Hey Vasco! Taking into account moral uncertainty over the neuron count exponent, your plot would still make the animal interventions you listed look far higher EV than GiveWell. The probability mass where the exponent is between 0 and 1, making the animal interventions look several OOMs better than GiveWell, would swamp the cases where the exponent is >1.
(Yes, this runs into the two envelopes problem, but I think there are good arguments for using human welfare as the unit of account.)
Furthermore, I personally don’t find neuron count exponents >1 as plausible as you do. If I’m interpreting this plot from your linked source post correctly, for broiler chickens, exponent 1 implies welfare range 1⁄500 and exponent 2 implies welfare range 1⁄100,000. I agree that these numbers would make GiveWell look better, but I don’t find those welfare ranges intuitively plausible.
Taking into account moral uncertainty over the neuron count exponent, your plot would still make the animal interventions you listed look far higher EV than GiveWell. The probability mass where the exponent is between 0 and 1, making the animal interventions look several OOMs better than GiveWell, would swamp the cases where the exponent is >1.
You are distributing the probability mass roughly evenly across the potential models (values of the exponent)? I worry about giving weights to models based on practically no evidence. In Bob Fischer’s book about comparing welfare across species, there is just this justifying the weights (I read the whole book).
When we generated the mixture model, we assigned 60 percent weight to the simple additive model, 30 percent to the neurophysiological model, and 10 percent to the equality model. We did this because we suspect that collecting empirical data on the presence or absence of welfare-related traits is a more reliable methodology for generating welfare range estimates than using either the neurophysiological or equality models. However, the proper weight to give is the subject of a reasonable debate.
People usually give weights that are at least 0.1/”number of models”, which is at least 3.33 % (= 0.1/3) for 3 models, when it is quite hard to estimate the weights. However, giving weights which are not much smaller than the uniform weight of 1/”number of models” could easily lead to huge mistakes. As a silly example, if I asked random people with age 7 about whether the gravitational force between 2 objects is proportional to “distance”^-2 (correct answer), “distance”^-20, or “distance”^-200, I imagine I would get a significant fraction picking the exponents of −20 and −200. Assuming 60 % picked −2, 20 % picked −20, and 20 % picked −200, one may naively conclude the mean exponent of −45.2 (= 0.6*(-2) + 0.2*(-20) + 0.2*(-200)) is reasonable. Yet, there is lots of empirical evidence against this which the respondants are not aware of. The right conclusion would be that the respondants have practically no idea about the right exponent because they would not be able to adequately justify their picks.
If I’m interpreting this plot from your linked source post correctly, for broiler chickens, exponent 1 implies welfare range 1⁄500 and exponent 2 implies welfare range 1⁄100,000. I agree that these numbers would make GiveWell look better, but I don’t find those welfare ranges intuitively plausible.
You are reading the graph correctly. Why do you find an exponent of 2 implausible? My position is not so much that I find it plausible. It is more that I do not know how to check the plausibility of values ranging from 0 to 2 or so, and therefore do not want to rule them out.
I’d still expect the magnitude of their direct welfare effects (ignoring indirect effects) to be huge relative to global health
You cannot rule out indirect effects if you are confident the exponent is 0 to 1? In this case, I estimate effects on soil invertebrates are much larger than those on target beneficiaries.
The graph above covers microarthropods (springtails and mites) and nematodes, which are not covered in Bob’s book. However, I have very little idea about whether cage-free egg campaigns increase or decrease welfare due to potentially dominant effects on soil ants and termites alone. These are macroarthropods like shrimps and black soldier flies (BSFs), which are covered in Bob’s book.
Hey Vasco! Taking into account moral uncertainty over the neuron count exponent, your plot would still make the animal interventions you listed look far higher EV than GiveWell. The probability mass where the exponent is between 0 and 1, making the animal interventions look several OOMs better than GiveWell, would swamp the cases where the exponent is >1.
(Yes, this runs into the two envelopes problem, but I think there are good arguments for using human welfare as the unit of account.)
Furthermore, I personally don’t find neuron count exponents >1 as plausible as you do. If I’m interpreting this plot from your linked source post correctly, for broiler chickens, exponent 1 implies welfare range 1⁄500 and exponent 2 implies welfare range 1⁄100,000. I agree that these numbers would make GiveWell look better, but I don’t find those welfare ranges intuitively plausible.
You are distributing the probability mass roughly evenly across the potential models (values of the exponent)? I worry about giving weights to models based on practically no evidence. In Bob Fischer’s book about comparing welfare across species, there is just this justifying the weights (I read the whole book).
People usually give weights that are at least 0.1/”number of models”, which is at least 3.33 % (= 0.1/3) for 3 models, when it is quite hard to estimate the weights. However, giving weights which are not much smaller than the uniform weight of 1/”number of models” could easily lead to huge mistakes. As a silly example, if I asked random people with age 7 about whether the gravitational force between 2 objects is proportional to “distance”^-2 (correct answer), “distance”^-20, or “distance”^-200, I imagine I would get a significant fraction picking the exponents of −20 and −200. Assuming 60 % picked −2, 20 % picked −20, and 20 % picked −200, one may naively conclude the mean exponent of −45.2 (= 0.6*(-2) + 0.2*(-20) + 0.2*(-200)) is reasonable. Yet, there is lots of empirical evidence against this which the respondants are not aware of. The right conclusion would be that the respondants have practically no idea about the right exponent because they would not be able to adequately justify their picks.
You are reading the graph correctly. Why do you find an exponent of 2 implausible? My position is not so much that I find it plausible. It is more that I do not know how to check the plausibility of values ranging from 0 to 2 or so, and therefore do not want to rule them out.
You cannot rule out indirect effects if you are confident the exponent is 0 to 1? In this case, I estimate effects on soil invertebrates are much larger than those on target beneficiaries.
The graph above covers microarthropods (springtails and mites) and nematodes, which are not covered in Bob’s book. However, I have very little idea about whether cage-free egg campaigns increase or decrease welfare due to potentially dominant effects on soil ants and termites alone. These are macroarthropods like shrimps and black soldier flies (BSFs), which are covered in Bob’s book.