Thanks for the post.

In discussions of industrial animal agriculture, chickens dominate.

Rethink Priorities (RP) estimates 472 billion insects were slaughtered in 2023, which is 5.53 (= 472*10^9/(85.4*10^9)) times the number of all other land animals slaughtered that year. 

Nice point, Ozzie. I think wild animal welfare interventions would have to target aquatic or invertebrate animals in order to be as cost-effective as the best invertebrate welfare interventions. So I would like to have a clearer picture of how helping a few birds and mammals in urban areas leads to that.

Thanks for the conversation!

Likewise!

Thanks for sharing! If relevant to anyone seeing this, I am open to betting up to 10 k$ against short AI timelines, or large unemployment (my median unemployment rate in the United States in 2027 is something like 4 %).

Thanks, Stijn!

C: happy animal farming: buying and eating products from farms where animals live happy lives. This option generates the second highest total welfare, lower than option D but higher than A and B.

D: animal sanctuary: eating vegan and donating money to a sanctuary that breeds animals and give those animals long and very happy lives. This generates the highest welfare.

I think C would have higher welfare than D for the reasons I mentioned in my last comment. The welfare per animal-year in D would be higher than in C, but the population in D would be much lower than in C.

I am ignoring effects on wild animals, but I think these are the driver of the overall effect. I think decreasing the consumption of animal-based food tends to increase the population of wild animals due to decreasing cropland, and the counterfactual land having more wild animals. So I believe it is beneficial/harmful if wild animals have positive/negative lives.

So I assume you are an act utilitarian.

I confirm I am an act utilitarian (in terms of my moral theory), although I use rules/heuristic all time to make decisions (for example, being honest).

As a total act utilitarian, you should choose option A for yourself.

I agree I should follow a plant-based diet. I do not know whether this is good or bad for animals due to unclear possibly dominant effects on wild animals. However, it is cheaper and healthier, and therefore allows me to donate more to the most cost-effective invertebrate welfare interventions, which I think is the most important consideration for maximising my altruistic impact.

This brings me to the second issue: which option should you choose for other people? Most people currently choose option B. Persuading them to choose option C might be feasible with campaigns that require a bit of money, whereas shifting people to option D might be unfeasible, no matter how much money you throw at such a campaign. At first sight, total utilitarianism seems to suggest that the best strategy is to have people switch to option D, but that may be unfeasible or too costly. Note that in terms of total welfare, it is better to persuade people to choose option C instead of option A. 

B is the cheapest, and therefore allows for the largest population of farmed animals, and smallest population of wild animals. So, if these have negative lives, B can be the best. I have little idea about whether wild animals have good or bad lives, so I am mostly agnostic about whether decreasing the consumption of animal-based food is good or bad, regardless of whether it comes from farmed animals with positive or negative lives.

Most people have the moral intuition that the right to bodily autonomy trumps the right to live.

Killing a foetus violates their right to bodily autonomy? I agree most people support abortion, but there is still a significant fraction against it. It looks like 36 % in the United States think it should be illegal in all/most cases. A much smaller fraction of the population would be against killing farmed animals for food in all/most cases if they had good lives. So I think farming animals would be good under mild welfarism if the discounts to welfare were determined democratically.

If you endorse expectational total hedonistic utilitarianism, you should be against happy animal farming because those animals would be much better-off on a sanctuary where they live much longer with a higher welfare.

The annual welfare is proportional to the population (N), and welfare per animal-year. None of these factors should be maximised. A sufficiently large population would imply negative welfare per animal-year given a limited pool of resources. Having only a few animals with super high welfare would not be optimum due to implying a tiny population. I assume people would not want to farm lots of animals with super high welfare if they were not being raised for meat.

Here is a toy model about how to think about maximising welfare. If welfare per animal-year was proportional to the logarithm of the cost per animal-year (c = C/N) as a fraction of that for zero welfare (c_0), as self-reported life satisfaction is roughly proportional to the real gross domestic product (real GDP) per capita, the annual welfare would be proportional to N*ln(c/c_0). Note this does not explicitly depend on the lifespan of the animals. The ratio between annual welfare and cost, which is what I would want to maximise, would be proportional to N*ln(c/c_0)/C = ln(c/c_0)/c. This tends to 0 as c increases, which illustrates total welfare is not maximised by maximising the welfare per animal-year. The derivative of ln(c/c_0)/c with respect to c is (1 - ln(c/c_0))/c^2, which is 0 for ln(c/c_0) = 1, positive for ln(c/c_0) < 1, and negative for ln(c/c_0) > 1. So the ratio between annual welfare and cost would be maximum for ln(c/c_0) = 1, when the cost per animal-year is 2.72*c_0.

Eggs from barns cost around 1 $/kg in the European Union (EU), which means their production cost is lower than that (such that selling eggs is profitable), and I estimate hens in barns have roughly neutral lives. So the toy model above suggests the production cost which maximises welfare is lower than 2.72 $/kg. I have little confidence in this. I am just illustrating how I would think about it.

According to total utilitarianism, you may have a duty to sacrifice yourself by bringing into existence huge numbers of (slightly) happy beings and taking care of them.

I fully endorse expectational total hedonistic utilitarianism (ETHU), but I do not think it is that demanding. I think donating more and better is the best strategy to maximise impact for the vast majority of people, even ones working in impact-focussed organisations. Huge sacrifices like donating to the point of becoming extremely poor are a surefire way of not maximising future donations because they greatly decrease future earnings.

The fundamental difference between future human-years of existing and future people is based on the difference between existing and non-existing people. Existing people can complain against their welfare being discounted in those states where they exist, but possible future people who are never brought into existence, will never exist and hence cannot complain against their welfare being discounted in those states where they would exist. Hence, I don't think this difference based on existence is arbitrary. 

You seem to be implying that future human-years of existing people are necessary, and therefore can complain, whereas those of future people are not, and hence cannot complain. If so, what determines whether future human-years are necessary? Saying that future human-years are necessary if they belong to existing people would be assuming the conclusion.

I'm not against abortion (except very late term abortion, to some degree), because unwanted pregnancy violates a woman's right to bodily autonomy (that trumps the right to live, to some degree).

Why does women's right to bodily autonomy trump foetuses' right to life? I understand one can have a system where welfare is discounted in such a way that all one's intuitions are preserved, but I do not think this is a good way of figuring out one's moral theory. I prefer starting from the most fundamental intuitions, and then figure out what follows. Increasing happiness is fundamentally good to me, so I am open to the possibility of farming happy animals being good too (depending on the indirect effects).

I wonder, if you say happy animal farming is permissible, what do you think about happy slavery (breeding people to be used as slaves, guaranteeing a net-positive welfare) or happy human cannibalism?

I strongly endorse expectational total hedonistic utilitarianism, so I am in favour of increasing impartial welfare regardness of what it entails. I doubt human slavery or cannibalism would increase welfare, but I would support them if they did (accounting for direct and indirect effects).

Future human-years can only be discounted (in this context) if that discounting results in choosing the state where those humans do not exist.

I feel like this rule is arbitrary, as I do not see any fundamental difference between future human-years of existing and future people.

Thanks for clarifying, Hannah! As a result, I plan to recommend donating to RP's work on invertebrate welfare going forward together with my other top recommendations.

Thanks for the clarifications.

We see cost-effectiveness analysis as a valuable tool, but in some cases, it requires so many uncertain assumptions that it risks obscuring more than it reveals. In certain cases, two reasonable models can differ so much that one suggests a massive impact, and the other little to no impact.

Our hope is that donors to use cost-effectiveness estimates as one input, rather than treating them as final scores. Our goal isn’t to say which charity is “best,” but to offer high-quality information that empowers donors to make informed decisions. 

You estimated the animals helped per $ for the Fish Welfare Initiative (FWI), and SWP. I think your analyses would be more valuable if you got estimates for the increase in welfare per $. I know one needs to make contentious assumptions to compute these, but I still think producing them is valuable such that people can see which organisations increase welfare the most per $ under their preferred assumptions. Making calibrated adjustments is harder without an underlying model. Animal Charity Evaluators (ACE) estimated the cost-effectiveness of some charities in terms of suffering averted per $, which I believe is quite similar to welfare increased per $^[1].

Our goal isn’t to say which charity is “best,” but to offer high-quality information that empowers donors to make informed decisions.

I encourage you to clarify which are your criteria for recommending a charity. I think impact-focussed evaluators avoid large differences in cost-effectiveness among their recommendations^[2]. So people may infer you do not think there are large differences in cost-effectiveness among your recommendations.

1. ^{^}

   I think the impact of the interventions ACE assessed mostly comes from decreasing suffering, not increasing happiness.

2. ^{^}

   Vince Mak, ACE's charity evaluations manager, said "We do not recommend charities if there is a large enough gap between their expected marginal cost-effectiveness and that of our other charities". GiveWell's top charities save a life for 3.5 k to 5.5 k$.

Thanks for sharing!

The Shrimp Welfare Project (SWP) is one of my top recommendations too. Do you have any thoughts on my estimate that the cost-effectiveness of Fish Welfare Initiative’s (FWI’s) farm program from January to September 2024 was only 0.0111 % of SWP's past cost-effectiveness? I guess FWI will become more cost-effective in the future, but I do not see how they would come close to SWP.

I am surprised you are recommending the Animal Legal Defense Fund (ALDF) without doing any cost-effectiveness analysis of their work. Here is a quick illustration of why I think they are much less cost-effective than SWP:

• SWP has helped 1.50 k shrimp/year/$.
• You say that, "since 2021, ALDF has had average annual expenditures of $16,189,282". So they would have to have helped 24.3 billion animals per year (= 1.50*10^3*16.2*10^6) to have helped as many animals per year per $ as SWP.
• 9.91 billion land animals were slaughtered for meat in the United States (US) in 2023. So ALDF would have to have helped 2.45 (= 24.3*10^9/(9.91*10^9)) times as many animals per year as those slaughtered for meat in the US in 2023 to have helped as many animals per year per $ as SWP. This seems way too much to be feasible considering their interventions do not target aquatic animals, or land invertebrates.
• I understand impact is not directly proportional to the number of animals helped, but I believe the impact per animal helped may well be larger for SWP than ALDF. I estimate SWP averts the equivalent of 0.0426 DALYs per shrimp helped, which I think is as good as preventing the existence of 0.463 broilers (= 0.0426/(0.754*0.122)) in a conventional scenario.