (as long as, and to the extent that, they are capable of wellbeing).

(Read about this argument in more detail.)

So we can then consider whether each species has something that constitutes “wellbeing.”

There are three broad theories of wellbeing:

  • The hedonistic view, where what matters is the degree of positive vs negative experienced mental states (e.g. happiness vs pain).
  • The preference satisfaction view, where wellbeing means getting what you (perhaps most deeply) want.
  • Objective list theories, where wellbeing consists of achieving things on a list of goods, e.g. like friendship, knowledge, happiness, love, virtue, wisdom and health.

Overall, the hedonistic view appears to place the most confidence in animals having the capacity for wellbeing — it seems clear that if an animal is conscious and has pain responses or emotion, then they can have positive or negative experienced mental states.

It’s less obvious that animals have preferences — nailing down exactly what philosophers mean by preferences is really difficult — but that many animals act to avoid pain is at least suggestive of the concept of preferences. (That said, some preference utilitarians argue that preferences require language, e.g. R. G. Frey.)

It’s also very unclear the extent to which animals have the capacity for virtue or wisdom — so animals might be missing the capacity for many of the items on objective list theories of wellbeing. But in practice, most people (including philosophers of wellbeing) would include “lack of pain” and “happiness or pleasure” as an important good to obtain on any ‘objective list.’

This suggests that on every mainstream view, there’s an argument that animals are capable of wellbeing, though it’s especially clear in the hedonistic case.

There’s lots more to discuss here, and many more objections and counter-objections to consider. For a more in-depth discussion, we’d recommend Comparisons of Capacity for Welfare and Moral Status Across Species by Jason Shukraft.

Different moral approaches

Many people believe some version of “welfarism” — the view that the welfare of individuals is in fact the only source of moral value.

If welfarism is right, the above analysis suggests that animals do deserve our moral consideration, since they seem likely to be conscious and to be capable of welfare and suffering.

There are good arguments for welfarism, but we think it makes sense not to be too confident in any one moral theory — so it’s worth considering what other views have to say about the moral standing of animals.

Do any other popular theories suggest that, even if factory farming is large in scale, solvable and neglected, we should nevertheless not consider it to be a particularly pressing problem because animals shouldn’t be given much (or any) moral consideration?

There is too much here to give a comprehensive overview, but here areshort descriptions of some of the most well-known theories and what they have to say about animals:

  • Kantian personhood. Kant argues that fundamental moral value comes from being a rational being — and that only humans meet his criteria for being rational. In recent years, philosopher Christine Korsgaard has developed this idea into the claim that humans face a unique “problem of normativity” as a result of our ability to think about our desires. One common critique of this approach is the problem of “marginal cases”: many humans, like babies and people in comas, don’t have the capacity for rationality or self-reflection in the sense required by Kant and Korsgaard, but do seem to deserve moral consideration. Korsgaard rejects this critique, but nevertheless concludes that animals should be treated as “ends in themselves” under a Kantian view of ethics. (Read more on Kantian personhood.)
  • Contractarianism similarly only grants moral status to rational agents that can comprehend the idea of a social contract. Mark Rowlands argues that some forms of contractarianism do grant status to animals, because behind a veil of ignorance (where the moral choice is the one you would make if you didn’t know your ethnicity, gender, social status, or species), one would support the moral consideration of animals. That said, it’s common to interpret contractualism as excluding animals from moral concern because in a thought experiment of forming a social contract, they’re not sophisticated enough to participate. However, one prominent defender of a related view, T. M. Scanlon, notes that his version of contractualism only accounts for one domain of morality (what we owe to other people) — which doesn’t rule out that other domains (such as obligations with regard to the environment) could include reasons we have to respect the interests of animals.
  • Rights-based approaches. Many philosophers have argued for and against the concepts of animal rights. Tom Regan argues that all mammals of at least one year in age have rights, because they have certain cognitive abilities like perception, memory, sense of the future, and capacity for pain. Gary Francione argues that animals have the right not to be owned, and that just that one right is sufficient for a huge change to our practices. Critics, like Carl Cohen, argue that rights holders must have the capacity to comprehend rules of duty. It’s also common to think that rights are an important part of morality and that protecting the welfare of individuals is good too, even if it’s less important. So a rights-based approach seems to not lean clearly one way or another.
  • Capabilities approach. This approach, developed by Martha Nussbaum and Amartya Sen, suggests that a just society arises if people are given the chance to flourish and achieve things like health, safety, and social relationships. Nussbaum argues that this approach requires us to also let animals achieve, at the very least, a lack of suffering. (Read more in her book, Justice for Animals.)
  • Virtue ethics. Virtue ethics approaches tend to think that compassion and care for animals are virtues, but disagree on the extent to which this implies actions like refraining from factory farming.
  • Even if animals have moral standing, there are views of what it means for some state of the world to be better than another that might suggest that factory farming isn’t such a pressing concern —in particular, person-affecting views. These views argue that we only have moral obligations to help those who are already alive, not to enable more people to exist with good lives. Since the vast majority of animals on farms live very short lives (usually less than a year), most of the problem of factory farming arises with regard to the treatment of animals who aren’t yet born. As a result, even if we do give moral consideration to animals, the scale of the problem is substantially reduced under these person affecting views, which would make it much less pressing overall. That said, philosophers who defend person affecting views often think there’s an asymmetry where, while it’s not good to bring someone with a good life into existence, it is bad to bring someone into existence just to suffer. If that’s right, the problem of factory likely remains very large. (Read more about person-affecting views).
  • Finally, many “common sense” approaches to ethics suggest that many of the standard industry practices in factory farming (described above) are wrong, even if they don’t have a considered view about animals’ moral standing.

So it’s not the case that no matter what your view of morality, factory farming is an especially pressing issue. However, many moral theories put at least some weight on animals (often as a result of their consciousness and capacity to feel pain, and sometimes as a result of apparent preferences or capacity for flourishing). Since we’re philosophically uncertain about the nature of morality or the moral standing of animals, we don’t think we can rule out the basic argument above that animals do have moral standing, making factory farming a large-scale problem.

You also don’t need to treat every moral theory as having a vote in your deliberation — an alternative approach to moral uncertainty is to think about how confident you are in a given moral theory and take an expected choiceworthiness of an action.

For us, both approaches point to giving at least some non-negligible moral weight to animals, and plausibly quite a lot — though your views could be sufficiently different that you disagree.

How many animals count morally the same as one human?

Even if animals deserve our moral consideration, it doesn’t mean that we should count animals the same way we count humans.

Counting the number of individuals

Counting the number of people affected is usually important for how we compare the effects of actions that only affect humans — an action that affects two humans is considered twice as important as one that affects only one human (all else equal).

If moral status doesn’t have degrees (i.e. something is either worthy of consideration or not), a view argued for by philosophers such as Kant, Regan, and Francione, then this sounds like a reasonable way to count animals too, as long as the animals meet the above criteria for moral consideration.

However, this leads to some strange conclusions, like an overriding consideration for the welfare of insects because they are so numerous.

Counting the number of neurons

Perhaps we should reject the idea that moral status doesn’t have degrees and instead ask how much moral status each animal has. One common way of doing this is assuming each individual has some fixed capacity for wellbeing in fixed proportion to the number of neurons they have.16

This substantially changes how we compare animals to humans.17

Total populationNeurons per animalTotal neurons across the population
Humans8 billion85 billion700 quintillion
Chickens24 billion220 million5 quintillion
Pigs1 billion430 million0.5 quintillion
Cows1.5 billion6 billion9 quintillion
Farmed fish100 billion to 180 billion10 million1 to 2 quintillion

While this method is simple, and tempting, we think there are lots of good objections to using solely neuron counts as proxies to moral weight.

These objections focus on showing that while neuron count probably does correlate with the amount of moral consideration you should give an animal, these correlations are very imperfect.

For example, there are a large number of studies showing that decreased brain volume can increase the intensity of experiences (like chronic pain) in humans.18

As a result, we can’t conclude much just from neuron counts, other than that we should probably give somewhat more consideration to humans than to non-human animals. The extent of that additional moral consideration remains unclear — and neuron counts seem a poor proxy.

Using welfare ranges

The most recent and most comprehensive approach to comparing animals to humans was pioneered by Jason Schukraft and Bob Fischer, two philosophers working for Rethink Priorities’ Moral Weight Project.

They try to assess animal’s capacity for welfare, which they split into two components:

  • Their lifespan — if animals live longer, they will experience more.
  • Their welfare range — how much capacity they have for welfare in each moment.

To assess these welfare ranges, Rethink Priorities got together a team of three philosophers, two comparative psychologists, two fish welfare researchers, two entomologists, an animal welfare scientist, and a veterinarian. They listed 82 different possible proxies for welfare ranges (things like animals’ response to painkillers, the mirror test, or finding depression-like behaviour), and investigated them for a range of animals.

Next, they aggregated this information using 12 different assumptions, for example:19

  • Going by neuron count alone
  • Using multiple different quantitative measures (such as brain mass to body mass ratio and the maximum number of nociceptor spikes per second), and then taking a weighted average of the ratio between these values and the values in humans
  • Looking at the proportion of the qualitative proxies each animal has compared to humans (humans have all the proxies)
  • Looking at just certain subsets of the proxies or using quantitative adjustments that exaggerate the differences between species

In the end, they came up with the following welfare ranges, which are very uncertain but tell us more than anything else we’ve seen on the topic (these numbers are also adjusted for the chances that, in their view, each species is deserving of moral consideration at all20):

Species5th percentile50th percentile95th percentile
Humans111
Pigs0.0050.5151.031
Chickens0.0020.3320.869
Octopuses0.0040.2131.471
Carp00.0890.568
Bees00.0710.461
Salmon00.0560.513
Crayfish00.0380.491
Shrimp00.0311.149
Crabs00.0230.414
Black Soldier Flies00.0130.196
Silkworms00.0020.073

These numbers are surprisingly high – chickens have only a 3x smaller welfare range compared to a human (compared to 1000 times smaller when using the neuron count method). Even insects, like black soldier flies, have only approximately 100 times smaller ranges. (Rethink Priorities have argued that this result shouldn’t be surprising.)

Overall, these results are hugely uncertain. But it’s highly plausible that a good way of measuring the moral importance of things that affect non-human animals is by using these welfare ranges.

When we’re dealing with these kinds of uncertainty, we think it makes sense to consider the expected value of our actions — and that means we should probably think that the effects of our actions on animals are a highly important moral consideration.

How might factory farming change in the future?

Here’s a graph of the number of land animals being slaughtered each year:

The number of animals being slaughtered is increasing. This is also true for fish farming:

It seems likely that trends will be similar for shellfish and other invertebrates (there’s a nascent and growing insect farming industry.21

Ultimately, we’d guess this trend is primarily caused by the human population getting bigger and richer, increasing the demand for meat.

The number of animals on factory farms isn’t increasing as much in developed countries. This is probably primarily because their populations aren’t growing as much. It also seems like as income per capita increases, meat consumption per capita increases — but possibly this is only up to a point. In a few countries, meat consumption per capita is now falling.22 We’re not quite sure why this is, but we’d guess that, in part, as people get rich enough, they start being more willing to do things like change their diets in response to climate, health, or animal welfare concerns.

But overall, it looks like meat consumption — particularly from fish and chickens — is going to continue to substantially increase in the short-term.

We think roughly all of this increased demand will be met by factory farming — as opposed to bucolic family farms — since it’s by far the cheapest way to produce meat, and probably the only way to produce meat at the scale being demanded without turning most of the planet into grassland.

This could be changed by regulations — as we mentioned above, some countries have passed laws to protect the welfare of some farmed animals – but these are far from universal, are poorly enforced, and usually are only small initial steps.

Looking further ahead, meat consumption might stop rapidly increasing if much of the world becomes wealthy enough that average meat consumption per capita stabilises, and the global population stops increasing — which the UN projects will happen around 2080:23

But if you’re looking that far into the future, all the way to 2080, there are so many things that might change about factory farming by then — not just the number of humans who can buy meat. Society’s values might have changed hugely. There might have been a major catastrophe, like a nuclear war or a major pandemic. And perhaps most importantly, technology might completely change the picture.

How will factory farming be affected by technological advances?

We asked some experts what technological advances might affect factory farming in the future. Here are the main things they mentioned:

  • Widespread use of in ovo-sexing could mean farmers hatch only female chickens, preventing the culling of male chicks.
  • R&D in aquaculture and insect farming will likely reduce their costs, increase intensification and support the expansion of these industries. For example, we might see more productive fish and insect breeds, possibly with worse welfare for many (for the same reasons we see worse welfare for chickens that are more productive breeds).
  • We may develop uses for factory farming other than food for humans or feed for other animals — for example farming pigs for organ transplants.
  • We might use gene-editing of animals to enhance either welfare or productivity (potentially at the expense of welfare).
  • Increasing use of AI to make management decisions on factory farms could make farms more intensive and lower welfare, or could increase care and health for animals, depending on the details.
  • We might make big strides in plant-based and cultivated (i.e. lab-grown) meat alternatives — which could displace factory-farmed meat if they’re able to compete on the market, meaning fewer animals are farmed.

The two developments that seem most likely to completely change the picture are the development of price, taste, and convenience-competitive alternatives to animal products and the development of transformative AI. We’ll look at each of these in more detail below.

When might we develop really good alternatives to meat?

The development of a really good alternative to meat — one that is competitive with meat on price, taste and convenience grounds — would be a game-changer for factory farming.

We’d guess that, if there were more competetitive alternative products, it would be much easier to convince people to stop eating factory-farmed meat — and as a result, to push for increased welfare on farms.24

There are three broad approaches to creating alternatives to animal products:

  • Plant-based alternatives, made solely using ingredients from plants and fungi
  • Fermentation methods (also known as acellular agriculture), which use microorganisms to create proteins usually found in animal products
  • Cultivated meat (also known as cellular agriculture), which produces a product cellularly identical to animal meat through cultivating animal cells directly

Each of these approaches faces major challenges.

Plant-based and fermentation methods face difficult challenges in reaching parity with animal products on taste, texture, and on getting plant-based meat to cook like meat.25 Fermentation methods produce individual proteins rather than complete products, but combining these animal proteins that aren’t available in plants with plant-based products may overcome some of these challenges.

Ultimately, we’d guess that as long as there are substantial chemical differences between meat and its alternatives, there will be reasons not to switch. For example, people may have worries about health or chemical additives, or feel put off by differences in what the product looks like, feels like, or how it cooks. This is one reason why people are working on cultivated meat.

A 2020 report by David Humbird into cultured meat sparked controversy with its claims that there are severe technical challenges to growing cultivated meat at scale. Humbird’s arguments include:

  • The low growth rate of animal cells (especially compared to possible infectious microbial cells) means producers need to take expensive measures to ensure a sterile environment.
  • It’s difficult to transfer oxygen into, and waste out of, a bioreactor without rupturing animal cells, and this will limit the volume (and thus cost-efficiency) of bioreactors.
  • Scaling up the industry producing the raw ingredients (amino acids and/or plant protein hydrolysates) will be extremely expensive.

We read the full report, along with some responses to it (by the Good Food Institute and by Rethink Priorities) — and ultimately were reasonably convinced that these challenges are real.

That doesn’t mean that cultivated meat is a doomed project — it just means that without something radically changing (e.g. technological progress substantially speeding up in general, it might take a long time until we can use a competitive meat alternative to end factory farming.

Would a competitive meat alternative end factory farming, and if so — when?

There’s more to consider than just the technology. If we have the technology for products that are price, taste, and convenience-competitive with meat, would that lead to the end of factory farming? If so, how long would that take?

It’s pretty unclear.

First, it’s important to remember that making a real dent in meat consumption doesn’t just require developing the technology — it also requires building up new infrastructure. We currently consume around 1 million tons of meat every single day. Building up a replacement for that would be a gigantic project. Humbird estimates that it would cost hundreds of millions of dollars to build a facility to replace less than one-ten-thousandth of the world’s meat.26

There are also lots of other factors that might affect whether people will switch, like social norms, familiarity, food safety, and religion.27 These factors are much harder to predict — for example, it’s possible that the development of alternatives to animal products causes political polarisation and worse animal policy overall (although we’d guess that’s less likely than the opposite).

So, what does all this mean for the future of meat production?

It’s pretty hard to bring this information together into a meaningful prediction. We’ve seen two attempts to do this.

On Metaculus, a forecasting website, the aggregate of 268 predictions suggests there’s a 12% chance of a 50% decline in global meat production by 2040.

Rethink Priorities brought together a panel of forecasters to predict how likely it is we’ll have specifically cultivated meat by 2051. Aggregating their estimates suggests that, while there’s an around 50% chance of us producing over 100,000 tons per year by then, there’s more like a 9% chance of us producing over 50 million tons per year. (For comparison, we produced around 350 million tons of meat in 2022.)

But forecasting is really difficult when there are no clearly similar events to point to and when you’re trying to predict low-probability events far into the future, so it’s not clear what we should take away from this.

Perhaps more importantly, the estimates from Rethink Priorities were made assuming that there won’t be some kind of transformative AI (and we’d guess that transformative AI wasn’t on the minds of the Metaculus forecasters either). And AI could massively change the picture.

How could AI affect factory farming?

As we’ve argued, we think there’s a decent chance that we’ll develop AI systems that are hugely transformative for society within the next few decades.

We think there’s some chance that we could lose control of AI systems. But there’s also a good chance we don’t, and that developers are able to make AI systems that reliably do what their controllers want. If that’s the case, even fairly autonomous AI systems will do roughly what humans want them to do — and so the way any AI system treats animals will be guided by the users’ incentives, beliefs, and attitudes towards animals.

We’ll probably find ways of using AI to make factory farming more efficient through the use of precision livestock farming — we’d guess this is negative overall (because it might lead to increased stocking densities, and it would reduce the price of meat, making it harder for alternative products to compete). But there could be positives too, such as reducing rates of disease. Animal advocates might be able to use AI systems to analyse CCTV footage from slaughterhouses to identify and prosecute violations. (For more on how AI might affect animals in the short-term, see this article from Max Taylor, a researcher at Animal Charity Evaluators.)

We’ll likely see these effects just with AI systems that are not too far from the current state of the art.

But how might the development of truly transformative AI — for example, AI systems that can automate a very large proportion of the things that humans can do — affect the picture?

Such systems would have huge economic effects. It’s unclear exactly what these will be, but a simple model is that they would accelerate current economic trends and make society much richer — so we might see an acceleration of the trends we discussed above: per-capita meat consumption rising, peaking, and then declining as wealth increases.

More concretely, if AI systems are able to automate parts of the R&D processes, we’d see an acceleration of all the technological advances discussed above — in particular, the development of price, taste and convenience-competitive alternatives to animal products. As a result, we’d guess it’s highly likely that we’ll develop these alternatives this century.

This wouldn’t necessarily bring the end of factory farming. We’d probably also see technological advances that reduce the price of meat, so there could be a long period of time where meat and alternatives are competing for customers. And as we discussed above, it could take decades to build the necessary infrastructure to replace animal agriculture. And there would still be social barriers preventing the end of factory farming — for example, we could end up in a situation where a large portion of society sticks to traditional meat products for religious reasons.

Overall though, the possibility of transformative AI makes us much more optimistic about the possibility of ending factory farming in the coming century — but it’s by no means a certainty.

What might happen to factory farming in the very long-run future?

We’ve argued before that there are strong reasons to consider the interests of all future generations when assessing the scale of problems — and those arguments apply to future animals, as well as future humans.

That means that, to the extent that we can, we should try to look further ahead than just the coming century.

Here’s a simple sketch of why this matters. If we could end factory farming this century, it would be a fantastic achievement. However, if factory farming were destined to end in the coming centuries anyway or if factory farming were to re-appear, the impact of this achievement would be significantly reduced.

So: should we expect factory farming to remain in the long-run future (or re-emerge), or not?

Of course we cannot know for sure — or anything close to sure. But there are some long-run factors we can consider to try to come to a provisional view.

One rule of thumb says that we should expect the future to be guided by the values, cultures, and beliefs of people alive today. We’re all in a metaphorical tug of war about the direction of the future, and how society changes over time depends on the winners and losers of that game over the centuries.

It seems like right now things aren’t trending towards an end to factory farming — in fact, factory farming is growing.

Of course, technological changes might affect this: perhaps the real barrier to change is just a lack of alternative products, and today’s society cares enough about animals that it would end factory farming if those were developed.

But current evidence suggests that if a meat alternative was created tomorrow that was competitive on price, taste and convenience with animal products, people wouldn’t switch — so if how the future goes is just an extrapolation of what society is like today, we won’t see an end to factory farming.

A different rule of thumb looks to technology and economic efficiency. Ultimately, growing whole animals seems like an inefficient process.28 So even if demand for meat remains high, eventually we should expect to develop far cheaper ways of producing meat. Then over time, we might expect people’s values and desires to fall in line with these commercial incentives, and eating animal products will become a weird thing people used to do. Even more radically, perhaps people in the future won’t need food for survival or entertainment — because maybe most people in the future will be digital.

This second line of argument seems more persuasive than the first — which suggests we shouldn’t expect factory farming to last forever.

That said, there’s some chance that society’s values don’t remain malleable, but instead become fixed or put permanently on a particular trajectory at a certain point in time, a possibility known as “value lock-in.” It’s been argued that the invention of artificial general intelligence could make it possible to fix the values with which society is governed over very long periods of time.

Ultimately, I think the chances of long-term value lock-in related to animals are very low. It seems that to prevent the drift of values over time, you would need a very controlling system like totalitarianism. We’ve written elsewhere about the chances of a perpetual totalitarian regime and estimated that there’s an around 1 in 30,000 chance of this occurring in the next 1,000 years. The chances that such a perpetual regime will enforce particular norms about how we treat animals seems even smaller.

So, overall, over the very long-run, we’d guess that factory farming will come to an end.

How can we compare the scale of factory farming to existential risks?

We think that reducing risks to the continued existence of civilisation are particularly important moral issues.

This is reflected in our list of the world’s most pressing problems. Our top problems include:

All of these are prioritised so highly at least in part because we think they are existential risks.

So, in order to figure out how highly we should prioritise the problem of factory farming, we think it makes sense to try to compare it with existential risks.

Here, we’ll focus on the scale of these problems. To compare their overall pressingness, we’d also need to consider solvability and neglectedness; we’ll look at the solvability and neglectedness of factory farming below, but won’t make explicit comparisons to existential risks. You can check out our problem profiles on other issues to see how we think about their solvability and neglectedness.

Ultimately, the question of comparing factory farming to existential risks is really difficult. To simplify things, let’s try to compare factory farming directly with just the problem of nuclear weapons as an illustration of the kind of analysis needed.

We think the risk of nuclear war presents a large-scale problem. There are two reasons for this:

  1. A large nuclear war would have catastrophic immediate consequences, killing billions and causing unimaginable suffering — and it would reduce the size and quality of life of the population for many decades.
  2. Then there’s a small but real chance (something like 1 in 10,000 that such a war could lead to the complete collapse of civilisation, and even human extinction. This would have very long-run implications: there would be no more human civilisation, ever. Since the number of future individuals whose lives matter could be vast, the expected value of reducing a risk to those lives could also be large.

Let’s start by comparing the scale of factory farming with the catastrophic immediate consequences of a large nuclear war.

Above, we looked at how many animals count morally the same as one human. We saw that from a welfarist perspective, there are two components to this:

  1. Animals are worthy of moral consideration if they are conscious, so we need to consider the chances that each animal is conscious.
  2. Animals have different welfare ranges (i.e. capacity for wellbeing). We looked into the most recent and most comprehensive research that’s been done on this topic — Rethink Priorities’ Moral Weight Project.

Using the moral weights from the Moral Weights Project (that include both the welfare ranges and a weighting for the probability that each animal is conscious), and multiplying those weights by the number of animals of each species studied, we find that we kill the moral equivalent of 50 billion humans each year through factory farming, from a welfarist perspective.

There’s huge uncertainty on this figure. Just looking at uncertainty in the welfare ranges, this figure might vary from 60 million to around 800 billion (using the 5th and 95th percentile estimates respectively) — and that doesn’t include uncertainty on the weighting for the probability that each animal is conscious.

Let’s compare that to the immediate consequences of a nuclear catastrophe (putting aside the long-run implications for now). A large-scale nuclear catastrophe could only kill up to around 8 billion humans once — not every year.

And putting aside deaths, the other suffering caused by the way we treat animals seems plausibly even worse than the suffering a nuclear war would cause its survivors.

As a result, we think the scale of factory farming seems likely larger than (or at the very least similar to) the scale of the immediate consequences of a large nuclear war. Personally, I’d guess that, morally speaking, the scale of factory farming over the course of a few decades is something like 100 times larger than the scale of the immediate consequences of a large nuclear war.

It’s much harder to compare the scale of factory farming with the immediate consequences plus the the longer-term consequences of nuclear war arising from existential risk.

We think some kind of value lock-in that perpetuates factory farming into the very long-run future is highly unlikely, and ,over the very long-run, we expect factory farming to end. That said, we might expect the effects of our actions on the welfare or number of farmed animals to persist for decades, and it could be centuries.

How you compare this with the long-run effects of nuclear war depends on how much you buy the arguments for thinking existential risks are top priorities due to their scale .

In short, those arguments say that:

We argue for the first two of these claims in our article on longtermism and think they are very likely to be true.

Whether future individuals’ lives are likely to be good seems much harder to answer — we’d guess the answer is yes, but that’s highly uncertain. In my view, this uncertainty substantially decreases the relative importance of existential risk reduction compared to other concerns.

Finally, thinking in terms of expectations is a theoretical ideal, not a practical methodology. Making explicit estimates of expected scale (like by multiplying the size of a plausibly existential risk by the value of the future) is sometimes useful as a method, but we should also look for useful rules of thumb and robust arguments, or even use gut intuitions which can incorporate implicit understanding from experience rather than (just) explicit reasoning.

And it’s not clear that all rules of thumb — even ones focused on improving the long-term future — suggest the scale of existential risk is larger than that of factory farming.29

I think a reasonable position to take is that we should think of nuclear war as a large-scale problem for both its short-run consequences and the existential risk it poses. If that’s right, then we should think of these effects as comparable in scale — so that the existential risk’s contribution to the scale of the problem is between 10% and 90% of the total scale of the problem of nuclear war. Combining that with my view that the short-run effects of factory farming are plausibly 100 times more important than the short-run effects of nuclear war suggests that factory farming is overall larger in scale.

That said, while comparing factory farming to existential risks in general is extremely difficult, and it’s plausible factory farming is a larger scale problem than nuclear war, we do think that factory farming seems smaller in scale than the existential risk posed by AI — as the risk from AI seems substantially larger than the risk posed by nuclear war.

There are promising ways of solving this problem

As in many areas, the best approaches to solving to factory farming are probably hundreds of times more cost-effective than others.

Below, we’ll go through a few rules of thumb for finding the best approaches:

  • Don’t forget about technological advances (especially the plausible development of alternative proteins in the near future).
  • Find interventions that offer more leverage.
  • Find interventions that the animal agriculture industry won’t fight.
  • Don’t cause harm.
  • Work in neglected areas.

Later, we’ll look at what you could do with your career to help make these happen.

First, we should try to find interventions that take a bet on alternative proteins. As we’ve argued, it seems like factory farming is going to increase over the next few decades unless we develop clear alternatives to animal products. But we think it’s likely that we will develop these price, taste and convenience-competitive alternatives to meat within this century (not least because of the plausibility of transformative AI within the next few decades).

The development of these products in the near future is by no means guaranteed — and competitive products probably won’t solve everything by themselves.

But this means there’s an opportunity: it’ll probably be much easier to reduce meat consumption and improve the lives of the remaining farmed animals once competitive alternatives are developed.

As a result, we think it’s worth taking a bet: interventions whose theories of change assume that alternatives to meat will be developed to a point of market competitiveness seem likely to be more effective overall.

For example, you might work on finding ways to pre-emptively reduce barriers to the uptake of cultivated meat, such as finding ways to ensure cultivated meat adheres to religious dietary restrictions, or preventing a possible EU-wide ban on cultivated meat (which, if passed, could last decades or more).

Second, we should try to find interventions that offer more leverage — that is, ways of changing how a lot of resources are used, either via knock-on effects or directly.

One great example of this is corporate campaigns. These campaigns try to convince companies to commit to using higher-welfare animal products. Many companies will agree to make commitments without any public campaign, but if they don’t, a campaign might include protests, negative advertising, and criticism on social media. These campaigns are a cost-effective way of helping improve the welfare of large numbers of farmed animals, as they’re fairly cheap to run and can affect the entire supply chain of a large company.

These campaigns have been surprisingly successful. For example, companies such as Burger King, Unilever, and Chipotle have agreed to the Better Chicken Commitment, switching to higher-welfare slaughter methods and slower-growing breeds of broiler chicken (there’s a similar campaign in Europe called the European Chicken Commitment). Companies also seem to stick to the pledges they make in similar campaigns — for cage-free campaigns for egg-laying hens, around 90% of companies fulfil their cage-free pledges (although we’d guess more work needs to be done to make this happen in the future).

Similarly, you might be able, using a small amount of resources, to get governments to act (and they have many, many more resources). The EU is revising its animal welfare laws, and other European countries are also considering steps like banning caged hens. We think there are opportunities to help shape these reforms. In the US, it seems like ballot initiatives are almost as cost-effective as corporate campaigns. We’re particularly excited about the Good Food Institute’s work to leverage government funding for alternative proteins.

Third, we should look for interventions that the animal agriculture industry won’t fight — or, at the very least, don’t require their support. This isn’t always doable, but if you can find ways to improve the lives of animals that won’t be fought, they’re likely to be particularly cost-effective.

One great example here is developing laws in countries with relatively high-welfare farms that restrict the import of low-welfare products, as this benefits both animals around the world and domestic farmers.

Another example is working on technological progress.

Some technological progress, such as technologies to prevent the conception of male chicks or immunocastration for piglets, could substantially improve the lives of farmed animals — and many have only small costs for farmers (e.g. in-ovo-sexing, which ensures all chicks born are female, increases the costs of egg production by around 1–3 cents per egg).

We’d also like to see research on ways of reducing keel bone fractures, as these are a major cause of suffering in laying hens, and may be worse in cage-free systems. More speculatively, we might be able to gene edit animals to prevent them from feeling pain.

If we can find suffering-reducing technologies that are cheap to use, it may be easy to get farmers to use them (as we’re seeing with in-ovo-sexing).

Fourth, find solutions that don’t cause harm.

As an example, we might be able to convince people not to eat beef on environmental grounds or by raising the price of beef through welfare interventions. But this might lead to people eating chicken instead — and not only do chickens live in much worse conditions, but they’re smaller, so many more chickens are killed for the same amount of meat. The same problem applies even more so for small fish and smaller invertebrates like shrimp.

a cute kitten
Source: Our World In Data.

Finding solutions that are very unlikely to cause harm is much harder than it sounds — there are plausible reasons why many of the solutions we’ve discussed above could cause harm. If you develop a technology that improves animal welfare (like immunocastration or in-ovo sexing), but this technology lowers costs for farmers, it also might decrease prices, increasing the demand for animal products and thereby increase the number of animals in factory farms, resulting in more harm. Weighing up these effects is hard, and price-decreasing technologies may well end up being good overall — but because of the plausibility of harm, they’re unlikely to be among the best approaches to solving factory farming, so we’d avoid working on them if you have other options.

As another example, corporate campaigns to get farms to switch to slower-growing breeds of chicken improves the welfare of the chickens, but these slower-growing breeds live longer, which might lead to more suffering overall. Corporate campaigns might also support humanewashing — convincing people that meat is high-welfare when it’s not, encouraging people to buy low-welfare products.

Overall, we still think many corporate campaigns are worth doing — but it’s important to investigate possible sources of harm when choosing what to work on.

In some cases, people have stopped their work because of this risk. Anima International were campaigning to end live sales of carp in Poland. They became worried about carp being replaced with salmon, a carnivorous fish, and as a result increasing the number of fish farmed overall — and so ended their campaign. We think this is extremely commendable!

Some writers on animal welfare, such as Brian Tomasik go further and suggest avoiding most welfare interventions and instead focusing on improving slaughter — as slaughter improvements don’t introduce other complicating factors like changing the length of time animals live.30

We’re not saying you need to be certain that the intervention you work on doesn’t cause harm — that’s just not possible. But it’s important to do what you can to think through the downsides and try as much as you can to avoid interventions that sound good and which might even do some good, but which are bad ideas overall because they may cause more harm.

Finally, if you can work in a particularly neglected sub-area of the issue, your solutions are more likely to be cost-effective overall. This is because solutions have diminishing returns to effort.

As we’ve seen, the vast majority of the animals in factory farms are fish and invertebrates (like crustaceans and insects). Work on these animals (like getting any aquatic welfare standards or finding better ways of slaughtering shrimp) seems pressing. We also think work to prevent the rise of insect factory farms seems particularly pressing.

Similarly, work in non-western countries is likely to be more cost-effective overall.

Work on factory farming is highly neglected

We’ve argued so far that factory farming is similar in scale to small existential risks, and that there are promising ways of solving the problem. But work in this area is also pressing because it’s so neglected that the impact of an additional person working on it will tend to be high.

Philanthropic spending on preventing factory farming is around $290 million a year31 — which is similar to philanthropic spending other issues we prioritise (like risks from AI and nuclear war), and over 100 times less than areas like global development (around $70 billion) or climate change (around $60 billion). Open Philanthropy, the largest single funder of farmed animal welfare work,32 told us that they expect total spending in the area to grow to around $450 million a year over the next decade.

This spending corresponds to something like 1,750 to 2,000 people working on this problem full-time, growing to something like 2,500 to 3,000 people over the next decade.33

There’s also work being done in the private sector, especially in alternative proteins. There are something like 55,000 people total working for companies in the area (like Quorn, Gardein, and Beyond Meat). We’re uncertain how to count these people — most of what they’re doing is delivering products into supermarkets, which is helpful, but not really focused on advancing the field or solving factory farming. Only around 300 of these 55,000 roles are focused on cultivated or plant-based meat research.

If we focus only on work we think directly addresses the issue of factory farming (e.g. alternative protein research roles) spending from the private sector is more like $120 million per year, with around 750 people working on the problem.33

That means altogether, we’d estimate there’s ~3,000 people working on reducing harms from factory farming and around $410 million dedicated to this issue.

Within that, there are particular areas that we see as particularly neglected. There are around 10 people working full time on insect farming and maybe around 20–30 on shrimp farming.34 And it seems like Open Philanthropy’s donors have decided not to fund work on invertebrate welfare, so these areas are particularly neglected in funding.

What are the major arguments against this problem being (especially) pressing?

It’s a bit hard to argue that factory farming isn’t a fairly pressing problem. Given the scale of the issue, the fact that it is caused by human practices (and therefore would seem to admit of human solutions), and how animals are treated, in order to think this issue doesn’t deserve much attention you’d need to have either a very pessimistic outlook on our ability to do anything about the issue if we try, or think that animals’ moral standing is extremely tiny compared to humans (or that they don’t matter at all morally).

That said, there are more arguments that you could make that other issues are even more pressing than factory farming. And if you’re thinking about how you should spend your career, that matters. Here are a few such arguments:

  • You might think that the wellbeing of each animal matters much less than the wellbeing of each human. If you think the difference is big enough, then you’ll think the scale of this problem isn’t as large, in terms of moral importance, as the scale of problems that affect current generations of humans, like global health or any imminent risks from pandemics, war, and AI.
  • You might think that the long-run future has an extremely large and positive expected value. If so, and you think that there are sufficiently tractable ways to preserve that value, this suggests that you should work on reducing existential risks instead.
  • You might agree with prioritising work on the long-run future, but think that there are other ways of affecting the long-run future other than working on existential risks — some people prioritise improving the trajectory of civilisation over preserving its existence. For example, you might think the way we treat digital sentience could have a bigger impact on the long-run future than the way we treat animals, if you expect there to be more digital minds than animal minds. If so, this suggests that you should work on digital sentience (or another issue affecting the long-run future) instead.

There are also some reasons that you might be very pessimistic about our ability to make progress on ending factory farming — at least in a way that doesn’t worsen other issues:

  • You might worry that promoting concern for other beings increases risks of astronomical future suffering. This suggests you should work on s-risks instead.
  • You might think that work on farmed animal welfare could make lives worse for wild animals.
  • You might think that the lack of growth in veganism and pro-animal attitudes over the past decades (since an initial surge in popularity in the West) suggests that moral attitudes to animals are highly resistant to change. If you’re also sceptical about technological change ending factory farming, you might just think the problem is intractable.

What can you do to help?

When people think about working in animal welfare, they might picture looking after chickens in a sanctuary or handing out leaflets in a shopping mall.

But there are options that we think can achieve many times more impact than either of these options. A really high-impact job that’s working on reducing factory farming would focus on the ways of solving the problem that we discussed above.

As a result, the main options we recommend in this area are:

We’ll go into some detail about each of these below.

If you’re moving into the area for the first time, or early in your career, you could also try:

We’d also recommend taking a look at Animal Advocacy Careers as well as Hive, a Slack workspace and newsletter focused on animal careers.

If you already have experience, it might be worth looking at what the main skill bottlenecks are to making progress on the problem, in order to figure out where you can contribute best. When we spoke to people we know working on solving factory farming, they told us the main bottlenecks are:

  • Funding
  • Management and leadership
  • Government and policy expertise
  • Entrepreneurship and starting new organisations

We expect these to change over time, so they’re a less useful guide if you think you should currently be spending time building career capital and only aiming to have an impact later (which we often recommend for people who are early in their careers).

Of course, no matter what your job, if you have more wealth than you need you can help by donating a portion of your income to the most cost-effective animal charities you can find. (See here for a few recommendations.)

Earning to give

As we mentioned, there’s very little funding for factory farming relative to many other causes.

So one of the best ways to help might be by earning to give, which means taking any job that’s higher paying in order to donate more.

Our top recommendations for earning to give — because they can pay so much — are:

These paths are highly competitive, but if you’re exceptionally successful, you could earn millions of dollars a year. If you donate most of that, that would mean increasing the funding for the whole cause area by something like 1% by yourself. If you take the time to make sure you’re donating to particularly effective opportunities, your impact could be even higher than that suggests.

We’d be particularly excited about people earning to give to support invertebrate welfare (e.g. insects, shrimp, and other crustaceans) because these areas aren’t currently being funded by large donors like Open Philanthropy.35

Helping to run nonprofits

Many of the organisations working on the most cost-effective solutions are nonprofits.

These organisations need people to help run them. That might mean working in management, HR, finance, accounting, law, fundraising, software, operations, marketing, communications, and more.

According to Animal Advocacy Careers, nonprofits working on ending factory farming currently find it hardest to hire for leadership and fundraising positions, with software engineering positions also being quite difficult to fill. So if you’re a good fit for one of those, you might find your skills are especially useful.

The best way to get started if you don’t already have experience is probably by finding any role that will let you start learning relevant skills — even if that’s not initially at a high-impact organisation.

To learn more, including more information on how to get started, take a look at our article on organisation-building skills.

You could also look at:

Founding something new

Founding something is a difficult path — but if you succeed, it can be hugely impactful.

There are two broad routes here outside founding a company for earning to give:

  • Founding a new nonprofit focusing on a gap in what’s needed to tackle factory farming
  • Founding an alternative protein startup, especially if you can find investment that wouldn’t otherwise be spent on alt protein research

Even if it doesn’t work out, you’ll probably learn a huge amount — so this can also be a great route for career capital.

If you’re starting a nonprofit, one option is to join an incubator, such as Charity Entrepreneurship or, if you’re in Southeast Asia, Welfare Matters’ Farmed Animal Welfare Incubator.

If you’re interested in founding an alternative protein startup, take a look at these resources from the Good Food Institute. The Good Food Institute also offers a mentorship programme for alternative protein entrepreneurs, a database of accelerators and incubators, and a community and events.

To learn more about this career path, read our career review on founding impactful organisations.

Government and policy

There are two broad areas for government and policy work on factory farming:

  • Improving welfare standards on farms — as we mentioned earlier, there are lots of exciting opportunities for legislative reform.
  • Removing barriers to and leveraging government support for alternative proteins — most of this work is currently carried out by the Good Food Institute and its affiliates around the world.

Working in policy might mean working in the executive branch, lobbying from the outside, or possibly even running for office.

Our article on policy and political skills goes into more detail, including a bunch of ideas about how to get started in the area.

Corporate campaigning and activism

Corporate campaigns – like those we mentioned above working on improving the lives of caged chickens — are one of the most cost-effective ways people have found of improving welfare for currently farmed animals.

So learning to effectively carry out these campaigns is a particularly useful skill set, especially if you can then go on to run entire teams or organisations of campaigners.

We haven’t looked into this area much, but we’d guess the best way to get started would be to volunteer with an organisation doing corporate campaigns, like The Humane League.

Scientific and engineering research

We think there are opportunities for scientists to help reduce factory farming by:

  • Developing alternative proteins
  • Working in animal welfare science

This work will probably require you to have a relevant degree. In both fields, biology or chemistry undergraduate degrees are useful; you can also do research into animal welfare with philosophy or neuroscience backgrounds.

About half of alternative protein job postings say they require a relevant postgraduate degree — either a relevant PhD or a master’s with multiple year’s additional industry experience. According to the Good Food Institute, relevant areas include: mycology, plant biology, molecular biology, cell biology, biochemistry, food science, genetic engineering, chemical engineering, mechanical engineering, bioengineering, tissue engineering, and materials engineering.

To get started on an alternative protein research career path, we’d recommend this newcomer’s guide from the Good Food Institute..

There are also lots of unanswered questions in animal welfare science. What are the relevant indicators for animal consciousness? How much pain do animals feel on farms in different situations? We have some vague answers to these questions, but better answers could importantly guide people’s actions — and we know far less for some animals, like fish and invertebrates.

For this article, we leaned heavily on the work of the Welfare Footprint Project, who carefully analyse the effects of things like the Better Chicken Commitment or banning gestation crates for sows. On insects, we’re excited about the work of the Arthropoda Foundation.

Most research in this area happens in academia (and even at organisations like the Welfare Footprint Project, most researchers have an academic background) — which again means you’ll likely need a relevant postgraduate degree. For more on academic careers, read our article on how to become an academic researcher.

There’s other potentially useful research such as the development of in-ovo sexing and immunocastration. More speculatively, The Far Out Initiative is researching ways of genetically engineering animals who don’t suffer — I’d guess that at least some people should be engaging in this sort of research, but it’s much less likely to pan out than the other research we’ve mentioned in this section.

If you’re interested in a career in research and just getting started, we’d suggest reading our article on building research skills.

Strategy research and grantmaking

These roles involve figuring out what we should be doing to most help farmed animals, and then making that happen.

This might mean something like economics research in academia looking at the price elasticity of animal products to figure out where cost-increasing interventions should focus to have the biggest impact (or where cost-decreasing interventions could cause substantial harm), or becoming a grantmaker at a foundation looking to fund solutions to factory farming.

For academic research you’ll want to pursue a PhD in economics, or get a degree in public policy or maybe philosophy. Doing similar research at a think tank or nonprofit might not require an advanced degree, but it’s probably still useful.

For becoming a grantmaker, you should probably start by doing any other work in animal welfare, building up your expertise and connections. (Read more about how to become a grantmaker.)

You could also try testing your fit for research and strategy roles by doing a research project in your spare time or funded by a grant, perhaps on a topic on Open Philanthropy’s list of social science research topics for animal welfare.

Find vacancies on our job board

There are also other job boards that you might find helpful, including:

Key organisations

The main funders and grantmaking organisations (donating $500,000+ a year) are:

There are more animal grantmakers on this list.

Organisations working on alternatives to animal products:

Some particularly relevant governmental and intergovernmental organisations:

Lobbying and campaigning organisations, focusing on both the government and corporations. Here are some of the top organisations around the world:

Organisations focused on dietary change:

  • Veganuary has played a significant role in promoting vegan food in the UK and now has campaigns across the EU, India, South Africa, and the US.
  • Dansk Vegetarisk Forening promotes plant-based products in Denmark.

Meta organisations, which help build the animal movement. Here are some of the top organisations around the world:

  • Ambitious Impact runs programs like the Charity Entrepreneurship incubation programme, starting new charities focused on global health and animal welfare.
  • Faunalytics conducts research useful for animal advocates.
  • New Roots Institute educates students in the US about animal agriculture.
  • Catalyst is building an animal welfare movement specifically in Thailand.
  • Welfare Matters is building an animal advocacy movement across Southeast Asia.
  • Animal Alliance Asia is also focused on creating a cross-Asia animal welfare community.
  • Animal Advocacy Africa is trying to build an animal advocacy movement in Africa.
  • Rethink Priorities and Animal Ask do research that tries to resolve important strategic questions for animal advocacy.

Organisations focused on fish and crustacean welfare:

Other nonprofit organisations working in the area:

Learn more

Top recommendations

Further recommendations

Notes and references

  1. Data on the proportion of animals in factory farms is hard to find — not least because the exact definition of factory farm is difficult to define. In this article, we’ll try to explain how animals are usually treated in the vast majority of commercial farms.

    We can get some data on this question from the United States Environmenal Protection Agency.

    The USDA defines a “concentrated animal feeding operation” (CAFO) as a farm where animals are confined for 45 days or more in any 12-month period, and where the number of animals confined reaches a certain threshold, depending on the species, and where wastewater is managed in a certain way.

    We’d guess that almost all CAFOs are factory farms, but that many non-CAFOs are also factory farms (because they have animals in high-density conditions but have few animals, or because they have good wastewater management — so they don’t have the sort of environmental or polluting impact that the EPA cares about).

    The Sentience Institute combined EPA data on CAFOs with USDA Census of Agriculture data on total animal populations (see the numbers in this spreadsheet to find the % of animals in factory farms in the US.

    They found that 98% of land animals in the US were factory-farmed in 2017:

    • 99.96% of broiler (meat) chickens
    • 98% of egg-laying hens
    • 99.9% of turkeys
    • 98% of pigs
    • 70% of cows

    Unfortunately, there’s little data on the rest of the world. Based on our understanding of the regulatory landscape, we’d guess the situation is better in the EU, and worse elsewhere.

    We’d also guess that the situation for fish is even worse than for broiler chickens. As far as we can tell there is very little regulation about fish welfare and very little consumer demand for high-welfare fish.

    We discuss more about how each animal is commonly treated here.

    If the number of animals is roughly proportional to population, 90% of animals in the EU are factory farmed (which I’d guess is an underestimate), 98% of animals in the US and elsewhere are factory farmed, then approximately 97.5% of animals globally are factory farmed.

    For more discussion, see How many animals are factory farmed by Hannah Ritchie at Our World in Data.

  2. In this article, all estimates of global captive vertebrate numbers are from this post by Saulius Šimčikas.

    We chose this because it’s a collation of figures from the most widely used source – the Food and Agriculture Organization of the United Nations statistics database, FAOSTAT. Where FAOSTAT data was unavailable Šimčikas used reasonable sources and noted his uncertainty, giving ranges instead of point estimates. For example, for the number of farmed fish, Šimčikas based his estimate on data from Fishcount (who later published that data under peer review and noted its high uncertainty.

    Šimčikas noted issues with FAOSTAT and discussed at least one of these with the Food and Agriculture Organization who agreed there was a mistake.

    Overall, this source seems comprehensive and fairly reliable, and unlikely to be biased in any particular direction (even though Šimčikas was an animal advocacy researcher when he wrote the post).

  3. Reliable estimates on the number of invertebrates in farms are harder to find than estimates for vertebrates. We relied on three sources:

    We’re not aware of any estimates of the number of cephalopods (e.g. squids, octopuses), krill or bivalves (e.g. clams, oysters, mussels) in farms.

  4. We kill around 5 to 70 trillion wild marine animals every year (mainly shrimp), and between 100 trillion and 10 quadrillion smaller invertebrates, primarily through the use of pesticides. For completeness, we could also consider human effects on nematodes and zooplankton, which are likely extremely large in scale, but these tiny animals are generally poorly studied and everything about them (how many there are, how humans affect them, and whether they feel pain) is extremely uncertain.

  5. We farm many other invertebrates, like cephalopods (e.g. squid), bivalves (e.g. clams, oysters), but we don’t have a good estimate of how many of these animals are in farms.

  6. Phrase borrowed from Luke Muehlhauser’s 2017 Report on Consciousness and Moral Patienthood.

  7. This report was conducted while Muehlhauser was a Research Analyst at Open Philanthropy. Open Philanthropy is 80,000 Hours’ largest funder. We’re referencing this report not because Muehlhauser is an expert in the area (he’s not), or because we agree with its assumptions or conclusions (I don’t — I’m in particular suspicious of assuming functionalism and illusionism), but because it’s an unusually broad and transparent review. If you’re interested in reading more about animal consciousness, we’d recommend taking a look at the full report, as well as its extensive list of sources.

  8. Meuhlheuser notes that play behaviour was proposed as an indicator of consciousness in Rial et al., 2007.

    …play shows several traits indicative of consciousness. Besides of being an onerous activity, play seems to be always pleasant. The only explanation for the play paradox lies in considering that the expenditure of energy must have a wide variation in hedonic value, from rather unpleasant to extremely pleasurable, that is, it shows a wide range of alliesthesia. An animal confronted with the possibility of playing should rank the costs and the benefits of each alternative and its final decision will aim at maximizing pleasure. Therefore, the presence of play should be a sign of consciousness.

    Rial, R. V. et al. “The Evolution of Consciousness in Animals.” Consciousness Transitions: Phylogenetic, Ontogenetic, and Physiological Aspects, Elsevier, 2007, pp. 45–76.

  9. This is because many theories of consciousness require complex neural behaviour, and this likely correlates with complex cognitive ability.

    If you are interested in investigating this in more detail, Muehlhauser’s table of potentially consciousness-indicating features includes a very long list of cognitive abilities that have been linked to consciousness, as well as sources for those claims.

  10. This is because, even though there are theories where consciousness itself isn’t necessary for moral consideration, most moral theories think that pain is morally relevant.

  11. More comprehensive attempts can be found in Meuhlhauser’s report and the “Proxy references” sheet of this spreadsheet put together by Rethink Priorities.

  12. See my overall thoughts on PCIF arguments in Meuhlhauser, 2017 for more.

  13. Meuhlheuser considers whether we can find any necessary or sufficient conditions for consciousness (he investigates the view that a cortex is required for consciousness), but doesn’t come to any clear conclusions.

    Meuhlhauser also considers some big-picture considerations on how rare consciousness is.

    For example, he investigates whether, for all behaviours that arise from conscious systems, there is the possibility that an unconscious system could behave the same way — if there is, it’s much harder to find evidence for consciousness. He also investigates how complex we should consider consciousness to be: if it’s more complex, we should think it’s less likely.

  14. Excluding mealworms and some bivalves. Mealworms are the larval form of the yellow mealworm beetle, Tenebrio molitor. I haven’t really investigated this, but my intuition is that insect larvae are less likely to be conscious than adult insects. Some bivalves like oysters and mussels also seem plausibly nonconscious.

  15. It’s notable that for those animals where we think they are less likely to be conscious, they’re often farmed in much larger numbers (e.g. fish, shrimp and insects). This makes it harder to dismiss the overall scale of the moral problem.

  16. We don’t necessarily have to reject the idea of binary moral status in order to do this — we could assign moral status to neurons themselves rather than to individuals. I don’t know of any philosophers who have argued for this position.

  17. Cow neuron counts from Vertebrate Neuron counts by Jaoa Fabiano — for other data, see the table above. We couldn’t find good sources for neuron counts in invertebrates.

  18. According to Shriver, “What Neural Counts Can and Can’t Tell Us About Moral Weight”:

    For there also are a large number of studies showing an inverse relationship between brain volume and the intensity of particular affective states. In particular, when it comes to chronic pain, by far the most commonly cited relationship between brain volume and chronic pain is a decrease in brain volume in regions commonly associated with the experience of pain. For example, see the Davis et al. (2008) article “Cortical thinning in IBS: implications for homeostatic, attention, and pain processing.”

  19. You can find all the data used, and all 12 models’ results here.

  20. They assumed the following probabilities of moral consideration:

    Species

    Humans
    1
    Pigs
    0.780
    Chickens
    0.700
    Octopuses
    0.600
    Carp
    0.600
    Bees
    0.250
    Salmon
    0.070
    Crayfish
    0.070
    Shrimp
    0.070
    Crabs
    0.250
    Black Soldier Flies
    0.100
    Silkworms
    0.050

    These are largely in line with the subjective probabilities estimated by Meuhlhauser.

    The team note that:

    There is deep uncertainty about consciousness generally and sentience specifically. In the face of that uncertainty, we think there’s no good argument for assigning a credence below 0.3 (30%) to the hypothesis that normal adult pigs, chickens, carp, and salmon are sentient. Likewise, we think there’s no good argument for assigning a credence below 0.01 (1%) to the hypothesis that normal adult members of the invertebrate species of interest are sentient. So, skepticism about sentience might lead you to discount our estimates, but probably by fairly modest rates.

  21. Much of the initiative behind insect farming is based around the idea that insects are a more sustainable option than other food sources. We’re sceptical of this claim. According to a preprint literature review by Biteau et al., 2024a:

    Many of the benefits commonly mentioned by companies and proponents of insect farming are challenged by current scientific evidence… There are significant uncertainties, with many authors highlighting the fact that the future environmental impact of large-scale insect production is largely unknown. This is especially true given claims that insects can be fed on food waste and that insect frass can be used as fertiliser, both of which have considerable challenges to overcome at scale. Lastly, most insect based foods replace plant-based products with limited environmental impact rather than meat, and several studies indicate that insects-based feeds and pet food can have a larger environmental impact than conventional products.

    A core claim here is that feeding insects food waste is difficult. In a peer-reviewed literature review by Biteau et al., 2024b:

    While the idea of turning trash into treasure for insect agriculture may be appealing in theory, the reality appears to be more challenging. Only some species of insects can be farmed using food waste, while others perform poorly. The inconsistent availability and quality of food waste pose significant obstacles to the establishment of large-scale insect farms aimed at consistently yielding high-quality products. Consequently, insect-farming companies often resort to utilising high-quality feeds already in demand by other sectors. Moreover, competition intensifies for the limited pool of food waste suitable for insect agriculture, as various industries, including agriculture, aquaculture, pet food production, and biogas manufacturing, vie for the same resources. Additionally, concerns regarding food safety due to contamination risks constrain the types of food waste viable for insect cultivation.

    Corentin Biteau, et al. Have the Environmental Benefits of Insect Farming Been Overstated? A Critical Review. 4 Apr. 2024, https://doi.org/10.32942/x2w60r.

    Corentin Biteau, et al. “Is Turning Food Waste into Insect Feed an Uphill Climb? A Review of Persistent Challenges.” Sustainable Production and Consumption, 1 July 2024, https://doi.org/10.1016/j.spc.2024.06.031.

  22. See Table S1 from Whitton et al., 2021.

    Meat consumption per capita fell from 2000 to 2019 in six investigated countries: Canada, Ethiopia, New Zealand, Nigeria, Paraguay and Switzerland.

    It’s plausible that the explanation in Canada, New Zealand, and Switzerland is related to income per capita.

    Whitton, Clare, et al. “Are We Approaching Peak Meat Consumption? Analysis of Meat Consumption from 2000 to 2019 in 35 Countries and Its Relationship to Gross Domestic Product.” Animals, vol. 11, no. 12, 6 Dec. 2021, p. 3466, https://doi.org/10.3390/ani11123466.

  23. This is only a median scenario — the UN thinks there’s around a 40% chance that the global population won’t start to decrease this century.

  24. One source of evidence for this is the so-called “meat paradox” — the phenomenon that people use animals in ways that harm them, despite also caring for animals and wishing them no harm.

    A literature review by Gradidge, et al. (2021) finds support for the prevalence of this phenomenon.

    The ‘meat paradox’ (MP) is the phenomenon of people using animals in ways that harm them (e.g., meat consumption), despite caring for animals and wishing them no harm (Loughnan et al., 2014)1. This theoretical MP represents a form of cognitive dissonance (hereon dissonance), describing the discomfort arising from a contradiction between one’s beliefs and behaviours (Loughnan et al., 2014). For instance, most US participants (n = 1,024) are very or somewhat concerned about animal welfare across contexts (e.g., research, 67%; zoos, 57%; food production, 54%; Riffkin, 2015), indicating most people care about animals. In fact, people empathise more with dogs than adult human victims (Levin et al., 2017). Yet, even though care for animals sometimes exceeds care for hu­mans, 90-97% of people consume meat (Food Standards Agency [FSA], 2012; The Vegan Society [TVS], 2019)… most articles within this review directly or indirectly supported the MP (70 articles; 95.89%).

    Gradidge, Sarah, et al. “A Structured Literature Review of the Meat Paradox.” Social Psychological Bulletin, vol. 16, no. 3, 23 Sept. 2021, https://doi.org/10.32872/spb.5953.

  25. The main challenges are flavour and texture (see Wang et al., 2022), as well as nutrition, food safety, cost, and consumer confidence (see Ahmad et al., 2022).

    For more on the science of plant based meat, take a look at this introduction by the Good Food Institute.

    Wang, Yaqin, et al. “Flavor Challenges in Extruded Plant‐Based Meat Alternatives: A Review.” Comprehensive Reviews in Food Science and Food Safety, vol. 21, no. 3, 26 Apr. 2022, https://doi.org/10.1111/1541-4337.12964.

    Ahmad, Mudasir, et al. “Plant-Based Meat Alternatives: Compositional Analysis, Current Development and Challenges.” Applied Food Research, vol. 2, no. 2, Dec. 2022, p. 100154, https://doi.org/10.1016/j.afres.2022.100154.

  26. Humbird, 2020. See table 4.7 for the capital expenditure (CAPEX) and production rate for a fed-patch production facility and table 4.14 for a perfusion process facility.

    The facility is assumed to contain 24 bioreactors of 20 m3. This minimises the cost of production within a single facility subject to the physical constraints outlined by the report (see figure 4.4b).

    Humbird, David. Scale-up Economics for Cultured Meat: Techno-Economic Analysis and Due Diligence. 29 Dec. 2020, https://doi.org/10.31224/osf.io/795su.

  27. This report from Jacob Peacock at Rethink Priorities suggests that price, taste, and convenience-competitive plant-based meat would not currently replace meat. We’re a bit more optimistic than that, as we’d guess that if there were clear alternative products, it would be much easier to both push for increased welfare on farms and to convince people to stop eating factory-farmed meat.

  28. It’s very hard to predict technological changes over the very-long-run. That said, one way of doing so is to look at fundamental limits. In this case, we can look at the energy efficiency with which animals convert calories in their feed into calories in meat. This calorie efficiency ranges from around 5% (in prawns) to up to 30% (in broiler chickens) (see figure 2 in Fry at al., 2018). While we’ll likely find ways to improve this efficiency (for example by developing more efficient breeds), it seems fundamentally impossible to grow an animal which moves around and has a complex brain more efficiently than you could grow just the parts you’d want to eat.

    Even more precisely, sensory systems, including nociception (neural pathways which cause pain), use energy. According to Niven and Laughlin, 2008:

    Energy consumption affects all aspects of animal life from cellular metabolism and muscle contraction to growth and foraging (Alexander, 1999). Yet despite early studies on energy metabolism in neural tissue (e.g. Kety, 1957), the impact of energy consumption upon the evolution of nervous systems has only recently begun to be generally appreciated (Laughlin, 2001). Recent studies have made substantial advances in relating the energy consumption of neural tissue to neural function. Together these studies show that there are high energetic costs associated with the nervous system both at rest and whilst neurons are signalling (Laughlin et al., 1998; Attwell and Laughlin, 2001; Niven et al., 2007). Crucially for the evolution of the nervous system, and in particular sensory systems, these costs are incurred even during activity. Thus, animals pay an energetic cost associated with nervous system irrespective of the demands of other tissues such as skeletal muscle.

    It doesn’t seem fundamentally impossible to produce meat without the development of these neural pathways and doing so would be more efficient, so we’d guess that a sufficiently advanced future society would do so.

    Fry, Jillian P, et al. “Feed Conversion Efficiency in Aquaculture: Do We Measure It Correctly?” Environmental Research Letters, vol. 13, no. 2, 1 Feb. 2018, p. 024017, https://doi.org/10.1088/1748-9326/aaa273.

    Niven, Jeremy E., and Simon B. Laughlin. “Energy Limitation as a Selective Pressure on the Evolution of Sensory Systems.” Journal of Experimental Biology, vol. 211, no. 11, June 2008, pp. 1792–804, https://doi.org/10.1242/jeb.017574.

  29. Perhaps a good rule of thumb is to think of current-day civilization as if we’re holding a baton in a relay race. At some point soon we’re going to hand over that baton to a future generation — and our job is to put those people in the best position possible to improve the long-run future.

    I think this is an attractive rule of thumb because it helps avoid concerns that our uncertainty about how the long-run future will unfold makes it much harder to influence than the present. That’s because, under this rule of thumb, we focus only on the effects of our actions when we have the baton (and when we hand that baton over) — that is, the relatively near future (e.g. the next 100 years).

    This rule of thumb suggests that ensuring the next generation exists to take over in the race is crucial (reducing existential risk), but it also seems to support ensuring they make good decisions and ‘safeguarding’ civilisation while we are responsible for it (making sure we don’t do anything morally awful). I think this points less towards working on existential risk than the explicit expected value calculation.

    There are other plausible rules of thumb for identifying particularly important issues to work on (from a longtermist perspective). For example:
    * Identify the main process that’s going to substantially affect the future and work on its trajectory. We’d guess this points towards working on improving the trajectory of transformative AI.
    * Identify the worst thing we’re currently doing and find a way to stop it — especially if stopping it seems unlikely to be reversed. We’d guess this points towards working on factory farming.

    It’s hard to know what to conclude from all this. I think the explicit expected value estimate is pretty useful in this case because it’s capturing something important that the other rules of thumb miss. On the other hand, in the case of nuclear war, the existential risk is pretty small — approximately 1 in 10,000. Also, I don’t really expect existential risk to decrease very much in the next few centuries, and I worry about many future lives possibly being so bad that they’re not worth living, which makes me less confident in the value of the future.

  30. This makes slaughter improvements seem particularly likely to be cost-effective interventions. Tomasik even argues that ending animal farming might be bad overall, because of the effects on wild animals. (I think this is an interesting argument, but I’m less confident than Tomasik about whether these effects on wild animals would be good or bad overall.)

  31. According to Open Philanthropy in written correspondence.

  32. Open Philanthropy spent around $100m on farmed animal welfare in 2023.

  33. Also according to Open Philanthropy in written correspondence.

  34. According to Prof Bob Fischer at Rethink Priorities in written correspondence.

  35. The linked post doesn’t explicitly mention an intention to drop invertebrate funding but for other reasons we believe this is included in the areas exited by Open Philanthropy.