On the longtermist case for working on farmed animals [Uncertainties & research ideas]

I also considered the following title for this post, which might be more fitting: Does expanding moral circles to one type of being also expand them to other types of beings?


Some people think that longtermists should prioritise work focused on farmed animals in the near-term future. The argument for this typically includes the premises that the vast majority of all the suffering and wellbeing that ever occurs might be experienced by beings which humans might have little to no moral concern for (e.g., artificial sentient beings), and that some work focused on farmed animals could increase the chance that humans have moral concern for those beings.

I find this argument and conclusion plausible, but also quite speculative. Part of my uncertainty has to do with whether expanding moral circles to include farmed animals would also expand them to include the relevant other types of beings, and whether it’d do so more effectively than other actions would.

Below, I outline several ideas for research projects that someone could do to reduce our uncertainty on those points. These include reviews of relevant literature, expert elicitation, surveys, experiments, and historical research.

Note: This post is adapted from a doc I wrote around July 2020. Since then, I’ve become somewhat less confident about the importance of the ideas raised and projects suggested here (though I still think they might be important). And if I were writing this from scratch now, I’d probably frame it somewhat differently.

I wrote this post in a personal capacity, and it does not necessarily represent the views of any of my employers.

My thanks to Tobias Baumann and Megan Kinniment for helpful comments on an earlier draft of this post.

A longtermist argument for farmed animal welfare work

Some people think that longtermists should prioritise work aimed at reducing near-term farmed animal suffering, and some people indeed seem to be doing such work for longtermist reasons. For example, I think this roughly describes the Sentience Institute’s views. (See also posts tagged Non-Humans and the Long-Term Future.)

I think that that position is typically based on something like the following argument (see, e.g., the post Why I prioritize moral circle expansion over artificial intelligence alignment):

  • Premise 1: It’s plausible that the vast majority of all the suffering and wellbeing that ever occurs will occur more than a hundred years into the future and will be experienced by beings towards which humans might “by default” have little to no moral concern (e.g., wild animals on terraformed planets; artificial sentient beings).

  • Premise 2: If Premise 1 is true, it could be extremely morally important to—either now or in the future—expand moral circles such that they’re more likely to include those types of beings.

  • Premise 3: Such moral circle expansion (MCE) may be not just important but also urgent. This is because there may be a “value lock-in” relatively soon, for example due to some ways the development of transformative artificial intelligence (TAI) may play out.

  • Premise 4: If more people’s moral circles expand to include farm animals and/​or if factory farming is ended, that would increase the chances that future actors’ moral circles will include all sentient beings (or at least all the very numerous beings).

  • Conclusion: Work that supports the expansion of people’s moral circles to include farm animals and/​or supports the ending of factory farming could therefore be both (a) extremely morally important and (b) urgent.

(Of course, one could also arrive at the same conclusion via different arguments, including ones that entirely focus on the intrinsic significance of near-term suffering and wellbeing. See also this comment thread. Also, one could use a longtermist argument similar to the above one in order to argue for focusing on near-term wild animal welfare work; I’ll return to this point below.)

Uncertainties about Premise 4

Personally, I find the four premises above plausible, along with the conclusion.

However, each of those premises also seems quite speculative. This is often hard to avoid, especially in longtermism. But it seems to me that there are tractable ways to improve our knowledge regarding Premise 4 (and related matters), and that the value of information from doing so would be very high. So I’ve generated some ideas for research on Premise 4 which I think it might be worthwhile for someone to pursue, which I’ll describe below.[1]

Essentially, I’d be very confident that Premise 4 was true if moral circles were effectively “unidimensional”. However, it seems to make more sense to think of moral circles as multidimensional, such that a person’s moral circle can expand along one dimension without expanding along others, and that two people could have differently shaped “moral circles”, without it being clear whose is “larger”. (See Moral circles: Degrees, dimensions, visuals for elaboration on these points.)

Thus, it seems plausible that expanding a person’s moral circle to include farm animals doesn’t bring the “boundary” of that person’s moral circles any “closer” to including whatever class of beings we’re ultimately concerned about (e.g., wild animals or artificial sentient beings). Furthermore, even if expanding a person’s moral circle to include farm animals does achieve that outcome, it seems plausible that that the outcome would be better achieved by expanding moral circles along other dimensions (e.g., by doing concrete wild animal welfare work, advocating for caring about all sentient beings, or advocating for caring about future artificial sentient beings).[2]

Research on these points could have major implications for which research and interventions should be prioritised by people focused on animal welfare work, MCE, and/​or benefitting non-humans in the long-term future.

Additionally, research on these points could suggest that some people who aren’t focused on animal welfare work, MCE, and/​or benefitting non-humans in the long-term future should focus on those things. This would occur if the research ends up providing further support for Premise 4 and/​or revealing more cost-effective interventions for long-term-relevant MCE than near-termist work on farmed animal welfare.

(I’m not the first person to have considered roughly these sorts of points. For example, there are somewhat similar ideas in this talk by Jacy Reese.)

How could we reduce those uncertainties?

I think it might be worthwhile for someone to conduct research aimed at answering the question of how interventions that expand moral circles along certain dimensions (or to certain types of beings) spill over into expanding moral circles along other dimensions (or to other types of beings).

This question could be tackled:

  • Relatively directly, though that might require expensive experiments; or

  • Somewhat indirectly, by investigating how expansions of moral circles (whether or not they’re caused by “interventions”) along certain dimensions spill over into expanding moral circles along other dimensions; or

  • Even more indirectly, by addressing the purely correlational question of how well the size of a person or group’s moral circles along one dimension predicts the size of their moral circles along another dimension

Ideally, this research would focus on:

  • The types of interventions that EAs (or related groups) are most likely to actually consider supporting

  • The types of “audiences” these interventions are most likely to target (e.g., the general public, AI researchers)

  • The types of beings those interventions are most likely to directly focus on (e.g., farmed animals)

  • The types of beings it might be ultimately most important to expand moral circles to include (e.g., wild animals, artificial sentience)

One could also investigate how results differ depending on differences in intervention types, types of audiences, and types of beings. This could inform decisions about things like:

  • Whether to prioritise clean meat research or advocacy against speciesism

  • Whether to target thought leaders, tech researchers, or the general public

  • Whether to focus on expanding moral circles to include farm animals, insects, wild animals, “all sentient beings”, or artificial sentient beings

Note that:

  • I won’t be pursuing those questions myself, as I’m busy with other projects.

  • It’s possible that some of the work I propose here is already being done.

Sketches of more specific possible research projects

1. Reviews of literature relevant to the above questions

The relevant literature might be mostly from psychology, history, sociology, and effective altruism.

For example, I know of at least one paper relevant to the extent to which inclusion of some entities in one’s moral circles predicts inclusion of other entities. I suspect there are also others. And the research on secondary transfer effects seems relevant too (my thanks to Jamie Harris for drawing my attention to that).

For another example, I suspect some writings on the history of MCE would contain clues as to how often each of the following things occur:

  • Expansion along one dimension leads to expansion along another

  • Expansion along many dimensions happens near-simultaneously due to some other underlying cause (e.g., economic growth)

  • Expansion occurs along one or more dimensions without occurring along (important) other dimensions

2. Expert elicitation focused on the above questions

This elicitation could be done via surveys and/​or via interviews.

The most relevant experts might be psychologists, sociologists, and historians who’ve published relevant research. Other types of people who might be relevant include EAs, animal advocates, futurists, and philosophers who’ve done relevant work.

3. Surveys focused on the above questions

These surveys would likely consist mostly of things like rating scale questions, though with at least some boxes for open-ended responses.

3a. Surveys simply focused on what types of entities people currently include in their moral circles (or related matters, like what entities they empathise with or eat).

Such surveys could provide evidence about Premise 4 because, if a person’s inclusion of one type of entity predicts their inclusion of other types of entities, this would push in favour of the hypothesis that moral circles are “effectively unidimensional”.

That said, that wouldn’t strongly indicate that expansion along one dimension will spill over into expansion along other dimensions. This is because the correlations could reflect how people’s moral circles started out (e.g. due to a genetic predisposition towards generalised empathy), rather than how they expanded.

(I think Lucius Caviola’s thesis “How We Value Animals: The Psychology of Speciesism” would be relevant here, though I haven’t actually read it myself.)

3b. Surveys asking people to recall what entities they included in their moral circles (or related matters) at various times.

This could provide slightly better evidence about how expansion along one dimension may or may not lead to expansion along other dimensions. But I don’t think I’d want to put much weight on self-reported distant memories.

3c. Longitudinal surveys on what entities people include in their moral circles (or related matters) at different times.

4. Experiments focused on the above questions

4a. Between-subjects experiments in which some participants are shown arguments, videos, or information which is intended to expand their moral circles to include a particular type of entity, and all participants are asked about which entities they include in their moral circles. The entities participants would be asked about would include ones not focused on by the arguments, videos, or information.

4b. Within-subjects experiments similar to the above, but with the intervention delivered to all participants, and participants being asked about their moral circles both beforehand and afterwards.

(These experiments are relevant inasmuch as Premise 4 focuses on expanding individual people’s moral circles to include farmed animals, rather than on ending factory farming as a whole. But some people might think the “active ingredient” in improving what values ultimately get “locked-in” in ensuring factory farming as a whole is ended by the time a value lock-in occurs. It would of course be very hard to run an experiment to relatively directly test that hypotheses.

5. Historical research focused on the above questions

Here I mean historical research other than just literature reviews, since I already discussed literature reviews earlier.

5a. Case studies.

For example, one could investigate the factors that seem to have led to a particular instance of MCE, and how many dimensions moral circles appear to expand along during that instance.

5b. Historical research using a more quantitative, macro approach, examining broader trends.

Some possible arguments against doing those research projects

  • Perhaps other people already have done or are doing similar research

  • Perhaps we can already confidently dismiss the above longtermist argument for farmed animal welfare work, without needing to do this research

  • Perhaps this research would give “pessimistic” results even if that argument was true, because the relevant effects on ultimate moral circles would only occur at a point such as the end of factory farming

  • Perhaps it’d be better to do some other research project focused on Premise 4 or one of the other premises

  • It would likely be hard to determine which of correlations these research projects identify are causal

  • Perhaps people’s current, self-reported attitudes on (e.g.) artificial sentient beings would bear little resemblance to their later, revealed attitudes

Possible next steps

If someone was interested in doing the sort of research I’ve proposed here, they might wish to take roughly the following steps, in roughly this order:

  1. Comment on this post

  2. Contact the Sentience Institute and/​or me to discuss ideas

  3. Do research along the lines of the first two project ideas listed above (partly to help orient themselves to the general area)

  4. Do research along the lines of the third project idea

  5. Maybe do research along the lines of the fourth and fifth project ideas, depending on whether that still seems worthwhile at that stage

See also Notes on EA-related research, writing, testing fit, learning, and the Forum.


[1] To be clear, I’m not saying I’m much more confident that the other premises are true than that Premise 4 is true, nor that research on the other premises wouldn’t be worthwhile. It’s just that I didn’t immediately see tractable ways to investigate those other premises. But see here for some other points about why moral circle expansion may be overrated and how one might investigate that matter.

[2] See also the post Beware surprising and suspicious convergence.