Philosophy, global priorities and animal welfare research. My current specific interests include: philosophy of mind, moral weights, person-affecting views, preference-based views and subjectivism, moral uncertainty, decision theory, deep uncertainty/cluelessness and backfire risks, s-risks, and indirect effects on wild animals.
I’ve also done economic modelling for some animal welfare issues.
If using precise credences, then I’d be a strong longtermist (probably focusing on existential risks of some kind) or chase infinities. I haven’t thought a lot from this perspective about practical donation recommendations, if I’m assuming not suffering-focused. If suffering-focused (like I actually am), then probably CLR.
Yes, absolutely right about 0 being possible and reaonably likely. Maybe I’d say “average welfare conditional on having any welfare at all”. I only added that so that X% likely to be negative meant (100-X)% likely to be positive, in order to simplify the argument.
A Pascal’s mugging by nematodes? Nematodes as utility monsters?
Pascal’s bugging and the Rebugnant Conclusion (Sebo, 2024). :P
Interested to hear from Insect Welfare and Wild Animal Welfare advocates why they disagree that nematodes are the primary moral concern of planet Earth.
I’m sympathetic to difference-making risk aversion and difference-making ambiguity aversion (although see here) and assign nematodes a quite low probability of mattering much at all to me, low enough for now that I’m inclined to ignore them altogether (and what would have gone to nematodes instead goes to mitigating s-risks). Mites, springtails, copepods and insect larvae seem substantially more likely to matter to me, based on my beliefs about their capacities.
Still, I’d rather not go 100% on invertebrates either, also due to my difference-making sympathies. I’d deal with this like normative uncertainty and use a kind of bucket approach, like the Property Rights approach and hedging, with normative uncertainty about difference-making and approaches to dealing with uncertainty, about the nature of consciousness and moral patienthood and how to deal with it (although also see this), and about aggregation. So, roughly in practice, based on the probabilities of making a difference, probabilities of moral patienthood, attitudes towards risk and aggregation, I have a humans bucket, a mammals and birds bucket, a fish bucket, a shrimp and insects bucket, a mites, springtails and copepods bucket, and an s-risks bucket.
Another potentially useful takeaway is that these interventions Vasco considered, or at least diet change interventions like Veganuary and School Plates, are not robustly positive in expectation, when considering exactly the near-term animal effects. So why would we support them?
These interventions don’t seem justified by their direct cost-effectiveness, unless we have adequate reason to single out those effects and ignore or discount the effects on wild terrestrial invertebrates. We’d need a good reason to single out the direct effects, or refer to even more indirect or longer term reasons (e.g. moral circle expansion, space colonization and s-risks).
If you discounted nematodes >10x or more, then SWP’s HSI would come out ahead, or roughly tied with HIPF, right?
For the comparison to Shrimp Welfare Project’s Humane Slaughter Initiative, how long are you assuming the stunners are (counterfactually) used for? If I recall correctly, some prior estimates only assumed 1 year, which seems very conservative, and would probably make the comparisons to other opportunities here unfair.
It’s not so much that there’s a specific threshold away from 50%, it’s more that if you’re wildly uncertain and it’s highly speculative, rather than assigning a single precise probability like 55%, you should use a range of probabilities, say 40% to 70%. This range has values on either side of 50%. Then:
If you were difference-making ambiguity averse,[1] then both increasing their populations would look bad (possibly more bad lives in expectation) and decreasing their populations would look bad (possibly fewer good lives in expectation). You’d want to minimize these effects, by avoiding interventions with such large predictable effects on wild animal population sizes, or by hedging.
If you were ambiguity averse (not difference-making), then I imagine you’d want to decrease their populations. The worst possibilities for animals in the near-term are those where wild invertebrates are sentient and have horrible lives in expectation and you’d want to make those less bad. But s-risks (and especially hellish existential risks) would plausibly dominate instead, if you can robustly mitigate them.
On a different account dealing with imprecise credences, when we reduce their populations, you might say these wild animals are neither better off in expectation (in case they have good lives in expectation), nor are they worse off in expectation (in case they have bad lives in expectation), so we can ignore them, via a principle that extends the Pareto principle (Hedden, 2024).
(I’m assuming we’re ruling out an average welfare of exactly 0 or assigning that negligible probability, EDIT: conditional on sentience/having any welfare at all.)
On standard accounts of difference-making ambiguity aversion, which I think are problematic. I’m less sure about the implications of other accounts. See my 2024 post.
I’ll add that even if I would make different methodological choices, I think it’s still useful to highlight the scale of indirect effects on wild animals. The default in the community seems to be to ignore these effects, and there doesn’t seem to be good justification for that. I think it’s great that Vasco is taking these effects seriously and seeing where they might lead.
(And, as in my other comment, the conclusions and analysis could hold approximately anyway for those sufficiently pessimistic about the lives of wild invertebrates, or who give enough weight to sufficiently suffering-focused views.)
I agree both with the specific point about using LLMs and the more general point about sensitivity to highly speculative and ambiguous values. I would endorse imprecise credences, and the use of approaches to decision-making with imprecise credences. See also Anthony’s piece against precise Bayesianism.[1]
That being said, if you’re sufficiently suffering-focused or confident that their lives are negative on average (or confident that their lives are positive on average), then you don’t have to worry about this too much.
On difference-making ambiguity aversion as one natural group of approaches, see my 2024 post, my 2020 post and Greaves et al., 2022. I’m not confident these are the best approaches for dealing with imprecise credences (if averse to fanaticism).
I wonder how things would change if you had short AI timelines or otherwise assumed the impacts would be cut off after 5 or 10 years, say.
I’d imagine HIPF would get adjusted downward a lot, if and because these are policy interventions targeting hypertension, diabetes (like you mention) and other chronic lifestyle diseases. Not only would policy changes take time to pass and be implemented, I’d guess their effects on chronic lifestyle diseases would be fairly small for several years after implementation, because those diseases take time to reverse or prevent.
GiveWell top charities would get adjusted downwards because many of the estimated years of life saved could actually be lost to AI, or have radically different effects on wild animals. Instead of 50-70 years, we only get 5-10, say.
Broiler welfare reforms, especially the breed component which drives the wild animal effects here, tend to have phase-in periods of multiple years, and welfare reforms in general are often estimated to last several years. We might only have a few years of effects, if any.
I estimate donating to HIPF from CEARCH is 12 times as cost-effective as GiveWell’s top charities. This is the mean between the lower and upper bound of 9 and 15 mentioned by Joel Tan, CEARCH’s founder and managing director, on 28 May 2025. Joel estimated the cost-effectiveness accounting only for effects on humans, as a fraction of that of GiveWell’s top charities, of “advocacy for top sodium control policies to control hypertension” to be 31, that of “advocacy for sugar-sweetened beverages [SSBs] taxes to control diabetes mellitus type 2” to be 55, and that of donating to Giving What We Can (GWWC) in 2025 to be 13.
Is the cost-effectiveness based on increases in life expectancies only, or also improvements in quality of life? Increases in life expectancy would be the main drivers for effects on animals, and quality of life changes would probably have minimal effects on animals.
Thanks for writing this!
Even for fish like salmon that are larger than chickens, they still spend far more time being farmed than chickens farmed for meat, because chickens farmed for meat live such short lives, around 40-60 days. Then the days in captivity per kg of meat are much higher, too. See Welfare Footprint Institute’s chart here (direct link to the interactive chart):
That being said, my impressions, perhaps wrong, is that broilers have really bad lives throughout, experiencing substantial disabling pain. I’m less sure about how bad the lives of farmed fish are.
For example, a preference utilitarian can think there is an objective moral fact that it is bad to, all else equal, do something to someone that they disprefer, even though the “badness” comes from their dispreference. It seems like that would be subjective only if the claim was that it was bad according to some observer.
I’m personally not sympathetic to such a claim. What makes it objective? Rather, to me, it’s just bad to the person who disprefers it (and possibly other individuals). They are an observer. They are observers of their own mental states and things in the world, and they have attitudes about them.
The view I describe in this piece could be made objective in the way you describe, though.
Why do you believe that?
On the welfare effects of slower growing breeds, I think suffering is reduced overall despite the increase in life expectancies, based on Welfare Footprint Institute’s analysis:
Adoption of the Better Chicken Commitment, with use of a slower-growing breed reaching a slaughter weight of approximately 2.5 Kg at 56 days (ADG=45-46 g/day) is expected to prevent “at least” 33 [13 to 53] hours of Disabling pain, 79 [-99 to 260] hours of Hurtful and 25 [5 to 45] seconds of Excruciating pain for every bird affected by this intervention (only hours awake are considered). These figures correspond to a reduction of approximately 66%, 24% and 78% , respectively, in the time experienced in Disabling, Hurtful and Excruciating pain relative to a conventional scenario due to lameness, cardiopulmonary disorders, behavioral deprivation and thermal stress.
(...)
In general, the slower the growth rate, the shorter the cumulative time in pain experience over a lifetime, despite differences in lifespan. Should breeds with growth rates slower than those assumed in the reformed scenario be used, the time in pain averted with the reform would be longer, despite a longer lifespan. By the same logic, slower-growing breeds growing faster (e.g. 50g/day) should endure a longer time in pain, despite their shorter lifespan. In all cases, reforms promoting a transition to slower-growing breeds should expect a reduction of the cumulative time in pain (net positive change) for all breeds considered under the BCC scheme: the slower the growth rate, the higher the expected welfare impact.
Also, for those interested in animal welfare specifically: https://www.animaladvocacycareers.org/
Seems fine to direct people to 80,000 Hours for AI/x-risk, Animal Advocacy Careers for animal welfare and Probably Good more generally.
Couldn’t a country just opt out unilaterally, and then others follow suit? And should we trust their assessment of s-risks even if proceeding by global consensus?
You could defend the idea that extinction risk reduction is net negative or highly ambiguous in value, even just within EA and adjacent communities. Convincing people to not work on things that are net negative by your lights seems not to break good heuristics or norms.
When I edited this comment, it removed my vote percentage from it.
(Edited to elaborate and for clarity.)
Thomas (2019) calls these sorts of person-affecting views “wide”. I think “narrow” person-affecting views can be more liberal (due to incommensurability) about what kinds of beings are brought about.
And narrow asymmetric person-affecting views, as in Thomas, 2019 and Pummer, 2024, can still tell you to prevent “bad” lives or bads in lives, but, contrary to antinatalist views, “good” lives and goods in lives can still offset the bad. Pummer (2024) solves a special case of the Nonidentity problem this way, by looking at goods and bads in lives.
But these asymmetric views may be less liberal than strict/symmetric narrow person-affecting views, because they could be inclined to prevent the sorts of lives of which many are bad in favour of better average lives. Or more liberal, depending on how you think of liberalism. If someone would have a horrible life to which they would object, it seems illiberal to force them to have it.
I think these papers have made some pretty important progress in further developing person-affecting views.[1]
I think they need to be better adapted to choices between more than 2 options, in order to avoid the Repugnant Conclusion and replacement (St. Jules, 2024). I’ve been working on this and have a tentative solution, but I’m struggling to find anyone interested in reading my draft.
Hi Nicolas, thanks for commenting!
Ah, good point. (You’re assuming the separate components can be added directly (or with fixed weights, say).)
I guess the cases where you can’t add directly (or with fixed weights) involve genuine normative uncertainty or incommensurability. Or, maybe some cases of two envelopes problems where it’s too difficult or unjustifiable to set a unique common scale and use the Bayesian solution.
In practice, I may have normative uncertainty about moral weights between species.
If you’re risk neutral, probably. Maybe not if you’re difference-making risk averse. Perhaps helping insects is robustly positive in expectation, but highly likely to have no impact at all. Then you might like a better chance of positive impact, while maintaining 0 (or low) probability of negative impact.
For my illustration, that’s right.
However, my illustration treats the components as independent, so that you can get the worst case on each of them together. But this need not be the case in practice. You could in principle have interventions A and B, both with ranges of (expected) cost-effectiveness [-1, 2], but whose sum is exactly 1. Let the cost-effectiveness of B be 1-”the cost-effectiveness of A”. Having things cancel out so exactly and ending up with a range that’s a single value is unrealistic, but I wonder if we could at least get a positive range this way.
Ya, the default doesn’t seem privileged if you’re a consequentialist. See this post.