Well, firstly, how much credence should we assign the actual analysis in that post?
Before we begin talking about how we should behave “even if”the cost per life saved is much higherthan 5k—is there some consensus as to whether the actual facts and analysis of that post are actually true or even somewhat credible? (separate from the conclusions, which, I agree, seem clearly wrong for all the reasons you said).
As in, if they had instead titled the post “Givewell’s Cost-Per-Life-Saved Estimates are Impossibly Low” and concluded “if the cost per life saved estimate was truly that low, we could have already gone ahead and saved all the cheap lives, and the cost would be higher—so there’s something deeply wrong here”… would people be agreeing with it?
(Because if so, shouldn’t the relevant lower bound for cpls on the impact evaluations be updated if they’re wrong,and shouldn’t that probably be the central point of discussion?
And if not...we should probably add a note clarifying for any reader joining the discussion late, that we’re not actually sure whether the post is correct or not, before going into the implications of the conclusions. We certainly wouldn’t want to start thinking that there aren’t lives that can be saved at low cost if there actually are)
Dropping in late to note that I really like the meta-point here: It’s easy to get caught up in arguing with the “implications” section of a post or article before you’ve even checked the “results” section. Many counterintuitive arguments fall apart when you carefully check the author’s data or basic logic.
(None of the points I make here are meant to apply to Ben’s points—these are just my general thoughts on evaluating ideas.)
Put another way, arguments often take the form:
If A, then B
A
Therefore, B
It’s tempting to attack “Therefore, B” with anti-B arguments C, D, and E, but I find that it’s usually more productive to start by checking the first two points. Sometimes, you’ll find issues that render “Therefore, B” moot; other times, you’ll see that the author’s facts check out and find yourself moving closer to agreement with “Therefore, B”. Both results are valuable.
I don’t think the post is correct in concluding that the current marginal cost-per-life-saved estimates are wrong. Annual malaria deaths are around 450k, and if you gave the Against Malaria Foundation $5k * 450k ($2.3B) they would not be able to make sure no one died from malaria in 2020, but still wouldn’t give much evidence that $5k was too low an estimate for the marginal cost. It just means that AMF would have lots of difficulty scaling up so much, that some deaths can’t be prevented by distributing nets, that some places are harder to work in, etc.
It does mean that big funders have seen the current cost-per-life saved numbers and decided not to give those organizations all the money they’d be able to use at that cost-effectiveness. But there are lots of reasons other than what Ben gives for why you might decide to do that, including:
You have multiple things you care about and are following a strategy of funding each of them some. For example, OpenPhil has also funded animal charities and existential risk reduction.
You don’t want a dynamic where you’re responsible for the vast majority of a supposedly independent organization’s funding.
You think better giving opportunities may become available in the future and want to have funds if that happens.
Well, firstly, how much credence should we assign the actual analysis in that post?
Before we begin talking about how we should behave “even if” the cost per life saved is much higher than 5k—is there some consensus as to whether the actual facts and analysis of that post are actually true or even somewhat credible? (separate from the conclusions, which, I agree, seem clearly wrong for all the reasons you said).
As in, if they had instead titled the post “Givewell’s Cost-Per-Life-Saved Estimates are Impossibly Low” and concluded “if the cost per life saved estimate was truly that low, we could have already gone ahead and saved all the cheap lives, and the cost would be higher—so there’s something deeply wrong here”… would people be agreeing with it?
(Because if so, shouldn’t the relevant lower bound for cpls on the impact evaluations be updated if they’re wrong, and shouldn’t that probably be the central point of discussion?
And if not...we should probably add a note clarifying for any reader joining the discussion late, that we’re not actually sure whether the post is correct or not, before going into the implications of the conclusions. We certainly wouldn’t want to start thinking that there aren’t lives that can be saved at low cost if there actually are)
Dropping in late to note that I really like the meta-point here: It’s easy to get caught up in arguing with the “implications” section of a post or article before you’ve even checked the “results” section. Many counterintuitive arguments fall apart when you carefully check the author’s data or basic logic.
(None of the points I make here are meant to apply to Ben’s points—these are just my general thoughts on evaluating ideas.)
Put another way, arguments often take the form:
If A, then B
A
Therefore, B
It’s tempting to attack “Therefore, B” with anti-B arguments C, D, and E, but I find that it’s usually more productive to start by checking the first two points. Sometimes, you’ll find issues that render “Therefore, B” moot; other times, you’ll see that the author’s facts check out and find yourself moving closer to agreement with “Therefore, B”. Both results are valuable.
I don’t think the post is correct in concluding that the current marginal cost-per-life-saved estimates are wrong. Annual malaria deaths are around 450k, and if you gave the Against Malaria Foundation $5k * 450k ($2.3B) they would not be able to make sure no one died from malaria in 2020, but still wouldn’t give much evidence that $5k was too low an estimate for the marginal cost. It just means that AMF would have lots of difficulty scaling up so much, that some deaths can’t be prevented by distributing nets, that some places are harder to work in, etc.
It does mean that big funders have seen the current cost-per-life saved numbers and decided not to give those organizations all the money they’d be able to use at that cost-effectiveness. But there are lots of reasons other than what Ben gives for why you might decide to do that, including:
You have multiple things you care about and are following a strategy of funding each of them some. For example, OpenPhil has also funded animal charities and existential risk reduction.
You don’t want a dynamic where you’re responsible for the vast majority of a supposedly independent organization’s funding.
You think better giving opportunities may become available in the future and want to have funds if that happens.