Yeah, I don’t think someone doing independent, principled cost-effectiveness analysis would find that the top interventions were anything like GiveWell does, or favor anything like the set of actions that GiveWell and OpenPhil do. I went into some detail in the posts I linked. The relevant standard here is whether which coherent model predicts their actions taken as a whole as the best actions to take, not whether any particular decision is defensible as better than some other particular proposed alternative. (Basically anything can be justified using the latter standard, it’s just a clever arguing contest.)
Suppose I have a radish, a carrot, and a prune. I eat the carrot. Someone asks me why. I respond that I ate the carrot because I like sweet foods, and the carrot was sweeter than the radish. They might reasonably be skeptical of this explanation, since if they’d tried to predict my behavior using the implied decision rule, they’d have predicted that I’d eat the prune rather than the carrot.
You mean something other than the cost-effectiveness process and analysis from their website?
Yeah, I don’t think someone doing independent, principled cost-effectiveness analysis would find that the top interventions were anything like GiveWell does, or favor anything like the set of actions that GiveWell and OpenPhil do. I went into some detail in the posts I linked. The relevant standard here is whether which coherent model predicts their actions taken as a whole as the best actions to take, not whether any particular decision is defensible as better than some other particular proposed alternative. (Basically anything can be justified using the latter standard, it’s just a clever arguing contest.)
I’m confused how you distinguish between predicting vs clever arguing if all our data is in the past?
Suppose I have a radish, a carrot, and a prune. I eat the carrot. Someone asks me why. I respond that I ate the carrot because I like sweet foods, and the carrot was sweeter than the radish. They might reasonably be skeptical of this explanation, since if they’d tried to predict my behavior using the implied decision rule, they’d have predicted that I’d eat the prune rather than the carrot.