Ben Stevenson

• Ben Stevenson 7 Mar 2026 22:11 UTC

  1 point

  0 ∶ 1

  in reply to: Andrew Knight’s comment on: Should sus­tain­able pet diets be a lead­ing EA cause area?

  Thanks for your continued engagement, Andrew! I think we’re getting closer to agreement on the scale question :)

  We agree that further research is needed to take your animal-equivalent numbers and figure out the actual reduction in animal farming if demand for meat-based companion animal food declined. I had Perplexity take a first pass, looking at a 13% drop in meat-based dog food consumption in North America and Europe, and it suggested 60 million to 130 million animals would be saved.^[1] To be clear, that’s a first pass, and it’s still a lot of animals! I would be interested to see somebody look into this; I know @Billy Nicholles is thinking about it.

  This is reasonable, on the face of it. e.g. if dogs consume 2 million farmed land animals annually, then switching those dogs to nutritionally sound vegan diets would prima facie (i.e. on the face of it), spare 2 million farmed land animals from slaughter annually [...] Those who wish to argue that something other than the prima facie outcome would occur, face the burden of providing the reasoning and data to demonstrate and quantify this.

  I don’t know if you can call that a prima facie outcome; the counter-argument isn’t unlikely or complex, but a fairly basic claim about economics. But in any case, the burden of providing reasoning and data about how many animals would be saved lies with the researcher making claims about the “number of ‘food animals’ spared from slaughter” and trying to argue that vegan pet food is “a leading EA cause area”.

  (I do think that tractability is a more interesting and important crux, but Andrew has not replied to my comments so there’s nothing further to discuss. I found some of Billy and Seth’s points in the other thread constructive, though!)

  1. ^

     I wanted to look at a 13% drop to reflect Andrew’s tractability section, at dog food because vegan cat food is less reputable for health, and at North America and Europe because I’m suspicious that the breakdown about premium meat -vs- byproducts applies far outside the West. But to be clear, the drop from 7 billion is more due to economics than my assumptions; I also asked it to consider Andrew’s assumptions (complete drop in dog and cat food demand globally) and it suggested 260 to 670 million animals would be saved.

• Ben Stevenson 4 Mar 2026 21:27 UTC

  4 points

  0 ∶ 0

  in reply to: Vasco Grilo🔸’s comment on: Aqua­cul­ture frag­men­ta­tion is a port­fo­lio prob­lem CEAs miss

  A direct welfare intervention targeting Scottish farmed salmon could be more cost-effective in 2025 than 2000 as the population has increased even as global aquaculture has diversified.

  I agree there’s some important sense in which species diversification constrains how many species we can help with fixed resources (if species diversified and total population remained still, then intervention cost-effectiveness would decrease). I think “portfolio problem” gestures at how this phenomenon is hard to notice if we analyse the cost-effectiveness only at the level of direct welfare interventions, and don’t take a step back to look at how we distribute resources across welfare interventions or consider approaches that target the diversification problem itself (like the post’s recommendations).

• Ben Stevenson 2 Mar 2026 23:27 UTC

  2 points

  0 ∶ 0

  in reply to: David T’s comment on: Should sus­tain­able pet diets be a lead­ing EA cause area?

  Feels like tractability is the key point here. It doesn’t matter a huge amount if 7 billion is or isn’t the total amount of animals that would counterfactually be saved if all pets were fed vegan diets^[1]

  Yeah, I think this is true

• Ben Stevenson 2 Mar 2026 20:19 UTC

  6 points

  2 ∶ 0

  in reply to: Mo Putera’s comment on: Mo Putera’s Quick takes

  I’d be keen to get takes from folks in the know on what seems underfunded here. Farmed insects jump out: just $135k out of $260m overall (~0.05%) seems nuts.

  Aquatic animals look really underfunded (see Tom and Aaron’s post). But ‘number of individuals’ isn’t necessarily the best proxy for how much money should be spent; e.g., we probably shouldn’t be funding a welfare ask for every farmed fish out there until we have a better idea how to help them.

• Ben Stevenson 2 Mar 2026 20:03 UTC

  2 points

  0 ∶ 0

  in reply to: Seth Ariel Green 🔸’s comment on: Is EA sleep­ing on mak­ing dogs and cats ve­gan?

  I think you’re right in a theoretical marketplace: there is a demand signal here and we should expect production to decrease slightly.

  I don’t know that this stretches to the real marketplace. For one, it turns on how much less the next available use case is willing to pay, and historically it looks like the rendering industry has done well to find substitute uses (e.g., when synthetic soaps took off). I asked Perplexity, so fact-check this, but it suggests that slaughterhouses sell human-grade chicken for about $1.15 to $1.25 per lb, and they’d lose about $0.02 per lb on their by-product profits (<2% of the profit margin from their main product) if they switched from selling to pet food manufacturers to selling to pork farmers.

  For another, I think animals are lumpy good. Selling by-products does slightly increase the profit margin, but it’s a thin part of the overall profit margin, and I don’t know if losing that valorisation stream really shifts the economics for the farmer.

  Relatedly, you say

  The “byproduct” designation is convenient when distinguishing animal parts of different value, but at the end of the day, if you’re paying someone for animal parts, you’re paying them to slaughter animals

  I definitely see your point, but I wonder if there’s another sense where: biological waste is a ticking time bomb. It will smell and start to attract vermin and disease. From the slaughterhouses’ perspective, it’s amazing that somebody will pay to come in quickly and take that away. But the slaughterhouses would still swallow the cost, and probably not seriously consider killing fewer animals, if they had to dispose of that waste themselves.

• Ben Stevenson 2 Mar 2026 19:25 UTC

  13 points

  1 ∶ 0

  on: Should sus­tain­able pet diets be a lead­ing EA cause area?

  Thanks for this response, Andrew. It’s made me realise I was misrepresenting your methodology, although I still have reservations. I hope it wasn’t disheartening to see criticism of your work on the Forum, and that you stick around and continue to engage in good faith. I’m appreciative of your efforts to help the animals, and want to reassure you that the commentators here share your goal. Far from being speciesists or “seeking to undermine the science,” we’re a very pro-animal, and very truth-seeking, community.

  I think we have three disagreements here:

  • Whether or not vegan pet food is a leading effective altruist cause area

  I don’t think that it is because I don’t think it clears the bar on cost-effectiveness, or looks competitive with other EA animal welfare interventions.

  But I think we should have a fairly broad church, and a spirit of investigation, not outright dismissal, so I’m glad that you’ve looked into this. I’m open to considering other interventions in this space, like trying to move cats and dogs towards smaller-bodied animals.

  • Scale

  You’re right here: I’m wrong to describe this as double-counting as your methodology doesn’t literally count a cow, pig or chicken body twice over. I’ll edit my original comment accordingly.

  That said, I continue to disagree with your methodology; let me take another stab at explaining why I disagree, and I’ll be curious to hear your response.

  Your methodology is attributive, not consequential. It tells us what proportion of the world’s livestock are physically consumed by companion animals (based on some assumptions).^[1] But, as you acknowledge, much of this is the consumption of by-products. Your papers don’t look into the economics of by-products, so your methodology doesn’t tell us how many animals would be counterfactually spared farming and slaughter with vegan pet food.

  To be honest, I found your writing a bit misleading about this before (e.g., your paper has a section titled, “Number of ‘food animals’ spared from slaughter”). I find this comment really helpful and clarifying:

  The commentator also assumes the drivers of farmed animal production and slaughter are relevant to the calculations, and that the calculations are wrong because ABPs do not primarily drive farmed animal production. This position is also incorrect. The calculations make no assumptions about which factors or components drive farmed animal production. These factors are, in fact, not relevant to the calculations. The calculations rely solely on knowledge of the actual farmed animal species used within pet food, and the actual proportions of those carcasses that supply the meat and ABPs used, and the total average carcasses that are actually required. These are actualities, determined by physical realities relating to pet food ingredients, carcass proportions, and (simple) mathematics. They’re unaffected by drivers of production.

  I think our disagreements about your methodology are really less about nitpicking the details—but, again, I concede that I did misrepresent your methodology by calling it ‘double counting’ - and more about what the methodology should be trying to figure out. To me, the proportion of farmed animals eaten by companion animals is an interesting proxy for scale, but the really relevant question is: how many animals would be saved if the demand fell away? Farm animal advocacy should be about actually getting animals out of the food system (or improving conditions in factory farms).

  My current sense is that scale isn’t as high as you suggest not because I disagree with your allocation model, but because:

  1. I think farm animal by-products will be realloacted to other uses. If it’s true that, as your paper claims, “if not consumed within pet food, all meat ingredients, ABPs [animal by-products] and their derivatives, would normally be consumed either directly by people, or within other sectors of society” then the marginal use case of the by-product will change but production will not be seriously disrupted. (@Seth Ariel Green 🔸 points out that there could still be a demand signal encouraging a decrease in production; it still wouldn’t be a perfectly elastic signal).

  2. Even if farm animal by-products become waste products, that will reduce the efficiency of animal farming and so probably decrease production, but it wouldn’t be perfectly elastic. The production would drop only a little.

  The economic allocation model seems a step closer to considering actual economic demand than a model based on physical consumption. I think the reason I like the Alexander paper and you don’t is that I’m more interested in thinking about how production would change given demand shifts. (Economic allocation still seems to struggle thinking at the margin, and I would still like to see a system expansion model).

  • Tractability

  The surveys you cite were a slight update for me, but they’re not a knock-down argument. For one, I could present different data from the same surveys to colour the argument quite differently. Dodd, et al. tells us, for example, that 75% of omnivores say they’ll never feed their pet a plant-based diet, and majorities of omniores, pescetarians and vegetarians are concerned that plant-based diets for companion animals are unnatural, unhealthy or incomplete.

  But also, more importantly, surveys aren’t a perfect way to measure consumer intention (given, e.g., social desirability bias), consumer intention isn’t a perfect proxy for consumer behaviour (given, e.g., the intention-behaviour gap), and consumer behaviour isn’t the only thing that matters here (consider also, e.g., scaling up production; achieving cost competitiveness; securing support from vets; making vegan food that’s healthy for cats).

  I’m open to hearing out the tractability case, but I don’t think enough evidence has been presented so far.

  1. ^

     I’ll also concede that some of these assumptions are conservative. But some of them are also broad /​ arbitrary /​ hard to assess, like the conversion factor, which makes me take the whole exercise as one model, but not authoritative.

  What links here?

  • Ben Stevenson's comment on Is EA sleep­ing on mak­ing dogs and cats ve­gan? by Alistair Stewart (20 Feb 2026 8:49 UTC; 32 points)

• Ben Stevenson 2 Mar 2026 18:02 UTC

  2 points

  0 ∶ 0

  in reply to: Seth Ariel Green 🔸’s comment on: Should sus­tain­able pet diets be a lead­ing EA cause area?

  Plus one to this message on norms, as another of the commentators. Welcome to the forum, Andrew, thanks for being willing to discuss your piece

• Ben Stevenson 28 Feb 2026 11:32 UTC

  22 points

  0 ∶ 0

  on: Pete Hegseth de­clares An­thropic a sup­ply-chain risk

  Department of War: “We don’t trade with ants”

• Ben Stevenson 26 Feb 2026 11:19 UTC

  2 points

  0 ∶ 0

  on: An­i­mal Micro­grants: Re­quest for Proposals

  PSA: It’s the last few days to apply for a microgrant! Submit your idea now :)

• Ben Stevenson 20 Feb 2026 14:21 UTC

  5 points

  0 ∶ 0

  in reply to: Seth Ariel Green 🔸’s comment on: Is EA sleep­ing on mak­ing dogs and cats ve­gan?

  Thanks Seth (good to hear from you)!

  I agree with everything here. In particular: a lot of animals are farmed directly for pets, and ~3 billion sounds like a reasonable guess (I think Knight is double counting with ~7 billion). This was an update for me when I looked into this because I had thought almost all pet food was byproducts.

  And I might be understating the scale case because I’m bugged by the methodology. But I don’t think the scale case is compelling when we think in terms of animals who could reasonably be taken out of the food system by vegan pet food, instead of number of animals farmed for pet food. This is something that motivates my (stray) thought that promoting beef > chicken/​fish for companion animals could be high impact.

  I like the Alexander paper but I think economic allocation makes more sense when you’re trying to attribute environmental harms than when you’re trying to reason about the number of animals killed in a counter factual scenario. Looking at the tools from LCA, I’d like to see somebody try a system expansion model here!

  Please share the draft with me, I’d like to read it!

• Ben Stevenson 20 Feb 2026 8:49 UTC

  32 points

  3 ∶ 1

  on: Is EA sleep­ing on mak­ing dogs and cats ve­gan?

  [Edit: Andrew has responded to several commenters in a new Forum post. His post made me realise I was misrepresenting his methodology here, so I’ve struck those parts. I still disagree with his point about scale, but I’ve found a better way to express it here.]

  Thanks for sharing, Alistair! I recently spent some time looking into this issue, and I don’t think it’s a good EA cause area. Pet food does directly result in the farming of more livestock that I had expected, but Knight overstates his case and doesn’t engage with the rest of the literature.
  
  On scale:

  a full global transition would apparently spare around 7 billion land animals each year

  This is not true. The paper linked doesn’t explain Knight’s methodology, but if you click through to Knight (2023) you’ll find that the author double-counts a lot of livestock carcasses. Copy-pasting my earlier notes because I’m dashing this comment off, hope it makes sense:

  In particular, it claims: “Human meat consumption is responsible for this many animals being farmed directly. And pet food meat consumption is responsible for this many animals being farmed directly, and this many animals being farmed indirectly (i.e., byproducts)”. The total sum of animals therefore double counts the animals who are farmed for both human and pet consumption.

  […]
  
  Andrew’s model in the 2023 paper assumes that animals farmed for pet food ~= animals farmed directly + (ABPs [animal by products] * R), where R is basically the ratio of ABP:meat in an animal body. (I know I’m oversimplifying because he also includes a co-efficient to account for food waste, and so on). This formula means that (a) when put alongside animals farmed for human food, we’re double-counting the animals whose bodies mostly are eaten by people and minorly eaten by pets, and that (b) ABPs are always assumed to have a greater cost in animals lives than meat, which reverses the causal relationship. Byproducts don’t drive production.

  [Edit: I think it was inaccurate to call this double counting. My criticism is mostly about the economics of by-products, and I think I’ve expressed it better here].

  The byproducts point is important to appreciate whether, setting the numbers aside, vegan pet food would counter factually reduce the number of animals being farmed. Knight claims that there are other economically productive uses for byproducts; if that’s true, then a reduction in demand for animal-derived pet food would change the marginal use case for byproducts but not reduce their production. That undercuts Knight’s point on scale because so much of the scale falls out of inflated byproduct numbers.

  On tractability:

  I haven’t dug into the surveys that Knight cites but I’m super skeptical. I know vegans who don’t have vegan pets, and I know how hard it is to make people go vegan. There are big barriers to getting humans to transition to alternative proteins at scale, and that’s only more true for companion animals.

  On whether this is EA:

  I’m super pleased that some people have looked into this, and I think being able to entertain weird ideas seriously is a great thing. I even think there could be some cool stuff in this space (campaigning for cats and dogs to eat less chicken and fish?).

  But to be an EA issue — especially a ‘leading’ issue — it has to be more cost-effective than other good opportunities, and I don’t think it is.

• Ben Stevenson 2 Feb 2026 17:15 UTC

  5 points

  1 ∶ 0

  in reply to: Vasco Grilo🔸’s comment on: An­nounc­ing ACE’s 2025 Char­ity Recommendations

  That’s a really interesting blog, Vasco. Worth its own post.

• Ben Stevenson 2 Feb 2026 17:06 UTC

  7 points

  5 ∶ 0

  in reply to: Toby Tremlett🔹’s comment on: to­bytrem’s Quick takes

  Thanks Toby. The winning topic was, “Countering democratic backsliding is now a more urgent issue than more traditional longtermist concerns”?

  I feel like this doesn’t violate your rules, because it’s discussing something “directly connected to cause EA areas”. Democracy isn’t a top three cause area, but it’s not not a cause area. In any case, in my opinion the rules are obviously well-intentioned and mostly helpful, but at the end of the day it would be fine to bend them here.

• Ben Stevenson 1 Feb 2026 22:24 UTC

  5 points

  0 ∶ 0

  in reply to: Vasco Grilo🔸’s comment on: An­i­mal Micro­grants: Re­quest for Proposals

  Hey Vasco, thanks. I have also thought about this but don’t have a clear answer. Each credible grant opportunity will get a shallow dive but it’s hard to reliably estimate a total time budget before we have a good sense how many people will apply. Happy to share post-hoc reflections afterwards.

• Ben Stevenson 31 Jan 2026 23:15 UTC

  7 points

  2 ∶ 2

  in reply to: Thomas Kwa🔹’s comment on: Un­falsifi­able sto­ries of doom

  Thanks for the reply. I agree with your specific point but I think it’s worth being more careful with your phrasing. How much we earn is an ethically-charged thing, and it’s not a good thing if EA’s relationship with AI companies gives us a permission structure to lose sight of this.

  Edit: to be clear, I agree that “it’s probably not much more than he could earn elsewhere” but disagree that “Eliezer isn’t necessarily out of line in drawing $600k”

• Ben Stevenson 31 Jan 2026 10:29 UTC

  27 points

  18 ∶ 7

  in reply to: Thomas Kwa🔹’s comment on: Un­falsifi­able sto­ries of doom

  $235K is not very much money. […] $600K is also not much money.

  
  This is false.

• Ben Stevenson 30 Jan 2026 16:01 UTC

  4 points

  0 ∶ 0

  in reply to: Jim Buhler’s comment on: If wild an­i­mal welfare is in­tractable, ev­ery­thing is in­tractable.

  Thanks Jim, very interesting. I also feel conflicted, but lean towards taking A.^[1]

  Here’s how I feel about that:

  1. Bracketing feels strange when it asks us to be led by consequences which are small in the grand scheme (e.g., +/​- $1; Emily’s shoulder), and set aside consequences which are fairly proximate and which clearly dominate the stakes (e.g., +/​- <=$1000; killing the terrorist/​kid). It doesn’t feel so strange when our decision procedure calls on us to set aside consequences which dominate the stakes but don’t feel so proximate (e.g., longtermist concerns).

  2. When I look at very specific cases, I can find it hard to tell when I’m dealing with standard expected value under uncertainty, and when I’ve run into Knightian uncertainty, cluelessness, etc. I’m bracketing out +/​- <=$1000 when I say I take A, but I do feel drawn to treating this as a normal distribution around $0.

  Ways in which it’s disaalogous to animals that might be important:

  1. Animal welfare isn’t a one-shot problem. I think the best things we can do for animals involve calculated bets that integrate concern for their welfare into our decision-making more consistently, and teach us about improving their welfare more reliably.

  2. I’m not sure we should be risk-neutral maximisers for animal welfare.

  1. ^

     Conditional on being a risk neutral maximiser who values money linearly. In the real world, I’d shy away from A due to ambiguity aversion and because, to me, -$1000 matters more than +$1000.

• Ben Stevenson 27 Jan 2026 18:47 UTC

  2 points

  0 ∶ 0

  on: AWASH has offi­cially launched!

  This is great! Really exciting

• Ben Stevenson 27 Jan 2026 13:59 UTC

  2 points

  0 ∶ 0

  in reply to: Mark Westcombe’s comment on: Why I Steal (Ideas) and You Should Too

  Look forward to reading this blog, Mark!

An­i­mal Micro­grants: Re­quest for Proposals