Comments
Welfare Biology and AI: What We Can Do Now
From soil-welfare research to the New World Screwworm: a practical portfolio for wild-invertebrate welfare.
This is part 3 of a five-part sequence on welfare ecology. Part 1 introduces the ethical premises. Part 2 covers the empirical landscape. Part 4 explores a model of invertebrate suffering. Part 5 covers AI.
In part 2, I argued that land use is a key lever for wild-invertebrate welfare and that pesticides applied at constant net primary productivity (NPP) can create a high-throughput killing field rather than a real population reduction. This post turns to the practical question: given that picture, what should we do?
The answer depends on which premises from part 1 you hold. I’ll write from my own view – asymmetric antifrustrationism with non-negligible probability mass on invertebrate sentience – but I’ll flag where the conclusions change for symmetric utilitarians or for people who don’t extend moral concern below the arthropod line.
A month ago, Rethink Priorities (RP) published its Wild Animal Welfare Intervention Database (WAWID), twenty-eight shallow reports on candidate interventions. It changes the picture. Several of the most promising entries are birth-prevention interventions that map cleanly onto the framework from parts 1 and 2, and a couple are surprises – the New World Screwworm (NWS) suppression case in particular is, on my read, the single strongest entry in the database.
Part 2 argued that the m²-years of agricultural land per food-kg explain almost all of the variance in welfare effects across foods. Converting natural biomes to agriculture reduces soil fauna density by 2–8×, and the sheer number of organisms affected (billions per m²-year) overwhelms everything else.
That mechanism is what makes land use change – not pesticide use, not “humane mowing,” not habitat enrichment – the central lever. Most of the interventions discussed below either work directly through it (existing donations, dietary choices, biofuels as a side effect) or are picked specifically because they do something land use cannot, like target a fly that causes excruciating pain to mammals it parasitizes.
Before we go through the candidates, two filters from the asymmetric antifrustrationist position do most of the prioritizing work. They aren’t decorative – they sharply rerank what looks promising.
Pro-choice. Or birth-prevention beats death-improvement: Under the asymmetry from part 1, creating frustrated lives is bad; preventing them isn’t. Improving an existing being’s mode of death matters, but the welfare delta is bounded by the difference between two deaths – a few hours of pyrethroid neurotoxicity versus a few hours of fungal infection, say. This has advantages for collaborations across ethical camps, though. Preventing that being from existing in the first place is bounded by their entire potential life. For r-strategists with high fecundity and short life expectancy at birth (LEB), almost the entire LEB is pre-reproductive death anyway, so the “future-life” side of the trade is large.
Soft Pareto. Or creating new frustrated lives in service of preventing others is bad: Bounded marginal costs to existing beings – slightly more expensive renewable fuel, a small Pigouvian tax, soil disturbance during a one-time land-conversion event whose victims would have died of other causes within weeks – are the kinds of trades society routinely makes. But bringing new sentient beings into existence as the means by which welfare gains for others are achieved is a different kind of action. The asymmetry weighs the entire frustrated life that gets created against the prevented frustrations elsewhere; there’s no bounded “marginal” framing that contains the harm. This is why eating beef to displace soy production registers differently from biofuel subsidies that reduce soil-fauna populations as a policy side effect: the first creates cattle in order to prevent invertebrate births; the second imposes bounded human costs (slightly higher food prices) to produce the same kind of welfare benefit.
There are two important caveats to this:
Taken together, these considerations confuse me. The first is clear enough, but once the second enters the picture I reliably lose my ethical footing. For the time being, I think of it along the lines of contractualism. If I can reasonably expect that someone expects me to make a certain choice, then that choice is my default to compare others against. In cases where there is no such expectation, I at least don’t judge people for the counterfactuals they choose.
As an example, humans in need sort of vaguely expect to perhaps be helped. Humans in general sort of vaguely expect not to be killed. But serial donors and serial killers are a thing. When a serial donor decides to donate somewhere, the decide not to donate elsewhere; likewise, when a serial killer decides to spare someone they decide to kill someone else. Naturally, in the age of financial self-determination, mood stabilizers, antipsychotics, and Dialectical Behavioral Therapy, we’re not locked into our nature, be it that of the serial donor or the serial killer – any serial killer could become a serial donor if they put their mind to it – but bracketing for the moment these changes to the whole setup, they face very similar decision: who gets to live and who gets to die – only serial donors get to make it at a vastly greater scale. This consideration has made me a lot less judgmental of serial killers.
In summary, the two filters – pro-choice and soft Pareto – favor: birth-prevention interventions, broad-spectrum population-reducing levers, direction-agnostic research, and bounded-marginal-cost trades. They disfavor: interventions that only mildly improve existing deaths without reducing future births, conservation-style interventions that increase aggregate populations, and any structure where new sentient beings are brought into existence to improve the welfare arithmetic.
A third filter follows from part 2’s empirical picture: prefer interventions that affect both nematodes and arthropods, because the dominant uncertainty in any cost-effectiveness analysis here is the welfare range of nematodes – the most numerous probably-sentient animals on Earth, and the ones we know least about. An intervention that is good whether nematodes turn out to have substantial or negligible welfare ranges is more robust than one whose whole raison d’être rests on that question.
Research stikes me as the most cost-effective category, and that’s why it goes first. The reason is structural: the entire portfolio below depends on comparisons of welfare ranges and on how fine-grained land-use changes propagate through trophic networks. A few hundred thousand dollars of well-targeted welfare-biology research could redirect tens or hundreds of millions of dollars of downstream allocation.
Three concrete research directions look strongest.
Soil-animal welfare research. This is the question the rest of the field is logically downstream of. We need functional characterization of nociception in nematodes – the connectome of Caenorhabditis elegans is mapped, but we don’t know whether its nociceptive processing produces phenomenal pain – along with welfare-per-animal-year estimates for soil mites and springtails, and on-the-ground measurements of how specific land-use changes propagate to soil-fauna densities and whether they are stable or wash out over time (short of millions of years) or get offset in some way. The current cost-effectiveness numbers, including the welfare range of 6.68 × 10⁻⁶ for nematodes that Grilo deferred to Gemini 2.5 to estimate, are placeholders waiting for empirical replacement. The Wild Animal Initiative (WAI) and Rethink Priorities are the obvious organizations to defer to.
Insecticide welfare research. The WAWID’s research-agenda entry by Simon Eckerström Liedholm proposes comparing insecticide classes on welfare effects on wild insects, not just mortality. Roughly 3.5 × 10¹⁵ insects are exposed to insecticides on US farmland each year. We have no idea which classes minimize per-individual suffering – pyrethroids cause hours of paralytic seizure, neonicotinoids 1–2 days of disorientation and twitching, Bacillus thuringiensis (Bt) leads to multi-day death by sepsis, insect growth regulators kill at the next molt by failure to ecdyse. Picking the least-bad class is direction-agnostic: the recommendation holds whether wild insect lives are net negative (then we want euthanasia, but as gentle as possible) or net positive (then we ought not to interfere, but failing that have a duty to minimize the suffering we impose).
Field-building. Mal Graham’s WAWID shallow on scientific field-building for wild animal welfare frames a different version of the research question: what does it cost to bring the field to maturity, with degree programs, peer-reviewed journals, and government and foundation funding? Her model assumes EA’s funding share decays from ~95% today to ~5% over thirty years as traditional science funders pick up the field, with stable annual budgets reaching ~$100M. The headline cost-effectiveness comes out to ~13 animal-years per dollar – within the same range as the 9–120 chicken-years per dollar Saulius Simcikas estimated for corporate cage-free campaigns. The argument I find persuasive in Graham’s piece isn’t the precise number; it’s the leverage. EA dollars spent on welfare-biology research now don’t just buy research; they pull in non-EA dollars later, the same way EA’s farmed-animal welfare investments eventually pulled in major-foundation and government attention.
The second tier is large-scale, high-confidence, well-understood existing donation targets. Their welfare effect on soil fauna might be enormously larger than their direct effect on the intended beneficiaries (depending mostly on nematode welfare ranges), but it runs through the same well-studied land-use lever, so we can be unusually confident in the direction of the effect.
Update: It’s important to note that Grilo has found a study that found surprisingly similar numbers of nematodes in the soil of different habitats. If true, this might ruin the robustness of this lever.
Funding GiveWell’s top charities. Grilo (2025a) estimates that GiveWell’s top charities increase cropland by ~137 m²-year per dollar, translating to 1.11 kQALY per dollar when accounting for soil animals – 1.74× the past cost-effectiveness of the Shrimp Welfare Project’s Humane Slaughter Initiative (HSI). The mechanism: saving lives increases the human population, which increases food demand, which increases cropland. Grilo estimates a 0.0157 human-year-per-dollar increase in human living time. But note that Roodman (2014) puts the sign of the mortality-fertility effect in some doubt.
Funding CEARCH’s High Impact Philanthropy Fund (HIPF). The Centre for Exploratory Altruism Research (CEARCH) estimates HIPF is ~12× as cost-effective as GiveWell’s top charities for human welfare. Grilo translates this into 13.4 kQALY per dollar with soil animals included – 20.9× HSI’s past cost-effectiveness. The same caveat as above applies, with the additional uncertainty Grilo flagged in his June 2025 update about whether HIPF’s chronic-disease-policy grants actually increase food consumption.
Chicken welfare campaigns. Broiler welfare and cage-free campaigns increase agricultural land by 93.9 and 16.4 m²-year per dollar respectively, because reformed systems require more feed per unit output. Grilo estimates these benefit soil animals 458× and 29× as much as they benefit chickens, though uncertainties over welfare ranges apply because most of them are nematodes. These chicken welfare reforms are likely good on their own terms; the soil-animal addition is a robust positive if soil animals have net negative lives and verly likely desirable from an asymmetric perspective.
Biofuel subsidies as a natural experiment. The WAWID shallow report on the US Renewable Fuel Standard (RFS) gives us a real-world confirmation of part 2’s mechanism: a policy-driven 2.1 Mha conversion of grassland to cropland produced an estimated reduction of ~12.9 trillion macroinvertebrates per year. The report’s own assessment is that movement counterfactual influence on the RFS is essentially zero – corn-state political coalitions and infrastructure lock-in mean nothing welfare-biology advocacy does will move that policy. So it isn’t an advocacy target. But it’s the cleanest empirical demonstration that the part-2 mechanism produces what part 2 says it produces, and worth knowing about for two reasons. First, EA-adjacent environmentalists who oppose the RFS on conservation grounds are working against welfare on this analysis; that’s an alignment worth noticing. Second, for asymmetric antifrustrationists worried about instrumental harm, the RFS structurally passes the second filter above: its costs to existing beings are bounded (food prices, displaced soy expansion), and no new sentient beings are created in service of the antifrustrationist gain. As such it can serve as a target for opportunistic altruism, in case a policy window does open.
The most interesting addition from the WAWID is a small group of interventions that combine high per-individual suffering severity, plausible birth-prevention mechanisms, and clean structural-asymmetry profiles. None has the broad-spectrum coverage of the donations above; none affects both nematodes and arthropods. But they are the cleanest single-shot applications of the framework to specific species whose lives we have unusually strong reasons to think are net negative.
New World Screwworm continental suppression. The WAWID shallow report on suppressing Cochliomyia hominivorax across the Americas is, on my read, the single strongest entry in the database. C. hominivorax larvae cause myiasis by feeding on the living tissue of warm-blooded mammals – the Welfare Footprint Institute Pain-Track gives 10–30% excruciating pain and 70–80% disabling pain over the 14 days from oviposition to death by sepsis or toxemia. Mortality is near 100% in untreated wild infestations, and 1–4% of warm-blooded wild animals in endemic regions are infested at any given time.
The 1950s–2000s campaign eradicated NWS from North and Central America using sterile insect technique (SIT). The Panama biological barrier collapsed in 2024, the fly is spreading north again, and the United States is now scaling sterile-fly production toward 300 million releases per week to push it back to Panama. Continental suppression – proactively eliminating NWS across South America rather than perpetually defending a barrier – is feasible with new male-only CRISPR strains and gene drives, and the political momentum from the 2024 outbreak creates a 5–10 year window. Screwworm Free Future is the dedicated organization.
Why this is exceptional on the asymmetric antifrustrationist framework: the suppression intervention is doubly birth-preventing. It prevents future cohorts of NWS flies in the wild (the obvious channel), and it ends the perpetual sterile-fly rearing program – currently ~100M flies/week, scheduled to scale to 300M/week, and rising indefinitely under climate change. Almost no other intervention reduces both wild-population births and movement-driven captive-population creation. The Pareto profile is also unusually clean: warm-blooded wild animals win, livestock win, humans win, and the welfare cost to NWS flies themselves is bounded by RP’s welfare-range estimates – ~0.013 (using Black Soldier Fly as a proxy) versus mammalian ~0.5.
Gene drive for Anopheles gambiae. The WAWID shallow report on gene drives for malaria explores two architectures: a suppression drive that pushes the A. gambiae population toward zero, and a replacement drive that leaves population sizes intact but renders the mosquitoes unable to transmit Plasmodium. The shallow report leans toward replacement on extinction-ethics grounds. From my framework, suppression dominates: trillions of A. gambiae per year will continue to die of natural causes for as long as the species exists in malaria-endemic Africa. A successful suppression drive prevents the births of all those future cohorts, and the human side of the trade – ~600,000 fewer malaria deaths per year, mostly children – is enormous and aligned. The lever for a welfare-biology movement isn’t building the drive; it’s shaping the funding and regulatory criteria so welfare-considerate designs are favored.
Mating disruption for rice yellow stem borer and spongy moth Stop the Spread. Two structurally similar Lepidopteran interventions belong together. The WAWID shallow report on Scirpophaga incertulas describes pheromone-based mating disruption as a non-lethal alternative to insecticides in South and Southeast Asian rice paddies. Field trials in Indonesia produced 40–56% fewer insecticide applications and yields equal to or higher than insecticide controls. The shallow report estimates 6.7 × 10⁹ to 1.4 × 10¹² yellow stem borer larvae per year on the ~73.8 Mha of insecticide-managed rice in the region, plus a much larger uncounted reduction in non-target insect deaths from displaced organophosphate, pyrethroid, neonicotinoid, and diamide applications – all of which the shallow’s pain table classes as “likely excruciating,” with deaths over hours to days. Provivi and Syngenta announced commercial deployment in India and Indonesia starting in 2026; ASSIST’s “Phero Rice” project is already running through Farmer Field Schools in Vietnam.
The RP report on spongy moth outbreaks estimates 35–385 trillion Lymantria dispar dispar caterpillars suffered through outbreak years 2 and 3 in the United States over the last 50 years. Most die of starvation, Entomophaga maimaiga fungal infection, or LdNPV virus (“larval melting”) rather than from active control. The two intervention arms that align with asymmetric antifrustrationism are technologies that kill spongy moth eggs before hatch and mating disruption at the invasion front; both are pure birth-prevention. Faster Bacillus thuringiensis var. kurstaki (Btk) alternatives are death-improvement, which matters but ranks lower on the framework.
A note on vertebrate analogs. The WAWID shallows on a combined contraception/TB vaccine for New Zealand brushtail possums and on fertility control for white-tailed deer show the same logical structure on a smaller scale: birth-prevention via reproductive disruption, replacing a more painful counterfactual (1080 poisoning, hunting/culling) with a less painful one. Per-individual welfare ranges are higher for vertebrates, but population scales are bounded (~30M possums total, ~5M killed/year for 1080). Worth knowing about; not the largest priorities.
Eat cheaply, donate the savings. Two arguments converge on the same recommendation, but they are independent and worth stating separately.
The cost-effectiveness argument is Grilo’s (2025b) own: a whole-food plant-based diet is cheaper than a standard one, so it allows him to donate more, and the donation impact dominates the dietary effect. On the per-kg numbers he computed:
Beef looks like an outlier – ~26× more soil-animal-years displaced than chicken, ~395× soy milk. On the cost-effectiveness story alone, this is what Grilo means when he calls the dietary choice “ambiguous”: a beef-heavy diet displaces enormous amounts of soil-fauna life.
The soft Pareto argument is the other reason to eat plant-based, and it doesn’t depend on the sign of soil-fauna welfare. Eating beef creates cattle whose entire frustrated lives enter the welfare ledger at full weight, in order to displace soil-fauna lives via the cattle-feed → soy-expansion pathway. That’s exactly the cow → soy structure the second filter above, the soft Pareto filter, discourages: bringing new sentient beings into existence as the means of producing antifrustrationist gains. Plant-based calorie shifting between plant foods is a redirection of existing land use without creating new sentient beings to instrumentalize the gain. Eat-cheap-donate-the-savings works through the second mechanism only.
The two arguments converge on plant-based, but only the soft Pareto one distinguishes “create cow lives to prevent insect lives” from “redirect existing land use to prevent insect lives.” This is a place where the cost-effectiveness framing alone would lead you to cause greater instrumental harm – with the caveats mentioned above.
A recurring theme: interventions that reduce both nematodes and arthropods simultaneously are more robust than ones that target one group, for three reasons.
Robustness across welfare ranges. The dominant uncertainty in Grilo’s model is the welfare range of nematodes. Rethink Priorities has estimated welfare ranges for some arthropod species, but nematode ranges have not been estimated. C. elegans has 302 neurons; soil mites and springtails have thousands; ants hundreds of thousands. Neurons are a crude proxy for welfare range, but it stands to reason that a nematode’s range may be much smaller than an arthropod’s. Conversely, though, nematodes are much more numerous. An intervention that affects both in the same direction is positive across this entire space.
Robustness against trophic backfire. Many soil-mite species are voracious nematode predators – some consume hundreds per day. If you selectively suppress mites, nematode populations can increase through predation release. Springtails compete with nematodes for fungal substrates. Broad-spectrum land-use change avoids these feedbacks because everything goes down together.
Robustness through existing policy levers. Land use, agricultural subsidy, and climate policy have massive existing political infrastructure. Building novel taxon-specific tools (nematode-only biocides, say) means starting from zero. Working through existing levers is enormously cheaper.
The targeted birth-prevention interventions are narrower on this axis than the broad-spectrum donations: they’re picked for severity of per-individual suffering, not breadth of taxon coverage. The portfolio reasoning says we want both kinds of bets.
A handful of interventions look reasonable from a generic conservation or animal-welfare framing but turn out badly under asymmetric antifrustrationism.
Stabilizing oscillating biomes. Boom-bust ecosystems with predator-prey cycles, insect outbreaks, or fire-driven succession involve mass starvation and die-offs that look like prime suffering targets. The catch is that the available stabilization interventions tend to increase aggregate biomass and biodiversity – the stable state is typically more productive than the average of the cycle. From a conservation perspective that’s success; from an antifrustrationist perspective, more stable, more productive ecosystems support more potentially-suffering organisms, not fewer. Fire suppression is another example: preventing fire not only allows succession, which raises soil-fauna density and thereby aggregate suffering, it also stockpiles fuel for worse eventual fires. Either way, this is a bucket to exclude from the portfolio, not a bucket to fund.
Mandating wind-turbine curtailment for insects. The WAWID shallow report on climate-energy infrastructure estimates ~23,200 trillion insects per year killed by wind turbines in instantaneous, low-pain collision deaths. Population-dynamics arguments suggest most of this mortality is density-compensated – the population would have been the same without turbines. From the asymmetric antifrustrationist view this is one of the least problematic mortality sources on Earth: a near-instant death substituting for what would otherwise be parasitism, fungal infection, or starvation. The intervention proposed in the shallow report – mandating curtailment of turbines during peak insect activity – would effectively impose more painful counterfactual deaths to save individuals whose populations would not be reduced. Photovoltaic panel mortality is a different story: polarotactic insects are attracted to the panels for oviposition, and the eggs and larvae die on the surface over hours. It’s unclear how sentient they are at that stage, but the process doesn’t change the NPP of the area, so the population size will barely decrease. The white-grid panel design tweak that breaks the polarized-light trap is probably good. So: panel design good-ish, turbine curtailment bad.
Banning neonicotinoid seed treatments. The WAWID shallow report on reducing neonicotinoid seed treatments (NSTs) is unusually direct: the author concludes that the net welfare effect of NST bans “may be negative.” The reasoning lines up neatly with part 2’s pesticide framework. NSTs are systemic and persistent; they make plant NPP physiologically inaccessible to phytophagous insects, which is genuine carrying-capacity reduction rather than constant-NPP killing. Counterfactual deaths for the insects that would otherwise have eaten treated plants are dominated by parasitoids, which the shallow report describes as “possibly more severe and longer than death by acute NST exposure.” The shallow report even cites evidence that sublethal NSTs function as fertility control. NST bans typically replace NSTs with foliar pyrethroid or diamide sprays, both of which are worse on each of these dimensions. Yet conservation NGOs and several US states are converging on NST bans as a “win for invertebrates.” On this analysis, that convergence is wrong.
There’s a pattern across the three. Conservation-default and welfare-default intuitions diverge cleanly from asymmetric antifrustrationist conclusions, and the divergence isn’t accidental: conservation values aggregate populations and biodiversity, asymmetric antifrustrationism values reduced expected suffering, and these come apart whenever you can choose between “more healthier organisms” and “fewer organisms full stop.”
It doesn’t have to be that way. No one is just a “conservationist” and most people have empathy, especially people who go into ecology. It’s just a matter of sensitizing them to the fact that conservation and welfare are not or rarely correlated.
There is a framing around autonomy that seems internally consistent to me: Imagine a parkour athlete has a bad fall next to you. They’re still breathing but unconscious and clearly badly injured. Some of my friends would argue that they should not call an ambulance or interfere in any other way because that would impinge on the athlete’s autonomy, and being unconscious, they cannot consent. But I think this is a niche view, and to apply it to some species (nonhumans) but not others (humans) would be speciesist.
Ranked by the joint score on importance, neglectedness, tractability, asymmetry-fit, and robustness to nematode-welfare-range uncertainty:
In part 4, I’ll take a detour into the philosophy of mind to ask what invertebrate suffering actually feels like – an answer that draws on the phenomenology of psychopathy and Buddhist insight practice, and that may reshape how we weight nematode versus arthropod welfare.
Reading your article, I got the impression that you assume soil animals lead lives not worth living, and build your recommendations on that premise. To my knowledge, there is no scientific evidence for such a bold claim and your post doesn’t provide any either. Could you elaborate on this? As written, I found the article quite unsettling.
Hiii! Did you read the article in isolation, or did you start with the first one in the sequence? It's written such that each builds on the previous ones.
I’ve read both posts and I found the first part a bit one-sided. In your definition of welfare biology, for instance, you only cited authors who believe soil animals have net-negative lives, while omitting those who argue the opposite.
While the quiz in that first post gave those opposing ideas some moral weight, this second part seems to start with the assumption that soil animals live net-negative lives. I noticed the robustness section doesn't really consider the possibility that this premise could be wrong.
How would your recommendations change if we assumed net-positive lives instead? Would we then have to oppose GiveWell’s top charities, biofuel mandates, or wind turbines by that same logic?
Ultimately, do you think an assumption that can neither be proven nor unproven is a good basis for recommending where people should donate?
I'm very much writing this sequence for antifrustrationists. I didn't set out to write a sequence that collects all the major moral views out there and then pivots the recommendations for each, but I noticed that while Brian started with a suffering-focused view, later researchers didn't continue that line of thinking while making progress on other fronts, and I wanted to plug that hole. The alternative would've been a much bigger undertaking. I wouldn't want to privilege just (say) antifrustrationism and hedonic act utilitarianism, but then my goal would've been to also pivot the recommendations for dominant systems like deontology (religious or Rawlsian), contractualism, and various more. (I do actually have an April Fools joke here that goes in that direction.)
Cf., when AMF provides LLINs to people in some regions in the southern DRC, then they're doing exactly that, and not, e.g., providing nets to people in Tanzania.
With the first article I hoped to make clear that this is a sequence for antifrustrationists to figure out where they stand on nematode and arthropod welfare (and that perhaps others can also take something away from it, more coincidentally).
This seems to be ad odds with what you wrote in part 1:
I must admit that I am still finding it difficult to fully grasp your position. Specifically, regarding the GiveWell top charities, you state:
If I interpret this correctly, it suggests that your recommendations depend, at least to some degree, on the unproven assumption that soil animals live net-negative lives. Would your recommendations change if their welfare signs were found to be positive, or is your view that preventing harm always takes precedence over any amount of positive experience?
Furthermore, would you apply this same logic to humans? For instance, if an advanced Super Intelligence created a weapon that could instantaneously end all life on Earth, would you recommend its use to ensure that no preference could ever be frustrated again?
What I'm doing is very much transfer my intuitions from humans, where I think they are well tested, to other species. I don't want special pleading for humans. The question you propose seems crazy high stakes. I'm glad I'll never face a decision like that. I'd probably start a think tank or something to figure it out. The actual decisions we face every day are more like: should we have as many children as possible (1) through sperm donations, (2) by banning contraception, (3) by impregnating women and vanishing, (4) by getting as rich as possible and then have so many children that each have a life barely worth living, etc. My answer is always no. Not everyone shares those moral intuitons, which is fine for me as antirealist, but I transfer those intuitions to other species, and so what's going on with r strategist species is horrifying to me!
While I share your moral intuitions, it seems like a bit of a stretch to transfer them directly to other species. Many of your intuitions involve reproductive contexts, such as sperm donation, banning contraception, or financial resources that simply aren't available to them.
Furthermore, r-strategists follow a vastly different evolutionary path. For a human woman, being impregnated and then abandoned clearly frustrates a preference for a partner’s support; however, if a female of another species lacks that preference to begin with, the situation seems less problematic. There is also the matter of dependency: a human mother must care for a vulnerable infant, whereas r-strategists are typically self-sufficient from the moment they hatch.
I’d like to offer a different thought experiment: Suppose an alien species observes all the suffering humans on Earth and concludes they could help us by significantly reducing NPP to ensure less humans, and thus less suffering, exists in the future. Most of us would view this as a catastrophic mistake on their part. Do you believe they could "help" us that way?
You adapt them to the other species. We can do the same with much better information with our friends: One friend enjoys tomato sauce; another friend doesn't enjoy it. You want to get pizza for all three of you, and it's supposed to be a surprise, so instead of being like, “Pizza has tomato sauce on it, and my friend doesn't like it, so I can't make any inference about whether they want pizza and should [get pizza for them anyway, get pizza only for myself and the other friend],” you can make adjustments like, “Pizza has tomato sauce on it, and my friend doesn't like it, so let's ask whether they can make pizza without tomato sauce.”
E.g., I give this example of eusocial vs. solitary insects. So when I wonder whether it's stressful for an insect to be caught in my flat and not finding the way out (despite some nutritious stuff I have lying around), my guess is that it's more likely stressful for the eusocial one who wants to get back to their hive than for the solitary one. When I had a fly over for a week or so in 2021, I gave her a name, enjoyed her company, and didn't worry much about her feeling trapped (I did leave a window open when I was awake), but when I had a bee over a few days ago, I was more concerned and helped the bee find their way out.
Likewise, humans can have some 0–3 children, cows some 5–10, and cats some 60–100 over their lifetimes. They all parent and alloparent. Turtles more like 1–2k, and no (allo-) parenting. So (just based on this data) I imagine that the loss of a child is almost as bad for a cow as it is for a human, that cats grieve somewhat less, and turtles not at all.
Or when it comes to pain, I mention in my last article that it probably doesn't make sense to have very strong pain signals, when an animal cannot react to them. So for nematodes it's not very useful to experience pain strongly; for a fly it may be very useful.
In your thought experiment there is a disconnect there that is crucial for my work: Antifrustrationism, as a form of preference utilitarianism, is all about the preferences of the individuals. If someone can fulfill their own preferences, there's nothing for me to do. If someone can tell me how I can help them fulfill their preferences, I'll happily do things for them that strike me as odd, like unusual kinks, because even though I don't share them, I trust their ability to know and communicate their preferences.
I only run into problems in cases where there is a power differential but the beings cannot communicate their preferences, e.g., because they are in the future, far away in the universe, or it's hard/impossible to build the sorts of experiments where one could elicit revealed preferences. Those are the cases where it gets uncomfortable because I need to make guesses and inferences.
It's like with a caring mother of an infant who can't talk yet: The infant can't help themselves (yet), so she has to do it, but to do that well, she has to infer the preferences rather than asking what they are.
With the aliens, the situation is reversed. They are powerful but really dumb and could just ask how they can really help us, and NPP is not an obvious lever for a K-strategist species. But we have this problem with aging. Maybe they'll stop by Earth 100k years ago, largely fail to communicate with us, but study our biology and notice that this aging thing really sucks, especially in the last couple of decades of an animal's life. So just leave a time capsule with information on how to reverse aging that's designed in such a way that they hope we can decypher it once we have the necessary foundational technologies.