I mostly do 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 and cluelessness, and indirect effects on wild animals.
I’ve also done economic modelling for some animal welfare issues.
Given your (partial) success in court (reported here and here), have your plans changed for what you’d use extra funding for? And has the funding gap changed? Your post says:
We will now explore bringing private prosecutions against mega-farms that use Frankenchickens.
I could imagine that a higher human population could induce demand for agriculture and increased trash output which could increase terrestrial invertebrate populations.
So this would be more food/net primary productivity available for terrestrial invertebrates to eat, and agriculture would have to increase net primary productivity overall (EDIT: or transform it into a more useful form for invertebrates), right?
My best guess is that more humans reduces wild terrestrial invertebrate populations in the near term / on Earth (so ignoring space colonization), largely through agricultural land use.[1] If you think:
these wild invertebrates or even just wild insects matter a decent amount, say about as much as RP does/would (interpolating their estimates), and
they have lives worth preventing, because you’re suffering-focused or just think they have net negative lives, say, due to their high fertility and mortality rates,
then increasing human populations could be good for animals in the near term, with the effect on wild animals outweighing those on farmed animals.
It’s unclear to me what’s going on with wild aquatic animals.
See especially Tables 3 and 4 from Attwood et al., 2008.
Net primary productivity is also typically lower in crops, across crops, based on “Land Use Change” greenhouse gas emissions from OWID / Poore & Nemecek, 2018.
Gross primary productivity decreases when replacing forest with crops, but increases if replacing grassland with crops, according to the globally representative study Krause et al., 2022.
Some other studies support increased and others decreased net productivity in crops compared to nature (Tomasik, 2013–2022, a, b)
But pesticides and fertilizers also plausibly reduce arthropod populations based on my lit reviews (but less clear in the long run with repeated use), so even if primary productivity increased, arthropod populations could still be lower overall.
I’d guess other forage fish’s life histories will give you a general idea of bristlemouths’ life histories.
About Peruvian anchoveta, one of the most wild-caught fish, representing 28% of the number of fish caught annually on average (Mood & Brooke, 2024), Molina-Valdivia et al. (2020) write:
Early larval growth and mortality of anchoveta is highly variable at intra-seasonal and latitudinal scales along the Chilean coast. In northern Chile (23 °S), larval growth varies between 0.50- and 0.85-mm day−1, with daily losses of 16–23% (Contreras et al., 2017); meanwhile, in central Chile (36 °S), larvae grew at 0.40-0.57 mm day−1, with daily losses of 4–7% (Castro and Hernández, 2000; Hernández and Castro, 2000).
See also Fig. 4. H (and G for another species) for the drop in abundance over time as larvae, leaving less than 1% after 40 days according to their fitted model.
Butler et al. (1993, Tables 1, 2, 3 and 4) report fecundity and mortality estimates for northern anchovies (Engraulis mordax) and Pacific sardines (Sardinops sagax) at various life stages. A female northern anchovy spawns 5.3 to 23.5 times per year, 4,000 to 14,000 eggs each time. I have some calculations for mortality rates at various life stages based on this paper here.
I think bristlemouths just make up a large share of all fish, by numbers of individuals. Lanternfish do, too. When someone says “a quadrillion”, this is an order of magnitude estimate, so it could easily be 3x too high or 3x too low. The estimates seem consistent to me as order of magnitude estimates, as long as bristlemouths do make up a decent share of all fish, like >30% of individuals.
The Micronekton Wikipedia page says:
Bristlemouths (Gonostomatidae), largely Cyclothone, account for more than 50% of the total vertebrate abundance between 100 and 1000 m. Twenty-one species of bristlemouths have been described globally. Lanternfishes are the secondmost abundant marine vertebrates, having diversified into 252 species.[24]
Futhermore, among pelagic fishes from 0-5000 m in the Sargasso Sea, northwestern Atlantic Ocean, “[t]he bristlemouth, Cyclothone braueri, dominated the catches both above (47%) and below (41%) 1000 m” (Sutton et al., 2010).
The number of bristlemouths has been estimated to be in the “hundreds of trillions — and perhaps quadrillions, or thousands of trillions” (Broad, 2015), so 10^14 to 10^16.
I estimated the number of lanternfish to be approximately between 10^14 and 10^16, based on a “total global biomass of 1.8 to 16 gigatonnes, accounting for up to 65% of all deep-sea fish biomass” (Lanternfish—Wikipedia, based on Hulley, 1998, pp. 127–128, ISBN 0-12-547665-5) and an average weight of “two to six grams” (Blacow, 2015).
This is all from my piece Which animals are most affected by fishing?
On whether they’re R-strategists: ecologists may have moved on from that classification, but either way, bristlemouths are fish, and almost all fish species, as far as I know, have many offspring (at least hundreds?) and high juvenile mortality rates (>90%?). Bristlemouths are forage fish, so food for predators. But they also probably eat zooplankton, like copepods.
We’re facing a funding gap of £236k this year and seek long-term funding to continue our work and give us greater financial stability.
Any updates on this since this was posted?
How much room for more funding do you expect to have in general for work in 2025 (after including funding coming from ACE)?
I’m clueless about the sign of the effects on wild aquatic animals (based on this project), and the effects on wild aquatic animals could be more important than the effects on farmed aquatic animals.
Reductions in animal agriculture, including aquaculture, have a high risk of backfiring overall, in case they’re bad for wild terrestrial arthropods. In particular, animal agriculture, including aquaculture, tends to decrease wild terrestrial arthropod populations through land use (for crops and/or pasture), and possibly substantially (Attwood et al., 2008, Tables 3 and 4), so reducing animal agriculture in general can increase wild terrestrial arthropod suffering by increasing their populations.
Aquaculture tends to have smaller effects on wild terrestrial arthropods than land vertebrate farming due to lower land use per kg of output/protein, and a targeted reduction of aquaculture — and not a broader reduction across animal agriculture in general — will, via substitution effects, shift some animal production towards land vertebrate farming and increase fishing pressure (and have highly uncertain net effects on wild aquatic animals, but actual wild capture production could go up or down overall [1][2]). This could reduce or even flip the wild terrestrial animal effects in 3.
Based on my own unpublished calculations, the targeted reduction of shrimp farming in particular seems good for farmed animals + wild terrestrial animals together (if you think wild arthropods have bad or ~neutral lives in expectation, and your moral weights for shrimp are high enough), although still highly uncertain for wild aquatic animals. However, my calculations need more review.
Not a backfire effect, but work to reduce aquaculture seems to reduce insect and brine shrimp farming in expectation, and this would, on my view, count in its favour. I don’t know if these effects or the effects on wild animals would be more important, because I haven’t really looked into it enough.
Even the best counterexample to (a theoretical version of) IIT consists in building a simple system with a high measure of “integrated information”: I entirely agree with that line of attack, that is fatal both for large monotonous matrices and tiny shrimp brains.
How does this argument apply to shrimp?
IIT is not a plausible theory of consciousness and it seems badly off track. It doesn’t aim to explain why we believe we’re conscious or really aim to identify a plausible mechanism for when things become access conscious or reportable to us. There was even an open letter signed by over a hundred relevant experts calling it pseudoscience. And you call a proposed counterexample to it fatal.
Why should we believe consciousness is superadditive in general, and to an extent that scales substantially with neuron counts in practice, if we reject IIT?
I’m guessing there isn’t much more we can gain by discussing further, and we’ll have to agree to disagree. I’ll just report my own intuitions here and some pointers, reframing things I’ve already said in this thread and elaborating.
It’s useful to separate the outcomes from the actions here. Let’s label the outcomes:
Nothing: the result of pressing neither button.
A: Bob getting an extra 1 util and Todd being created with a util, the result of only button 1 being pressed.
B: Todd being created with 3 utils, the result of both buttons being pressed.
On my person-affecting intuitions, I’d rank the outcomes as follows (using a different betterness relation for each set of outcomes, violating the independence of irrelevant alternatives but not transitivity):
When only Nothing and A are available, A > Nothing.
When only A and B are available, B > A.
When only Nothing and B are available, Nothing ~ B.
When all three outcomes are available, Nothing ~ B. I’m undecided on how to compare A to Nothing and B, other than that its comparison with Nothing and its comparison with B are the same. I have some sympathy for different ways of comparing A to the other two.
Now, I can say how I’d act, given the above.
If I already pressed button 1 and Nothing is no longer attainable, then we’re in case 2, so pressing button 2 and so pressing both buttons is better than only pressing button 1, because it means choosing B over A.
If starting with all three options still available, and I expect with certainty that if I press button 1, Nothing will no longer be available and I will then press button 2 — say because I know I will follow the rankings in the previous paragraph at that point —, then the outcome of pressing button 1 is B, by backward induction. Then I would be indifferent between pressing button 1 and getting outcome B, and not pressing it and getting Nothing, because B ~ Nothing.[1]
If starting with all three options still available, and for whatever reason I think there’s a chance I won’t press button 2 if I press button 1, then using statewise dominance reasoning:
If and because A > Nothing (and because B ~ Nothing) at this point, pressing button 1 would be better than not pressing either button.
If and because A < Nothing (and because B ~ Nothing) at this point, pressing button 1 would be worse than not pressing either button.
If and because A ~ Nothing (and because B ~ Nothing) at this point, I’d be indifferent.
Similarly if I’m not 100% sure that button 2 will actually even be available after pressing button 1.
My intuitions are guided mostly by something like the (actualist[2]) object interpretation and participant model of Rabinowicz and Österberg (1996)[3] and backward induction.
We might say I’m in case 3 here, because I’ve psychologically ruled out A knowing I’d definitely pick B over A. But B ~ Nothing whether we’re in case 3 or case 4.
Rabinowicz and Österberg (1996) write:
To the satisfaction and the object interpretations of the preference-based conception of value correspond, we believe, two different ways of viewing utilitarianism: the spectator and the participant models.
According to the former, the utilitarian attitude is embodied in an impartial benevolent spectator, who evaluates the situation objectively and from the ‘outside’. An ordinary person can approximate this attitude by detaching himself from his personal engagement in the situation. (...)
The participant model, on the other hand, puts forward as a utilitarian ideal an attitude of emotional participation in other people’s projects: the situation is to be viewed from ‘within’, not just from my own perspective, but also from the others’ points of view. The participant model assumes that, instead of distancing myself from my particular position in the world, I identify with other subjects: what it recommends is not a detached objectivity but a universalized subjectivity.
and
the object interpretation presupposes a subjectivist (or ‘projectivist’) theory of value. Values are not part of the mind-independent world but something that we project upon the world, or — more precisely — upon the whole set of possible worlds. In this sense, our intrinsic value claims, while not world-bound in their range of application, constitute an expression of a particular world-bound perspective: the perspective determined by the preferences we actually have.
Your argument is implicitly assuming IIA.
On a person-affecting view violating IIA but not transitivity, we could have the following:
button 1 neither, when exactly these two options are available
both buttons button 1, when exactly these two options are available
both buttons neither, when exactly these two options are available
button 1 both buttons neither, when exactly these three options are available
There’s no issue for transitivitiy, because the 4 cases involve 4 distinct relations (distinguished by their subsripts), each of which is transitive. The 4 relations don’t have to agree.
FIrst of all, so long as we buy the transitivity of the better than relation that won’t work.
This isn’t true. I can just deny the independence of irrelevant alternatives instead.
Second, it’s highly counterintuitive that the addition of extra good options makes an action worse.
It’s highly counterintutive to you. It’s intuitive to me because I’m sympathetic to the reasons that would justify it in some cases, and I outlined how this would work on my intuitions. The kinds of arguments you give probably aren’t very persuasive to people with strong enough person-affecting intuitions, because those intuitions justify to them what you find counterintuitive.
I find it crazy and I think nearly all people do.
This doesn’t seem like a reason that should really change anyone’s mind about the issue. Or, at least not the mind of any moral antirealist like me.
I suppose a moral realist could be persuaded via epistemic modesty, but if you are epistemically modest, then this will undermine your own personal views that aren’t (near-)consensus (among the informed). For example, you should give more weight to nonconsequentialist views.
By what standard are you judging it to be crazy? I don’t think the view that there are no good states is crazy, and I’m pretty sympathetic to it myself. The view that it’s good to create beings for their own sake is totally unintuitive to me (although I wouldn’t call it or really any other view crazy).
How I would personally deal with your hypothetical under the kind of person-affecting views to which I’m sympathetic is this:
We don’t have reason to press the first button if we’d expect to later undo the welfare improvement of the original person with the second button. This sequence of pressing both isn’t better on person-affecting intuitions than doing nothing. When you reason about what to do, you should, in general, use backwards induction and consider what options you’ll have later and what you’d do later.
If you don’t use backwards induction, you will tend to do worse than otherwise and can be exploited, e.g. money pumped. This is true even for total utilitarians.
Congratulations on the publication!
FWIW, I don’t find the denial of Sequential Desirability very counterintuitive, if and when it’s done for certain person-affecting reasons, precisely because I am quite sympathetic to those person-affecting reasons. The discussion in the comments here seems relevant.
Also, a negative utilitarian would deny the coherence of Generative Improvement, because there’s no positive utility. You could replace it with an improvement and generating a person with exactly 0 utility, or with utility less than the improvement. But from there, Modification Improvement is not possible.
I wonder if the effect is just too small to detect through the noise and other trends and shocks in animal product consumption, but it still exists and is still big in absolute terms for animals.
1.3 million people in Great Britain going vegan for the first time in January 2019 (source1, source2) out of 65 million people living in Great Britain is only about 2% of the population of Great Britain,[1] so we’d expect at most a 2% negative demand shift on meat purchses. Those going vegan were also probably eating less meat on average, so 2% would be an overestimate. And then you also have to adjust for elasticities, so the actual effect on meat production or sales would probably be <1%. That could be hard to detect, depending on how large differences in meat sales year-over-year and December to January typically are.
On the other hand, they report 1.31 million as 4.7% of the total GB adult population, but this seems wrong to me:
They came to the conclusion that 1.31 million people gave up animal products in Britain during January 2019 – that’s 4.7% of the total GB adult population and ten times the number of UK sign ups through the Veganuary website during the same time.
This would imply a GB adult population of 1.31 million / 0.047 = 27.9 million. But there were fewer than 16 million people under 18 in the UK in 2023 so GB should have an adult population of at least around 65 million (all GB) − 16 million (UK non-adults) = 49 million. I don’t see how they got 4.7%.
Also, this January spiking pattern had been most clear and pronounced in the UK, where Veganuary has been most active (from the Economist):
Related thread here.