Michael St Jules 🔸

Karma: 12,670

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.

Want to leave anonymous feedback for me, positive, constructive or negative? https://www.admonymous.co/michael-st-jules

Michael St Jules 🔸 25 Jan 2026 4:28 UTC
2 points
0 ∶ 0
in reply to: Linch’s comment on: Linch’s Shortform
Here are a few things you might need to address to convince a skeptic:
1. Humans currently have access to, maintain and can shut down or destroy the hardware and infrastructure AI depends on. This is an important advantage.
2. Ending us all can be risky from an AI’s perspective, because of the risk of shutdown (or losing humans to maintain, extract resource for, and build infrastructure AI depends on without an adequate replacement).
3. I’d guess we can make AIs risk-averse (or difference-making risk averse) for whatever goals they do end up with, even if we can’t align them.
4. Ending us all sounds hard and unlikely. There are many ways we are resilient and ways governments and militaries could respond to a threat of this level.

Michael St Jules 🔸 15 Jan 2026 21:55 UTC
8 points
0 ∶ 0
on: Improving wild animal welfare reliably
Thanks for sharing!
We could identify the most severe diseases through research and target them with vaccination campaigns. Yes, eliminating a disease would lead to population growth until another factor limits the population—but since it was such an unusually severe limiting factor, the net effect is likely to be positive.
This is interesting and seems possible to me, but I’d probably want to look more into any particular case and see population modelling to verify the logic more generally.
If this does work, I wonder if we’d have a general and reliable path forward for reducing wild animal suffering (whether disease or another cause, it sounds like you’re more agnostic about it being diseases in the paper): just iteratively and incrementally reduce the causes of suffering in a population, roughly in order from the most severe (worst conditional on their occurrence^[1]) to the mildest.
1. ^
  However, the worst conditional on their occurrrence could be rare enough that this wouldn’t look cost-effective. In that case, we might look for another animal population where it does look cost-effective to reduce the most severe cause of suffering.

Michael St Jules 🔸 28 Dec 2025 5:25 UTC
4 points
0 ∶ 0
on: It’s Good To Create Happy People: A Comprehensive Case
Of interest, see the comments on Thornley’s EA Forum post. I and others have left multiple responses to his arguments.

Michael St Jules 🔸 28 Dec 2025 5:18 UTC
2 points
0 ∶ 0
on: It’s Good To Create Happy People: A Comprehensive Case
Here’s how I’d think about “4 My argument”, in actualist preference-affecting terms^[1]:
My preferences will differ between pressing button 1 and not pressing button 1, because my preferences track the world’s^[2] preferences^[3], and the world’s preferences will differ by Frank’s. Then:
1. If I know ahead of time that if I press button 1, I will definitely press button 2 (because I will take Frank’s interests into account), then, by backward induction, my options at the start are actually just pressing both and pressing neither. If I press neither button, it’s not worse to Bob, and Frank won’t exist, so it won’t actually be worse to him either. So pressing button 1 is not better or obligatory, because if I don’t press it, it’s not worse to anyone. This matches the permissibility of not pressing a button that has the net effects of 1 and 2 together.
2. If I know ahead of time there’s some chance I won’t press button 2 even if I press button 1, then pressing button 1 is better, because it would be worse in expectation to Bob if I didn’t.
1. ^
  See this piece for more on actualist preference-affecting views.
2. ^
  Past, current and future, or just current and future.
3. ^
  Or also desires, likes, dislikes, approval, disapproval, pleasures in things, displeasures in things, evaluative attitudes, etc., as in Really radical empathy.

Michael St Jules 🔸 28 Dec 2025 1:11 UTC
2 points
0 ∶ 0
on: It’s Good To Create Happy People: A Comprehensive Case
I’d guess this is pretty illustrative of differences in how we think about person-affecting views, and why I think violations of “the independence of irrelevant alternatives” and “Losers Can Dislodge Winners” are not a big deal:
Narrow views imply that in the choice between 1 and 2, you can choose either. But why in the world would adding 3, which is itself impermissible to take, affect that?
Run through the reasoning on the narrow view with and without 3 available and compare them. The differences in reasoning, ultimately following from narrow person-affecting intuitions, are why. So, the narrow person-affecting intuitions explain why this happens. You’re asking as if there’s no reason, or no good reason. But if you were sufficiently sympathetic to narrow person-affecting intuitions, then you’d have good reasons: those narrow person-affecting affecting intuitions and how you reason with them.
(Not that you referred directly to “the independence of irrelevant alternatives” here, but violation of it is a common complaint against person-affecting views, so I want to respond to that directly here.) 3 is not an “irrelevant” alternative, because when it’s available, we see exactly how it’s relevant when it shows up in the reasoning that leads us to 2. I think “the independence of irrelevant alternatives” has a misleading name.
Adding some other choice you’re not allowed to take to an option set shouldn’t make you no longer allowed to choose a previously permissible option. This would be like if you had two permissible options: saving a child at personal risk, or doing nothing, and then after being offered an extra impermissible option (shooting a different child), it was no longer permissible to do nothing. WTF?
And then this to me seems disanalogous, because you don’t provide any reason at all for how the third option would change the logic. We have reasons in the earlier hypothetical.

Michael St Jules 🔸 23 Dec 2025 21:29 UTC
7 points
1 ∶ 0
in reply to: Bentham's Bulldog’s comment on: It’s Good To Create Happy People: A Comprehensive Case
I think we have very different intuitions.
I don’t think giving up axiology is much or any bullet to bite, and I find the frameworks I linked
1. better motivated than axiology, and, in particular, by empathy, and to better respect what individuals (would) actually care about,^[1] which I take to be pretty fundamental and pretty much the point of “ethics”, and
2. better fits with subjectivism/moral antirealism.^[2]
The problems with axiology also seem worse to me, often as a consequence of failing to respect what individuals (would) actually care about and so failing at empathy, one way or another, as I illustrate in my sequence.
Giving up axiology to hold on to a not even very widely shared intuition?
What do you mean to imply here? Why would I force myself to accept axiology, which I don’t find compelling, at the cost of giving up my own stronger intuitions?
And is axiology (or the disjunction of conjunctions of intuitions from which it would follow) much more popular than person-affecting intuitions like the Procreation Asymmetry?
Giving up the idea that the world would be better if it had lots of extra happy people and every existing person was a million times better?
I think whether or not a given person-affecting view has to give that up can depend on the view and/or the details of the hypothetical.
1. ^
  At a basic level better, not necessarily the things they care about by derivation from other things they care about, because they can be mistaken in their derivations.
2. ^
  Moral realism, that there’s good or bad independently of individuals’ stances (or evaluative attitudes, as in my first post) seems to me to be a non-starter. I’ve never seen anything close to a good argument for moral realism, maybe other than epistemic humility and wagers.

Michael St Jules 🔸 23 Dec 2025 18:08 UTC
14 points
2 ∶ 0
on: It’s Good To Create Happy People: A Comprehensive Case
The versions of person-affecting views that are to me best motivated, most intuitive and most plausible let go of axiology as normally conceived. They don’t have a single privileged “objective” impartial order over possible outcomes, and they violate the independence of irrelevant alternatives, but there are straightforward explanations for how those alternatives come to not actually be irrelevant. See:
1. My sequence Radical empathy, or, at least the first, third and fourth posts. The third’s motivation comes from the first. The third and fourth specifically defend the Procreation Asymmetry.
2. Rabinowicz & Österberg, 1996 and van Weelden, 2019, from which the first piece in my sequence draws.
3. Gloor, 2022.

Michael St Jules 🔸 18 Dec 2025 18:15 UTC
4 points
0 ∶ 0
in reply to: saulius’s comment on: Reducing aquatic noise as a wild animal welfare intervention
It does look like most studies suggested small or no effects after less than 10 meters away, but I wonder how much they focused on eggs, larvae and zooplankton, which are plausibly more sensitive. For example, from this study (discussion):
Experimental air gun signal exposure decreased zooplankton abundance when compared with controls, as measured by sonar (~3–4 dB drop within 15–30 min) and net tows (median 64% decrease within 1 h), and caused a two- to threefold increase in dead adult and larval zooplankton. Impacts were observed out to the maximum 1.2 km range sampled, which was more than two orders of magnitude greater than the previously assumed impact range of 10 m. Although no adult krill were present, all larval krill were killed after air gun passage.
This might be an outlier study, though. I had Perplexity attempt a systematic review here.

Michael St Jules 🔸 17 Dec 2025 21:39 UTC
4 points
0 ∶ 0
in reply to: Jim Buhler’s comment on: Reducing aquatic noise as a wild animal welfare intervention
I’d guess aquatic noise reduces populations across trophic levels. There’s some evidence across different animal size groups. It also seems a priori more likely that smaller animals, like zooplankton, will have the largest population-relative direct effects at a given noise volume because the force will be larger relative to their body and organ sizes (but they may be closer or farther from noise sources), and since they feed other animals up the food chain, higher trophic levels would have less available food, too.
Speculating pretty wildly: maybe larger primarily herbivorous species would be least affected or could increase in populations due to reduced competition for food with smaller herbivores. That could actually mean increasing the average welfare of aquatic animals.

Michael St Jules 🔸 17 Dec 2025 20:20 UTC
12 points
0 ∶ 0
on: Reducing aquatic noise as a wild animal welfare intervention
Figured I’d flag that it seems pretty likely to me that aquatic noise reduces populations (and unlikely that it increases them), both fish and invertebrates, by increasing mortality and reducing fertility. See this thread with Perplexity AI (I haven’t carefully verified the accuracy, though). So, reducing aquatic noise increases wild animal populations in expectation, which is plausibly bad if their lives are bad on average. There could be tradeoffs with average welfare, but I’d be at best clueless about whether reducing aquatic noise is good or bad overall for aquatic animals in the near term.

Michael St Jules 🔸 7 Dec 2025 18:47 UTC
8 points
0 ∶ 0
in reply to: Jim Buhler’s comment on: 4. Why existing approaches to cause prioritization are not robust to unawareness
I suspect we need to involve our criteria for defining and picking bracketings here.
In practice, I think it doesn’t make sense to just bracket in the bad long-term effects or just bracket in the good ones. You might be able to carve out bracketings that include only bad (or only good) long-term effects and effects outweighed by them, but not all bad (or all good) long-term effects. But that will depend on the particulars.
I think if we only do spatiotemporal bracketing, it tells us to ignore the far future and causally inaccessible spacetime locations, because each such location is made neither determinately better off in expectation nor determinately worse off in expectation. I’m not entirely sure where the time cutoff should start in practice, but it would be related to AGI’s arrival. That could make us neartermist.
But we may also want to bracket out possibilities, not just ST locations. Maybe we can bracket out AGI by date X, for various X (or the min probability of it across choices, in case we affect its probability), and focus on non-AGI outcomes we may be more clueful about. If we bracket out the right set of possibilities, maybe some longtermist interventions will look best.

Michael St Jules 🔸 26 Nov 2025 23:23 UTC
9 points
0 ∶ 0
in reply to: Vasco Grilo🔸’s comment on: Caring about Bugs Isn’t Weird
My sense is that if you’re weighing nematodes, you should also consider things like conscious subsystems or experience sizes that could tell you larger-brained animals have thousands or millions of times more valenced experiences or more valence at a time per individual organism. For example, if a nematode realizes some valence-generating function (or indicator) once with its ~302 neurons, how many times could a chicken brain, with ~200 million neurons, separately realize a similar function? What about a cow brain, with 3 billion neurons?
Taking expected values over those hypotheses and different possible scaling law hypotheses tends, on credences I find plausible, to lead to expected moral weights scaling roughly proportionally with the number of neurons (see the illustration in the conscious subsystems post). But nematodes (and other wild invertebrates) could still matter a lot even on proportional weighing, e.g. as you found here.

Michael St Jules 🔸 24 Nov 2025 17:39 UTC
8 points
1 ∶ 0
in reply to: NickLaing’s comment on: Caring about Bugs Isn’t Weird
Ya, bracketing on its own wouldn’t tell you to ignore a potential group of moral patients just because its probability of sentience is very small. The numbers could compensate. It’s more that conditional on sentience, we’d have to be clueless about whether they’re made better or worse off. And we may often be in this position in practice.
I think you could still want some kind of difference-making view or bounded utility function used with bracketing, so that you can discount extreme overall downsides more than proportionally to their probability, along with extreme upsides. Or do something like Nicolausian discounting, i.e. ignoring small probabilities.

Michael St Jules 🔸 24 Nov 2025 8:06 UTC
20 points
4 ∶ 0
on: Caring about Bugs Isn’t Weird
Great post! :)
Third, it’s easy for an org to think it’s helping when it’s hurting: we know so little about how to help that some caution is warranted. I don’t just want to do good in expectation: I want to do good.
FWIW, I think this would count against most animal interventions targeting vertebrates (welfare reforms, reductions in production), and possibly lead to paralysis pretty generally, and not just for animal advocates.
If we give extra weight to net harm over net benefits compared to inaction, as in typical difference-making views, I think most animal interventions targeting vertebrates will look worse than doing nothing, considering only the effects on Earth or in the next 20 years, say. This is because:
1. there are possibly far larger effects on wild invertebrates (even just wild insects and shrimp, but also of course also mites, springtails, nematodes and copepods) through land use change and effects on fishing, and huge net harm is possible through harming them, and
2. there’s usually at least around as much reason to expect large net harm to wild animals as there is to expect large net benefit to them, and difference-making gives more weight to the former, so it will dominate.
There could be similar stories for the far future and acausally, replacing wild animals on Earth with far future moral patients and aliens. There are also possibilities and effects of which we’re totally unaware.
That being said, I suspect typical accounts of difference-making lead to paralysis pretty generally for similar reasons. This isn’t just a problem for animal interventions. I discussed this and proposed some alternative accounts here.
Bracketing can also sometimes help. It’s an attempt to formalize the idea that when we’re clueless about whether some group of moral patients is made better off or worse off, we can just ignore them and focus on those we are clueful about.

Michael St Jules 🔸 22 Nov 2025 0:29 UTC
2 points
0 ∶ 0
on: Help the most neglected farmed mammals —replacing feeder rodents with alternatives that win on cost and convenience
Sounds exciting!
Do you have an estimate of the cost of the product per rodent spared you can share? This could help set a lower bound on the potential cost-effectiveness, where in the roughly worst case, donors, grantmakers or impact investors buy or subsidize the product for snake owners, similar to SWP buying stunners for shrimp producers.

Michael St Jules 🔸 16 Nov 2025 18:43 UTC
7 points
0 ∶ 0
on: What is the current best estimate of the cumulative elasticity of chicken?
(Edited: had the comparison of sizes flipped.)
One thought I’ve just had about this: these “cumulative elasticities” assume demand/price shifts just for one product at a time, and will therefore be too (EDIT) low if used for the effects of people going veg or reducing their consumption of multiple animal products. Here’s why:
1. Suppose one person (or a large group of them) just cuts out chicken, and for simplicity, they don’t eat more of anything else to compensate. That reduces the price of chicken. Other people will eat more chicken and less of everything else (including other animal products) to compensate, all else (or at least elasticities) equal. All else equal, it will also disproportionately affect chicken production.
2. Now, suppose this same person (or same large group of people) go vegan. This reduces the price of all animal products. Instead of others consuming more chicken and less of everything else in response, they consume more animal products all around (all else equal) and less plant-based food. The effect of other consumers on chicken in particular will be smaller than in 1, because they aren’t shifting away from other animal products to eat more chicken in particular. So more of the demand shift goes through.
It can also matter whether the elasticities were determined in a model with multiple products (general equilibrium), including with cross-price elasticities of demand, or just a single-product market. In models with multiple products, when you estimate the price elasticity of chicken demand, you’re adjusting for other prices and quantities besides chicken’s. In a model with just chicken prices and quantities, you aren’t adjusting for other products’ prices and quantities. These elasticities will differ.

Michael St Jules 🔸 20 Oct 2025 21:15 UTC
5 points
0 ∶ 0
on: Some thoughts on fanaticism
Great post, thanks for writing!
I buy that individuals should try to pick “policies” and psychologically commit themselves to them, rather than only evaluate actions one at a time. I think this totally makes sense for seatbelts and helmets. However, I’m not sure it requires evaluating actions collectively at a fundamental normative level rather than practically, especially across individuals. I think we can defend wearing seatbelts and helmets with Nicolausian discounting without supporting longtermism or x-risk work to most individuals, even if the marginal x-risk opportunity were similar to the average or best already funded.
In particular, I know that if I don’t wear my seatbelt this time in a car by some logic that is not very circumstance-specific, I could use similar logic in the future to keep talking myself out of wearing a seatbelt, and those risks would accumulate into a larger risk that could be above the discount threshold. So I should stop myself now to minimize that risk. I should consider the effects of my reasoning and decision now on my own future decisions.
However, I don’t have nearly as much potential influence over humanity’s x-risk strategy (causally or acausally) and the probability of an existential catastrophe. The typical individual has hardly any potential influence.
Also, separately, how would you decide who or what is included in the collective? Should we include the very agents creating the problems for us?

Michael St Jules 🔸 9 Oct 2025 12:35 UTC
2 points
0 ∶ 0
in reply to: JesseClifton’s comment on: [Linkpost] Bracketing Cluelessness
FWIW, unless you have reason otherwise (you may very well think some Fs are more likely than others), there’s some symmetry here between any function F and the function 1-F, and I think if you apply it, you could say P(F > ¹⁄₂) = P(1-F < ¹⁄₂) = P(F < ¹⁄₂), so P(F < ¹⁄₂) ≤ ¹⁄₂, and strictly less iff P(F = ¹⁄₂) > 0.
If you can rule out P(F = ¹⁄₂) > 0 (say by an additional assumption), or the bet were on F ≤ ¹⁄₂ instead of F < ¹⁄₂, then the probability would just be ¹⁄₂.

Michael St Jules 🔸 26 Sep 2025 19:22 UTC
9 points
1 ∶ 0
on: Giving multiplier of Giving What We Can
I recommend funding GWWC and the Centre for Exploratory Altruism Research’s (CEARCH’s) High Impact Philanthropy Fund (HIPF) due their effects on soil animals, which I think are practically proportional to the increase in agricultural-land-years per $. I estimate HIPF increases agricultural land 9.42 times as cost-effectively as GiveWell’s top charities, which is similar to my estimates for the giving multiplier of GWWC in 2023 and 2024.
A few quick comments:
1. GWWC donors don’t only give to GiveWell (or other life-saving) charities, so you’d want to discount by the proportion going to them.
2. There might be other charities that GWWC supporters donate to that will increase soil animal populations in expectation, e.g. some diet change and alternative protein work. You’d probably want to check the net effect combining them.

Michael St Jules 🔸 25 Sep 2025 19:34 UTC
7 points
0 ∶ 0
in reply to: Guillaume Corlouer’s comment on: Bracketing cluelessness: A new theory of altruistic decision-making
(Edited to elaborate.)
I think bracketing agents could be moved to bracket out and ignore value of information sometimes and more often than EV-maxers, but it’s worth breaking things down further to see when. Imagine we’re considering an intervention with:
1. Direct effects on a group of moral patients (or locations of value), and we’re clueless about those effects.
2. Some (expected) value of information for another group of moral patients (possibly the same group, a disjoint group or intersecting the group in 1).
Then:
a. If the group in 2 is disjoint from the group in 1, then we can bracket out those affected in 1 and decide just on the basis of the expected value of information in 2 (and opportunity costs).
b. If the group in 2 is a subset of the group in 1, then the minimum expected value of information needs to be high enough to overcome the potential expected worst case downsides from the direct effects on the group in 1, for the intervention to beat doing nothing. The VOI can get bracketed away and ignored along with the direct effects in 1.
And there are intermediate cases, with probably intermediate recommendations.