You Can't Tell How Conscious Animals Are By Looking At Them!

Follow up to the honey article

I say many things that get lots of people on the internet to write long screeds about my alleged wrongness so extreme that its total tonnage could stop a team of oxen in its tracks. But I don’t think I’ve ever gotten more backlash than I did recently when I made a case against honey. Now some of the criticisms were factual. One person correctly noted that I made a dumb error by saying the bees were indoors, which was totally false—I didn’t actually intend to say this, I think it was just kind of muscle memory from writing about indoor conditions for other factory farmed animals, but it was a pretty embarrassing error!

Others disagreed with my conclusions about the quality of bee life (see the best critical reply to me here, and my comment for why I think it’s very likely wrong). In short, I think it’s clear that bees are stressfully overcrowded, subject to frequently unpleasant transport, and frequently stricken by diseases and parasites. Insulation is poor, starvation is common, and most don’t live very long. Overall, their lives are quite bad, and eating just a kg of honey plausibly causes more extra years of unpleasant animal life than all almost all of one’s other annual animal consumption.

Isaac King correctly noted that the 97% number that I gave was pretty misleading. I originally said:

97% of years of animal life brought about by industrial farming have been through the honey industry (though this doesn’t take into account other insect farming).

Now this is false—it neglects farmed fish, farmed shrimp, and farmed insects. Really the correct statistic (which I have updated my article to say) was “Of all the farming years brought about by the honey, chicken, cow, sheep, turkey, duck, pig, and goat farming industries, 97% have been brought about by honey (though it’s comparable to numbers of shrimp, farmed fish, and farmed insects).”

So I definitely made a few dumb factual errors! But these don’t really affect the broader plausibility of my conclusion.

A more common target of attack was the 15% number for bee sentience. In my article I said:

The median estimate, from the most detailed report ever done on the intensity of pleasure and pain in animals, was that bees suffer 7% as intensely as humans. The mean estimate was around 15% as intensely as people. Bees were guessed to be more intensely conscious than salmon!

Shortly after publishing, I added the following note:

Edit 7/1: People have been getting too hung up on the exact numbers. You shouldn’t take them too literally. They’re just meant to be a sort of rough, order of magnitude estimate.

Very large numbers of people have been suggesting that this number is completely ridiculous and insane. My support for it supposedly discredits my seriousness as a thinker and means I should abandon the life of the mind and begin working in a coal mine. The wrongness of this claim alone allegedly has enough total tonnage to stop probably a team of horses in its tracks (though not of oxen—you’d need the wrongness of my entire article for that).

The genial Nathan Young also thought the number was absurd, but revised that assessment after he and I had a conversation.

Now, I hope you will pardon my upcoming snarkiness—each of the people making this argument that I’ve listed have been generally decent and trying to make the world a better place. I’d encourage you to, like Henry Howard, take the Giving What We Can pledge. But I think their criticisms have been stupendously epistemologically in error!

First of all, these people have been clearly misreading what I’ve said, even after it was pointed out that this was a misreading. I did not say that seven bees are more morally valuable than a human. I do not think seven bees are more morally valuable than a human! Humans are more valuable than bees for a lot of reasons including:

1. We’re able to experience far greater pleasures than those available to bees. We can get the pleasure of love, sublime poetry (well, I can’t, but some people can), and great achievement.

2. We live much longer than bees. Thus, even if bees had more pleasure per moment, a human life would still be vastly more important.

3. Humans have access to a vast range of goods that bees do not. I do not think the only goods are pleasure and pain. Humans can have relationships, deep wisdom, and achievements that intrinsically contribute to the value of our life even without bringing us pleasure. Of the non-hedonic goods, humans have a near-monopoly. Bees probably don’t have any of them.

4. Humans likely experience comparable pleasures and pains more intensely than bees. Probably if a bee hurts its wing and a human hurts her arm by similar amounts, the human feels considerably more pain.

5. Humans can improve the world in ways that animals can’t. Humans can cure cancer. Bees can’t.

The 15% number was about the welfare range. It was only about the 4th consideration. It was not about overall assessments of importance. It was just about the relative intensity of pleasure and pain.

Additionally note: the 15% number was the mean estimate not the median. It was high because there were some low probability scenarios of very intense bee valenced experience (even a 5% chance that they’re as intensely conscious as us quite thoroughly skews things)! And you shouldn’t take the numbers too literally. They’re mostly intended to be a rough order-of-magnitude estimate given our great degree of uncertainty.

Now, is this estimate absurd? I don’t think so. Perhaps my critics have some quasi-mystical faculty that enables them to directly intuit how intensely bees suffer, but I am bereft of such a faculty. I thus have to look at the published literature on the subject. The 15% number was the mean estimate from a very detailed months-long report compiled by large numbers of experts. The basic methodology was as follows:

There are many theories of valence, not all of which are mutually exclusive. For instance, some think that valenced experiences represent information in a motivationally-salient way (“That’s good” / “That’s bad” / “That’s really good” / etc.; Cutter & Tye 2011), others that valenced experiences provide a common currency for decision-making (“A feels better than B” / “C feels worse than D”; Ginsburg & Jablonka 2019), and others still that they facilitate learning (“If I do X, I feel good” / “If I do Y, I feel bad”; Damasio & Carvalho 2013). In all three cases, there are potential links between valence and conceptual or representational complexity, decision-making complexity, and affective (emotional) richness.

For example, suppose the function of pain is to represent information of the form “that’s bad.” We should then expect there to be a correlation between pain and organisms having representational information of various kinds. We should then expect pain to correlate with:

1. Cross-modal learning: being able to integrate multiple different senses into one broad picture of the world. Greater cross-modal learning=greater representation of the world=greater evolutionary function of pain.

2. A not B detour test, where an organism has to retrace its steps to get reward. If it does this, this raises the probability of it having a rich, integrated model of the world which makes pain serve a greater evolutionary function.

3. Anxiety-like behavior: if an organism displays strong indications of anxiety, this seems indicative of it having various extremely negatively-valenced experiences.

4. Valuing behavior: if an organism seems to value various things, this increases the likelihood that it phenomenologically represents information about how good and bad various states of the world are, and thus feels pain.

5. Prioritization of pain response: if an organism seems to prioritize pain response over other important evolutionary functions, then that seems to indicate it is feeling more intense distress—enough to overcome, in various contexts, its evolutionary pre-programming.

6. Modification of behavior in response to painkillers: if an organism behaves drastically differently after being given painkillers, then that seems to indicate it was previously in a lot of pain. Now, painkillers sometimes have other side effects (e.g. reducing motion) but if an organism displays drastically modified behavior in response to painkillers that can’t be explained by side effects (e.g. greater feeding, less nursing of the relevant wound) that is evidence that they were previously in lots of pain. Painkillers work in humans by eliminating our pain—it would be quite a coincidence if, though bees didn’t feel pain, they produced the same kinds of behavioral modification as diminished pain would. Thus, response to painkillers is quite good evidence that an animal feels pain.

Then they looked at eight different proxies for the complexity of mental representation—those being:

1. Brain mass.

2. Brain mass to body mass ratio

3. Brain volume

4. Critical flicker-fusion frequency in Hz (a measure of how rapidly organisms experience).

5. Encephalization quotient: “In lay terms, EQ is a measure of the deviation in brain size of a given species from a ‘standard’ species of the same taxon (RP).”

6. Neuron packing density

7. Total neurons

8. Working memory load which “refers to the amount of information a particular cognitive system can make accessible for a temporary amount of time.”

They then constructed twelve different models of welfare range from the relevant proxies and took an average across the models (given that it’s not obvious which model is correct) and explained in detail why they didn’t just rely on simpler proxies (like neuron counts). The report was compiled by a very large number of experts, over the course of months, averaging their findings given disagreement. One of the researchers, Bob Fischer (really the Bobby Fischer of welfare ranges, minus Bobby’s later descent into madness) said that compiling the report majorly changed his view on insect importance, so he started thinking insects were more important than he previously believed.

Now, you can definitely disagree with the report at the margins. I think it probably gave an underestimate of sentience in other mammals, for instance. But the idea that we should blithely dismiss the results of the report because they conflict with our naive intuitions about how intensely bees suffer strikes me as completely insane.

What in the world are our naive intuitions about bees based on that should give us any reason to give them much weight? We don’t really see bee behavior much. They’re too small for us to notice their behavior. Based on the behavior that we observe, bees don’t look that much less complicated than squirrels. This isn’t because they aren’t less complicated than squirrels—of course they are—but because we don’t get any great insight into bee complexity from looking at them. When we look at bees, just like when we look at squirrels, they’re mostly just moving around and performing their natural behavior.

Why is it that people are so confident that bees are not very conscious? I think the answer is: they’re small! It’s hard to imagine a thing that small could feel so intensely.

There have been very large insects—they’ve just mostly died off. Look at the prehistoric dragonfly below, for instance:

Now, if you swatted that dragonfly with a net, and watched it writhe around on the ground for several minutes before eventually dying, would anyone think it blindingly obvious that its experiences were no more than 1% as intensely unpleasant as a human’s experience? How in the world would one have any information about that? But bees are more mentally sophisticated than dragonflies—even large ones. They just have the defect of being small. But the mere fact that a creature is small tells us little about how intensely they experience.

Thus, I think our direct intuitions about bee consciousness are wildly unreliable and largely due to the fact that bees are tiny, caring about them is inconvenient, and we can’t really see their behavior that looks symptomatic of pain because they’re too small for it to be noticeable. If bees were 1,000 times larger, people wouldn’t have that intuition!

Bees probably have slightly greater cognitive sophistication than salmon. That’s why the report gave them a higher welfare range than it gave salmon. But few think it’s blindingly obvious that salmon only barely feel pain—well below 5% as intensely as humans. We should thus think the same about bees.

Now, does the 15% number (or the 7% number for the median estimate) pass the sniff test? I think the Rethink report is much more reliable than our direct intuitions. But my direct intuitions says that the number is pretty reasonable.

For one, we are really in the dark about consciousness. We don’t actually know which things give rise to consciousness in other animals. If you’re greatly uncertain about animal consciousness, then you shouldn’t confidently declare that it rounds down to zero. Maybe it does, but maybe it rounds up to be very significant.

One possibility is that all conscious creatures have equal welfare ranges. Now, this isn’t my preferred view. I don’t think it’s right. But it’s not so crazy as to be worth assigning nearly zero odds to. I’d give it maybe 3% odds. But that alone is enough to establish that bee welfare ranges are at least 3% that of humans on average—even ignoring all other possibilities of bee sentience.

In order to have a welfare range vastly below 15% on average, or 7% at the median, you must think it’s extremely obvious that organisms like bees—capable of transitive inference, pessimism bias, recognition of numbers, anxiety, novel responses to novel experience, and so on—have experiences that are barely intense at all. But given how mysterious consciousness is, what entitles us to such great confidence? Especially in light of the following facts:

1. Many of the Rethink models implied the presence of significant bee consciousness.

2. Bees when distressed certainly behave as if they’re in a lot of pain. They writhe around on the ground in agony and seem to struggle to a great degree.

3. There’s every evolutionary reason for bees to suffer intensely. If they only had mild pain experience, that wouldn’t produce as drastic of behavioral shifts.

4. Perhaps because of bees cognitive simplicity they need more intense pain to teach them a lesson, as Richard Dawkins has suggested.

5. When bees are in pain, given the simplicity of their brains, their pain likely takes up the totality of their experience. They can’t focus on anything else. This could make their pain very intense, as it means the only relevant signals their brains are taking in are pain signals.

6. There’s a long inductive trend of humans underestimating animal sentience. Until recently it was thought fish aren’t conscious (though that consensus has been overturned). In the 1980s, it was widely thought that animals don’t suffer at all. Similar, there was a longstanding belief newborn babies aren’t conscious.

In light of these factors, it seems wildly irresponsible to say that because bees don’t appear intensely conscious, on average their degree of valenced experience is far below 1% as intense as ours. Now, it could very well be that bees aren’t conscious at all or are only barely conscious. But being astronomically confident in such a proposition strikes me as totally unreasonable. Consciousness is the most mysterious thing in the universe—we are largely in the dark about the experience of other animals. We should not confidently declare that they’re only barely conscious! Especially if the basis for that judgment comes just from looking at them and is largely a byproduct of their diminutive size.

Now, lobsters aren’t the same as bees but they are somewhat similar. They’re both arthropods. Look at a lobster being boiled in a pot and tell me if it’s really so obvious that they’re only barely in pain. Certainly they behave as if they’re in a great deal of very intense pain, as they thrash around and try desperately to escape.

One of the posts that I listed before which I think most quintessentially captures what’s wrong with the way others were reasoning about bee consciousness was:

But obviously the right way to decide how intensely animals suffer is not to gauge how offensive it is to say animals are conscious. The facts don’t care about our feelings. Bees might be very conscious, and their interests might be important, even if it sounds offensive to say that they are comparably important to humans interests. Ethics should not be done by PR agencies. We shouldn’t revise our assessments of how many people die from Tuberculosis because it sounds offensive to say that tuberculosis is hundreds of times worse than 9/11.

But this seems basically like the methodology adopted by most of those who confidently declared bees to have supremely low welfare ranges. Because it sounds crazy that bees are intensely conscious, that their interests collectively dwarf our own, people blithely dismiss that possibility. They rejected intensely valenced bee experience not because of the evidence from bee behavior, but because of the implications that intensely valenced bee experience would have. But the facts of the world could turn out to be very inconvenient. You do not get to dismiss a factual conclusion because it would have troubling moral implications. You cannot derive an is from an ought.

In short, while bees and other insects might be only barely conscious, they are likely enough to be intensely conscious that we should take their interests seriously. Rethink Priorities did an admirable job estimating animal consciousness. While disagreements at the margins are perfectly alright, declaring their work obvious bullshit because it conflicts with your naive intuitions is insane and irresponsible. It is a somewhat concerning epistemological failure that lots of people reject the results of the most comprehensive report done on valenced experiences as obvious nonsense because it conflicts with their unreflective intuitions, dependent primarily on the observation that bees aren’t very large.