Thanks for engaging, James! I agree with all these limitations: we flagged several of them from the start and have mentioned others elsewhere. The only points I’ll add are:
The uncertainties cut in different directions. For instance, correcting for the “understudied species” problem would give us higher welfare range estimates for some animals, like fish and flies, whereas correcting for the “scalar not binary traits” problem would give us lower estimates. It’s an open question how those corrections would net out. But almost certainly not at the precise numbers we gave! (That’s one of many reasons why I’ve never taken the precise numbers all that seriously.)
One of many comms failures associated with the Moral Weight Project is that we didn’t do enough to stress the background considerations that led us to set up the MWP the way we did. The models are toy models; no one should trust them on their own. They make sense, insofar as they do, in the context of (a) more general arguments for thinking that, given hedonism, we’re likely to end up with small differences between species, and (b) a particular view about how best to navigate the uncertainties associated with making these estimates in the first place (basically, defaulting to an approach that minimizes differences we can’t justify, which we discuss both in the original post and in the book). If you think that, conditional on hedonism, nonhuman animals probably aren’t that different from human animals and that we shouldn’t give our priors that much weight, so the burden of proof is on any postulated differences, then I think what we did makes some sense.
I agree: low confidence is the right orientation. That being said, we should have low confidence in the alternatives! We’re all fumbling in the dark here. My hope is just that we can, as a community, do better over time.
Thanks for engaging, James! I agree with all these limitations: we flagged several of them from the start and have mentioned others elsewhere. The only points I’ll add are:
The uncertainties cut in different directions. For instance, correcting for the “understudied species” problem would give us higher welfare range estimates for some animals, like fish and flies, whereas correcting for the “scalar not binary traits” problem would give us lower estimates. It’s an open question how those corrections would net out. But almost certainly not at the precise numbers we gave! (That’s one of many reasons why I’ve never taken the precise numbers all that seriously.)
One of many comms failures associated with the Moral Weight Project is that we didn’t do enough to stress the background considerations that led us to set up the MWP the way we did. The models are toy models; no one should trust them on their own. They make sense, insofar as they do, in the context of (a) more general arguments for thinking that, given hedonism, we’re likely to end up with small differences between species, and (b) a particular view about how best to navigate the uncertainties associated with making these estimates in the first place (basically, defaulting to an approach that minimizes differences we can’t justify, which we discuss both in the original post and in the book). If you think that, conditional on hedonism, nonhuman animals probably aren’t that different from human animals and that we shouldn’t give our priors that much weight, so the burden of proof is on any postulated differences, then I think what we did makes some sense.
I agree: low confidence is the right orientation. That being said, we should have low confidence in the alternatives! We’re all fumbling in the dark here. My hope is just that we can, as a community, do better over time.
Thanks for the thoughtful response!