Thanks for the comment, Jason.

Could you say more about the relevance you perceive to the theory and/or practice of effective altruism? 

I was thinking the post could be helpful to make people reflect about whether they are someway elevating/suppressing content based too much on agreement/disagreement, and too little based on whether it could update views.

Thanks for comment, titotal. Agreed.

Now, I’m not saying this is that likely (though it does have surprisingly good arguments in its favor). But if this theory is true, it implies that brains which have about 86 billion neurons would contain around 2^86 billion conscious subsystems. It implies that a single African elephant with about 257 billion neurons has orders of magnitude more moral worth than all humans on Earth, on account of its staggeringly large number of conscious subsystems.

You seem to be using your intuition to assess the plausibility of different ways of comparing welfare across species, but this can easily lead to conclusions which strongly contradict empirical evidence. The gravitational force between 2 masses is proportional to "distance"^-2. An exponent like 0 instead of -2 may seem plausible to people who are not informed, but its probability in reality is practically 0. An exponent of 0 would imply there is no gravitational force.

I think individual welfare per animal-year proportional to 2^"number of neurons" would imply energy mattering implausibly little to produce welfare. A human with 1 k more neurons, or 1.16*10^-6 % (= 1*10^3/(86*10^9)) more, would have basically the same energy consumption, but their welfare per human-year would be 2^(1*10^3) = 10^(log10(2)*10^3) = 10^301 times as large.

If you think about the future of humanity as like the life of a person, then Longtermism starts to look really obvious. Of course the stuff after the first few seconds would matter more than the first few seconds. Of course the first five months matter less than what comes later.

What could a longtermist newborn who is a few seconds old do differently from a random newborn with the goal of maximising their expected future welfare? I see this as the relevant question in your analogy, and I believe there is nothing the longtermist newborn could do differently. Regardless of their goals, it would make sense for them to initially focus on themselves and their close family. From my perspective, the lower the level of foresight, the more it makes sense to focus on increasing the welfare of beings who are closer in time, space, or motivations, even if the ultimate goal is increasing the future welfare of the whole universe. I think humans currently have enough foresight to try to increase the welfare of all animals. However, I am sceptical humans currently have the ability to predict what will increase the welfare of the universe by the time their descendents (biological or not) have 10^50 QALY of total welfare.

Nice post, Richard! Do you have any takes on potential ways people in the effective altruism community are morally self-indulging, even if to a lesser extent than in the examples you mention?

Thanks for clarifying!

This is partly because we don’t rely solely on SADs averted per dollar when interpreting our CEAs, but also consider metrics like animals helped per dollar.

Could you elaborate on why you use the number of animals helped per $? In the context of interventions targeting humans, that is analogous to using people helped per $ in addition to QALYs per $. I wonder whether you could change the estimates for SADs such that they account for the reasons which make you independently value the number of animals helped. Using animals helped per $ favours interventions targeting smaller animals with a lower capacity for welfare, so you may want to use a higher welfare range for these.

WAI has funded many projects on invertebrates, but my impression is that most of its spending targets vertebrates. I would find it helpful to know which fraction of their marginal funding supports projects on invertebrates.

Below is very helpful context from @Casey Darnley. Historically, 9.39 % (= 0.46/4.9) of the funds have supported projects on vertebrates, but there is nuance.

Hi @Vasco Grilo🔸 

If you're hoping for a precise breakdown, such as "X% to vertebrates, Y% to invertebrates," our grants program history shows $4.9M allocated to vertebrates (including fish/rodents) and $0.46M allocated to invertebrates. That said, over the past two years, our support for invertebrate projects has nearly doubled compared to our first two years (1.85x). We've started seeing more strong proposals focused on invertebrates and fish, with researchers telling us they heard about us as a group keen on invertebrate welfare, which is a genuinely encouraging sign that our field-building efforts are working.

However, those numbers don't fully capture what we're actually trying to achieve. Many of our grants and internal research projects develop methods, data, and tools that serve wild animals broadly, rather than focusing on a single specific taxon. Many projects start with a particular species but have much broader applications as the science develops. Many of our grants are meta-projects (e.g., modeling frameworks or welfare measurement tools) with potential that extends beyond vertebrates alone.

Our goal is to establish a research ecosystem that benefits all wild animals, including invertebrates, while striking a balance between pushing new research areas and keeping people excited to contribute. Species-type tracking misses how resources multiply and ripple through the field.

We want to be mindful of how we spend our time, so unless there are significant updates or developments, we won't be posting more on this thread. We are always happy to reconnect down the line if there's something meaningful to add.

Thanks for clarifying, and providing a breakdown, Casey!

Hi Adam. I have not looked recently into the underlying claims of this comment. I agree it is good for organisations to correct mistakes. At the same time, metrics like animals helped per $ are quite uncertain due to the uncertain acceleration of the welfare reforms, and attribution of credit to the organisations involved. When an organisation says they help X animals per $, I understand it can easily help X/3 animals per $. So I do not worry much about an organisation taking a few months to update from, for example, X to 80.5 % of X (= 285/354) animals helped per $. To save time, I believe it often makes sense to collect feedback across a longer period of time, then look into it over a shorter period, and then make all the public updates at once. I also think organisations taking time to update metrics in situations like this is very far from what I would consider fraud.

Great points, James!

But overall, yes, there might be some period of time where supply is reduced, so prices are increased and demand drops slightly.

I estimate the reduction in consumption is (1 - "cumulative elasticity factor (CEF)")*"annual production of the target farm"*"expected years of delay of the start of the farm's operations (D)". 

CEF = "price elasticity of supply"/("price elasticity of supply" - "price elasticity of demand"). CEF = 1 if supply was infinitely more elastic than demand. In this case, there would be no reduction in consumption because this would be solely determined by demand. CEF = 0 if supply was infinitely more inelastic than demand. In this case, a leftwards shift in the supply curve would directly translate into a reduction in consumption because this would be solely determined by supply.

Figure 8.2 of Norwood and Lusk (2011) has values for CEF. For chicken meat, CEF = 0.76, which means a leftwards shift in the demand curve by 1 kg decreases consumption by 0.76 kg. It also means a leftwards shift in the supply curve by 1 kg decreases consumption by 0.24 kg (= 1 - 0.76). This suggests the years of impact from targeting a broilers' farm are 24 % of those one would expect if consumption of chicken meat was solely determined by supply.

The delay can be calculated from D = "probability of i) farm being built in the original place"*"delay of the start of the farm's operations given i)" + "probability of ii) farm being built elsewhere"*"delay of the start of the farm's operations given given ii)" + "probability of iii) farm not being built"*"lifetime of the farm".