Thanks for the post, Adam. I have eaten fully plant-based for 6 years, and haveargued for GiveWell considering effects on farmed animals. However, I now think it would be better for GiveWell to keep focussing just on humans. I believe considering effects on farmed animals would tend to decrease agricultural-land-years via decreasing animal farming, and therefore decrease animal welfare due to increasing the suffering of soil animals much more than it could decrease the suffering of farmed animals (for my best guess that soil animals have negative lives). Moreover, saving human lives cheaply is the most cost-effective way of increasing animal welfare I am aware of, and GiveWell considering effects on animals would tend to decrease the funds going towards that due to alienating some of its funders.
While I appreciate your concern for animal welfare, I think it is dubious with a large backfire risk to suggest this. Getting accidental benefits might be expedient in the short term (by accident) but at the risk of not making the requisite advances to get these benefits. I would see it as equivalent to guessing the correct answer to a math question. While it may be correct, you didn’t develop the tools to get there and thus are not likely to get good answers in the future.
As an example, I could imagine Givewell finding that charities that limit human births to be the most cost effective and this would backfire against you.
Thanks, Marcus. In any case, I think the effects on soil animals (positive or negative) are much larger than those on farmed animals. So I would say it would make much more sense for GiveWell to account for soil animals than farmed animals. As a starting point, they could estimate the cropland- and pasture-years per $ for the interventions they fund.
Hi Vasco, I have not read everything you have written on this topic in detail so forgive me if I have missed you addressing this somewhere.
It seems reasonable to me to claim that the welfare of soil animals can dominate these calculations. But, as you have noted, the action-relevance of this depends entirely on if soil animals live positive or negative lives. From what I’ve seen, you outsource this determination to the Gemini LLM. It doesn’t seem appropriate to me to outsource such a difficult question to an LLM. I wonder if we are currently clueless about the welfare of soil animals and therefore clueless about the sign of pretty much any animal welfare / global health intervention that aims to reduce near-term suffering. What do you think?
There is more than exactly zero evidence about whether soil animals have negative or positive lives. In this sense, I would not say we are clueless about it. I agree the uncertainty is very large, to the point I guess the probability of any intervention being beneficial/harmful is close to 50 % due to the probability of soil animals having positive/negative lives being close to 50 %. One can avoid this problem with a best guess that soil animals have neutral lives in expectation, but I do not think this is reasonable. It would be a huge coincidence because there are lots of positive and negative values, but a single neutral value (0).
In one of my posts, I used guesses from Gemini for the welfare per animal-year of soil animals as a fraction of the welfare per animal-year of fully healthy soil animals. However, I would now guess soil animals to have negative lives regardless of Gemini’s or other LLMs’ guesses. My sense is that most people working on wild animal welfare would guess soil animals have negative lives. In addition, Karolina Sarek, Joey Savoie, and David Moss estimated in 2018, based on a weighted factor model, that wild bugs have a welfare per animal-year equal to −42 % of that of fully happy wild bugs. In my last post about soil animals, I assumed −25 %, which is less negative than they supposed.
It would be great if Rethink Priorities (RP), the Welfare Footprint Institute (WFI), Wild Animal Initiative (WAI), or others investigated whether soil animals have positive or negative lives. I would be happy to donate myself. I emailed and tagged people from those organisations about this, but only Cynthia Schuck‑Paim from WFI replied, saying Wladimir Alonso from WFI is working on a project related to assessing differences in hedonic capacity, which I guess relates to this post.
estimate the cropland- and pasture-years per $ for the interventions they fund.
What would they do with such an estimate? I don’t think anyone, you included, knows with any more than very slim confidence, if it’s good or bad for soil animals to turn wild land into cropland or vice versa.
I would support interventions resulting in more m2-years of cropland and pasture per $. I guess soil animals have negative lives, and cropland and pasture are the 2 biomes besides desert with the least soil arthropods per unit area according to the means in Table S4 of Rosenberg et al. (2023), so I think increasing cropland and pasture implies less soil animals with negative lives in expectation.
I am not confident at all about whether soil animals have negative or positive lives. I havebeenhighlightingthat decreasing the uncertainty about this would be great. However, I still recommend interventions based on my best guess. I endorse maximising expected welfare (I see any alternatives as way worse), and I believe the expected effects on soil animals are much larger than those on target beneficiaries for the vast majority of interventions, so it makes sense I account for effects on soil animals despite their uncertainty.
Sorry for being this blunt, but EA is about using evidence and reason to identify the most effective ways to help others. I can’t possibly see how operating on a vague guess is on par with that.
This criticism is independent of the fact that I still claim a “negative life” is not a concept we should incorporate into moral theories, and that we definitely shouldn’t aim to just cull all animals whose lives we somehow think are negative.
My sense is that most people working on wild animal welfare would guess soil animals have negative lives. In addition, Karolina Sarek, Joey Savoie, and David Moss estimated in 2018, based on a weighted factor model, that wild bugs have a welfare per animal-year equal to −42 % of that of fully happy wild bugs. In my last post about soil animals, I assumed −25 %, which is less negative than they supposed.
It’s at least somewhat germane here. As I understand Adam’s post, he is urging GiveWell to weight certain types of animal-welfare harm in its analyses. But that would make GiveWell’s work less valuable to many people whose views materially differ in certain ways from whatever specific views and weights regarding animal welfare GiveWell incorporated into its analyses. I think Vasco’s views represent a valid (albeit unusual) example of those circumstances for a specific potential(?) donor.
I downvoted. Saying that you’re downvoting with a smiley face seems overly passive aggressive to me. Your comment also doesn’t attempt to argue any point, and I believe when you have done so in the past you have failed to convince Vasco, so I’m not sure what use these comments serve.
I also personally think that Vasco raises a very important consideration that is relevant to any discussion about the cost effectiveness of both animal welfare and global health interventions. I’m not sure what the conclusion of considering the welfare of soil animals is, but it’s certainly given me food for thought.
Thanks, Nick. I do not think this is a general animal welfare thread. The post is arguing for GiveWell considering effects on animals, and my comment relates to this.
Thanks, Vasco. I have no problem at all with GiveWell focusing on human welfare. I am opposed to donations of animals as a means of doing so, though, and I think that GiveWell should only consider this if they update their moral weights to account for the welfare of animals.
Thanks for the post, Adam. I have eaten fully plant-based for 6 years, and have argued for GiveWell considering effects on farmed animals. However, I now think it would be better for GiveWell to keep focussing just on humans. I believe considering effects on farmed animals would tend to decrease agricultural-land-years via decreasing animal farming, and therefore decrease animal welfare due to increasing the suffering of soil animals much more than it could decrease the suffering of farmed animals (for my best guess that soil animals have negative lives). Moreover, saving human lives cheaply is the most cost-effective way of increasing animal welfare I am aware of, and GiveWell considering effects on animals would tend to decrease the funds going towards that due to alienating some of its funders.
While I appreciate your concern for animal welfare, I think it is dubious with a large backfire risk to suggest this. Getting accidental benefits might be expedient in the short term (by accident) but at the risk of not making the requisite advances to get these benefits. I would see it as equivalent to guessing the correct answer to a math question. While it may be correct, you didn’t develop the tools to get there and thus are not likely to get good answers in the future.
As an example, I could imagine Givewell finding that charities that limit human births to be the most cost effective and this would backfire against you.
Thanks, Marcus. In any case, I think the effects on soil animals (positive or negative) are much larger than those on farmed animals. So I would say it would make much more sense for GiveWell to account for soil animals than farmed animals. As a starting point, they could estimate the cropland- and pasture-years per $ for the interventions they fund.
I think this is more sensible.
Hi Vasco, I have not read everything you have written on this topic in detail so forgive me if I have missed you addressing this somewhere.
It seems reasonable to me to claim that the welfare of soil animals can dominate these calculations. But, as you have noted, the action-relevance of this depends entirely on if soil animals live positive or negative lives. From what I’ve seen, you outsource this determination to the Gemini LLM. It doesn’t seem appropriate to me to outsource such a difficult question to an LLM. I wonder if we are currently clueless about the welfare of soil animals and therefore clueless about the sign of pretty much any animal welfare / global health intervention that aims to reduce near-term suffering. What do you think?
Hi Jack,
There is more than exactly zero evidence about whether soil animals have negative or positive lives. In this sense, I would not say we are clueless about it. I agree the uncertainty is very large, to the point I guess the probability of any intervention being beneficial/harmful is close to 50 % due to the probability of soil animals having positive/negative lives being close to 50 %. One can avoid this problem with a best guess that soil animals have neutral lives in expectation, but I do not think this is reasonable. It would be a huge coincidence because there are lots of positive and negative values, but a single neutral value (0).
In one of my posts, I used guesses from Gemini for the welfare per animal-year of soil animals as a fraction of the welfare per animal-year of fully healthy soil animals. However, I would now guess soil animals to have negative lives regardless of Gemini’s or other LLMs’ guesses. My sense is that most people working on wild animal welfare would guess soil animals have negative lives. In addition, Karolina Sarek, Joey Savoie, and David Moss estimated in 2018, based on a weighted factor model, that wild bugs have a welfare per animal-year equal to −42 % of that of fully happy wild bugs. In my last post about soil animals, I assumed −25 %, which is less negative than they supposed.
It would be great if Rethink Priorities (RP), the Welfare Footprint Institute (WFI), Wild Animal Initiative (WAI), or others investigated whether soil animals have positive or negative lives. I would be happy to donate myself. I emailed and tagged people from those organisations about this, but only Cynthia Schuck‑Paim from WFI replied, saying Wladimir Alonso from WFI is working on a project related to assessing differences in hedonic capacity, which I guess relates to this post.
What would they do with such an estimate? I don’t think anyone, you included, knows with any more than very slim confidence, if it’s good or bad for soil animals to turn wild land into cropland or vice versa.
Hi Guy,
I would support interventions resulting in more m2-years of cropland and pasture per $. I guess soil animals have negative lives, and cropland and pasture are the 2 biomes besides desert with the least soil arthropods per unit area according to the means in Table S4 of Rosenberg et al. (2023), so I think increasing cropland and pasture implies less soil animals with negative lives in expectation.
I am not confident at all about whether soil animals have negative or positive lives. I have been highlighting that decreasing the uncertainty about this would be great. However, I still recommend interventions based on my best guess. I endorse maximising expected welfare (I see any alternatives as way worse), and I believe the expected effects on soil animals are much larger than those on target beneficiaries for the vast majority of interventions, so it makes sense I account for effects on soil animals despite their uncertainty.
Sorry for being this blunt, but EA is about using evidence and reason to identify the most effective ways to help others. I can’t possibly see how operating on a vague guess is on par with that.
This criticism is independent of the fact that I still claim a “negative life” is not a concept we should incorporate into moral theories, and that we definitely shouldn’t aim to just cull all animals whose lives we somehow think are negative.
My sense is that most people working on wild animal welfare would guess soil animals have negative lives. In addition, Karolina Sarek, Joey Savoie, and David Moss estimated in 2018, based on a weighted factor model, that wild bugs have a welfare per animal-year equal to −42 % of that of fully happy wild bugs. In my last post about soil animals, I assumed −25 %, which is less negative than they supposed.
Strong downvote for the usual reasons brother :). Doing think it’s helpful to always post this argument on on general animal welfare threads.
It’s at least somewhat germane here. As I understand Adam’s post, he is urging GiveWell to weight certain types of animal-welfare harm in its analyses. But that would make GiveWell’s work less valuable to many people whose views materially differ in certain ways from whatever specific views and weights regarding animal welfare GiveWell incorporated into its analyses. I think Vasco’s views represent a valid (albeit unusual) example of those circumstances for a specific potential(?) donor.
I downvoted. Saying that you’re downvoting with a smiley face seems overly passive aggressive to me. Your comment also doesn’t attempt to argue any point, and I believe when you have done so in the past you have failed to convince Vasco, so I’m not sure what use these comments serve.
I also personally think that Vasco raises a very important consideration that is relevant to any discussion about the cost effectiveness of both animal welfare and global health interventions. I’m not sure what the conclusion of considering the welfare of soil animals is, but it’s certainly given me food for thought.
Thanks, Nick. I do not think this is a general animal welfare thread. The post is arguing for GiveWell considering effects on animals, and my comment relates to this.
Thanks, Vasco. I have no problem at all with GiveWell focusing on human welfare. I am opposed to donations of animals as a means of doing so, though, and I think that GiveWell should only consider this if they update their moral weights to account for the welfare of animals.