I love this a lot. Something that often gets my hackles/Spidey sense up is when someone seems very confident about a particular cause being the “best” or “better”, especially when making difficult (arguably even impossible like you say) comparisons between animals/current humans/future humans. I think it is helpful to make these difficult comparisons but only with deep humility and huge acknowledged uncertainty.
Another benefit of plurality is that it’s easier to have a “bigger EA tent” both in human resource and funding. Us humans will always have different opinions not only about impact, but also the kind of things that we lean towards naturally and also where our competitive advantages are. The more we prioritize the more we may exclude.
I’m not sure having a “bigger EA tent” leads to more funding/interest, if anything, people may be less likely to fund/support/be interested in a group that supports many different areas rather than the cause they mainly care about. At least it seems like cause specific orgs get much more funding than multi-cause/EA orgs.
Yeah I’m not sure we’re really disagreeing here? I agree people are less interested in as group that supports different areas and that orgs should mostly be cause specific. I’m talking about having a lot of grace for a wide range of “high impact’ causes under the broader EA tent depending on people’s epistemics and cost effectiveness calculation methods. I think this is more helpful than doubling down on prioritisation and leaving groups our causes feeling like they might be on the “outer” edge of EA or excluded completely
I think it is helpful to make these difficult comparisons but only with deep humility and huge acknowledged uncertainty.
I think acknowledging uncertainty implies that the best animal welfare interventions are way more cost-effective that the best human welfare interventions. Below are RP’s welfare ranges of chickens conditional on sentience for the 12 models RP has considered. I estimate broiler welfare and cage-free campaigns are 168 and 462 times as cost-effective as GiveWell’s top charities using Rethink Priorities’ (RP’s) median welfare range of chickens of 0.332. The welfare range of chickens under the neuron count model, the one leading to the lowest welfare range, is 0.528 % (= 0.002*0.876/0.332) as large as the one I assumed, in which case I would estimate cage-free campaigns to be 2.44 (= 462*5.28*10^-3) times as cost-effective as GiveWell’s top charities. The welfare range of chickens under the cubic model, the one leading to the 2nd lowest welfare range, is 40.1 % (= 0.152*0.876/0.332) as large as the one I assumed, in which case I would estimate cage-free campaigns to be 185 (= 462*0.401) times as cost-effective as GiveWell’s top charities. As a result, the best animal welfare interventions are way more cost-effective than the best human welfare interventions under 11 of the 12 models RP considered. I think one has to be super confident that the neuron count model, or other outputting a similarly low welfare range, is right for the best human welfare interventions to be competitive. I suspect people supporting human welfare interventions would not be that confident on such models on reflection.
I don’t really understand how you’re achknowleging the uncertainty here? Basing on RPs weights is one way of making these comparisons, and even those have enormous uncertainty. I was just saying I don’t like it when someone is very confident that one cause or set of causes are better than others. I think there’s still enormous uncertainty that animal welfare interventions are better than human ones. Are you saying that’s not the case?
As a side note I don’t think there really are a wide range of meaningfully different RP “Models” in the way I think of models anyway. A separate “model” for me implies genuinely different inputs and assumptions, which are basically all the same for RPs models—which is why they differ far less than even an order of magnitude.
”After the project decided to assume hedonism and dismiss neuron count, the cumulative percent of these 90 behavioral proxies became the basis for their welfare range estimates. Although the team used a number of models in their final analysis, these models were mostly based on different weightings of these same behavioral proxies.[12]. Median final welfare ranges are therefore fairly well approximated by the simple formula.
(Behavioral proxy percent) x (Probability of Sentience) = Median Welfare range”
To me then it becomes a question of whether or you agree with the assumptions that RP make along the way (and how strongly)
I think there’s still enormous uncertainty that animal welfare interventions are better than human ones. Are you saying that’s not the case?
I think it is clear that the best animal welfare interventions are much more cost-effective than the best human welfare interventions.
After the project decided to assume hedonism and dismiss neuron count, the cumulative percent of these 90 behavioral proxies became the basis for their welfare range estimates. Although the team used a number of models in their final analysis, these models were mostly based on different weightings of these same behavioral proxies.
The welfare range of chickens is higher than RP’s median under the 2 models besides the neuron count one which do not rely on behaviour:
If one puts at least 10 % weight on the quantitative model, which “aggregates several quantifiably characterizable physiological measurements related to activity in the pain processing system”, the welfare range of chickens will be at least 16.9 % (= 0.1*1.69) of RP’s median.
I love this a lot. Something that often gets my hackles/Spidey sense up is when someone seems very confident about a particular cause being the “best” or “better”, especially when making difficult (arguably even impossible like you say) comparisons between animals/current humans/future humans. I think it is helpful to make these difficult comparisons but only with deep humility and huge acknowledged uncertainty.
Another benefit of plurality is that it’s easier to have a “bigger EA tent” both in human resource and funding. Us humans will always have different opinions not only about impact, but also the kind of things that we lean towards naturally and also where our competitive advantages are. The more we prioritize the more we may exclude.
I’m not sure having a “bigger EA tent” leads to more funding/interest, if anything, people may be less likely to fund/support/be interested in a group that supports many different areas rather than the cause they mainly care about. At least it seems like cause specific orgs get much more funding than multi-cause/EA orgs.
Yeah I’m not sure we’re really disagreeing here? I agree people are less interested in as group that supports different areas and that orgs should mostly be cause specific. I’m talking about having a lot of grace for a wide range of “high impact’ causes under the broader EA tent depending on people’s epistemics and cost effectiveness calculation methods. I think this is more helpful than doubling down on prioritisation and leaving groups our causes feeling like they might be on the “outer” edge of EA or excluded completely
Maybe I’m being too vague here though...
Hi Nick.
I think acknowledging uncertainty implies that the best animal welfare interventions are way more cost-effective that the best human welfare interventions. Below are RP’s welfare ranges of chickens conditional on sentience for the 12 models RP has considered. I estimate broiler welfare and cage-free campaigns are 168 and 462 times as cost-effective as GiveWell’s top charities using Rethink Priorities’ (RP’s) median welfare range of chickens of 0.332. The welfare range of chickens under the neuron count model, the one leading to the lowest welfare range, is 0.528 % (= 0.002*0.876/0.332) as large as the one I assumed, in which case I would estimate cage-free campaigns to be 2.44 (= 462*5.28*10^-3) times as cost-effective as GiveWell’s top charities. The welfare range of chickens under the cubic model, the one leading to the 2nd lowest welfare range, is 40.1 % (= 0.152*0.876/0.332) as large as the one I assumed, in which case I would estimate cage-free campaigns to be 185 (= 462*0.401) times as cost-effective as GiveWell’s top charities. As a result, the best animal welfare interventions are way more cost-effective than the best human welfare interventions under 11 of the 12 models RP considered. I think one has to be super confident that the neuron count model, or other outputting a similarly low welfare range, is right for the best human welfare interventions to be competitive. I suspect people supporting human welfare interventions would not be that confident on such models on reflection.
I don’t really understand how you’re achknowleging the uncertainty here? Basing on RPs weights is one way of making these comparisons, and even those have enormous uncertainty. I was just saying I don’t like it when someone is very confident that one cause or set of causes are better than others. I think there’s still enormous uncertainty that animal welfare interventions are better than human ones. Are you saying that’s not the case?
As a side note I don’t think there really are a wide range of meaningfully different RP “Models” in the way I think of models anyway. A separate “model” for me implies genuinely different inputs and assumptions, which are basically all the same for RPs models—which is why they differ far less than even an order of magnitude.
As I said in my post here
”After the project decided to assume hedonism and dismiss neuron count, the cumulative percent of these 90 behavioral proxies became the basis for their welfare range estimates. Although the team used a number of models in their final analysis, these models were mostly based on different weightings of these same behavioral proxies.[12]. Median final welfare ranges are therefore fairly well approximated by the simple formula.
(Behavioral proxy percent) x (Probability of Sentience) = Median Welfare range”
To me then it becomes a question of whether or you agree with the assumptions that RP make along the way (and how strongly)
I think it is clear that the best animal welfare interventions are much more cost-effective than the best human welfare interventions.
The welfare range of chickens is higher than RP’s median under the 2 models besides the neuron count one which do not rely on behaviour:
For the quantitative model, it is 1.69 (= 0.641*0.876/0.332) times as high.
For the equality model, it is 2.51 (= 0.953*0.876/0.332) times as high.
If one puts at least 10 % weight on the quantitative model, which “aggregates several quantifiably characterizable physiological measurements related to activity in the pain processing system”, the welfare range of chickens will be at least 16.9 % (= 0.1*1.69) of RP’s median.