The error bars on the Rethink Priorities’ welfare ranges are huge. They tell us very little, and making calculations based on them will tell you very little.
I think without some narrower error bars to back you up, making a post suggesting “welfare can be created more efficiently via small non-human animals” is probably net negative, because it has the negative impact of contributing to the EA community looking crazy without the positive impact of a well-supported argument.
Hi Henry! While the 90% confidence intervals for the RP welfare ranges are indeed wide, this is because they’re coming from a mixture of several theories/models of welfare. The uncertainty within a given theory/model of welfare is much lower, and you might have more or less credence in any individual model.
Additionally, if we exclude the neuron count model, the welfare ranges from the mixture of all the other models have narrower distributions.
The error bars on the Rethink Priorities’ welfare ranges are huge. They tell us very little, and making calculations based on them will tell you very little.
I have now added the 5th and 95th percentiles. Thanks for the nudge!
I think without some narrower error bars to back you up, making a post suggesting “welfare can be created more efficiently via small non-human animals” is probably net negative, because it has the negative impact of contributing to the EA community looking crazy without the positive impact of a well-supported argument.
I think the post is still beneficial, because I am not endorsing taking any specific actions to create welfare via small non-human animals. However, I think you have a good point, and I agree the post could plausibly be harmful (although my best guess is that it is beneficial!). I would only disagree with views strongly asserting that the post is harmful.
The error bars on the Rethink Priorities’ welfare ranges are huge. They tell us very little, and making calculations based on them will tell you very little.
I think without some narrower error bars to back you up, making a post suggesting “welfare can be created more efficiently via small non-human animals” is probably net negative, because it has the negative impact of contributing to the EA community looking crazy without the positive impact of a well-supported argument.
Hi Henry! While the 90% confidence intervals for the RP welfare ranges are indeed wide, this is because they’re coming from a mixture of several theories/models of welfare. The uncertainty within a given theory/model of welfare is much lower, and you might have more or less credence in any individual model.
Additionally, if we exclude the neuron count model, the welfare ranges from the mixture of all the other models have narrower distributions.
Here’s a document that explains the different theories/models used: https://docs.google.com/document/d/1xUvMKRkEOJQcc6V7VJqcLLGAJ2SsdZno0jTIUb61D8k/edit
And here’s a spreadsheet with all the confidence intervals from each theory/model individually (after adjusting for probability of sentience): https://docs.google.com/spreadsheets/d/1SpbrcfmBoC50PTxlizF5HzBIq4p-17m3JduYXZCH2Og/edit
Hi Henry,
To be honest, that is a quite funny meme!
I have now added the 5th and 95th percentiles. Thanks for the nudge!
I think the post is still beneficial, because I am not endorsing taking any specific actions to create welfare via small non-human animals. However, I think you have a good point, and I agree the post could plausibly be harmful (although my best guess is that it is beneficial!). I would only disagree with views strongly asserting that the post is harmful.
PS: I upvoted your comment.