Thank you. I really value this kind of thoughtful engagement, and I think you’re raising exactly the right questions about efficiency and trade-offs.
I want to explore whether Sudan might actually meet the cost-effectiveness bar, even when compared to GiveWell top charities. Here’s what the independent research suggests:
Efficiency indicators that might rival malaria prevention:
Volunteers working unpaid for over two years = essentially 0% overhead
Operating in areas where international responders cannot reach = extremely high counterfactual impact
Community-based accountability without expensive M&E infrastructure
Direct cash to services: when funding goes through intermediaries, ERRs complain that 10% administrative fees could instead fund more relief supplies or volunteer stipends
The epistemic challenge: You’re right that we lack the kind of rigorous cost-per-life-saved calculations GiveWell produces. But I want to distinguish between two scenarios:
“We studied this and it’s not cost-effective”
“We can’t study this during an active war, so we’ll assume it’s not cost-effective”
The independent research from SSHAP (London School of Hygiene & Tropical Medicine) and ACAPS confirms ERRs are operating at massive scale. Between 2023-2024, they provided first aid, delivered medicines, mapped evacuation routes, supported IDPs, ran communal kitchens, distributed food, and operated hospitals. We just can’t quantify it the way we’d like.
A question for the community: Should EA develop any framework for responding to acute crises where traditional cost-effectiveness analysis isn’t possible? Or is our position that if we can’t measure it with near-certainty, we won’t fund it—even during famines?
I’m genuinely asking, because I think this gets at something important about what EA is for. If the answer is “we focus exclusively on interventions we can measure precisely over long timeframes,” that’s a legitimate choice, but it means explicitly ceding all emergency response to non-EA actors. I’m not sure that’s the right call, especially when volunteer networks with exceptional efficiency characteristics are already operating at scale.
Those are all great questions. i think with option 2 i wouldn’t assume it’s not cost effective, its often just that we don’t know. I for one would be surprised if there weren’t really cost effective places to donate in Sudan, to just that it’s hard to know which ones.
With acute crisis as well i think there is often an assumption that they are getting relatively well funded anyway, but like you say that might not be the case any more.
i think if you made a decent case for one particular situation that was cost effective and needed funding you night be able to convince folks here.
“Should EA develop any framework for responding to acute crises where traditional cost-effectiveness analysis isn’t possible? Or is our position that if we can’t measure it with near-certainty, we won’t fund it—even during famines?”
This is tricky. I think that most[1] of EA is outside of global health/welfare, and much of this is incredibly speculative. AI safety is pretty wild, and even animal welfare work can be more speculative.
GiveWell has historically represented much of the EA-aligned global welfare work. They’ve also seemed to cater to particularly risk-averse donors, from what I can tell.
So an intervention like this is in a tricky middle-ground, where it’s much less speculative than AI risk, but more speculative than much of the GiveWell spend. This is about the point where you can’t really think of “EA” as one unified thing with one utility function. The funding works much more as a bunch of different buckets with fairly different criteria.
Bigger-picture, EAs have a very small sliver of philanthropic spending, which itself is a small sliver of global spending. In my preferred world we wouldn’t need to be so incredibly ruthless with charity choices, because there would just be much more available.
[1] In terms of respected EA discussions/researchers.
Thank you. I really value this kind of thoughtful engagement, and I think you’re raising exactly the right questions about efficiency and trade-offs.
I want to explore whether Sudan might actually meet the cost-effectiveness bar, even when compared to GiveWell top charities. Here’s what the independent research suggests:
Efficiency indicators that might rival malaria prevention:
Volunteers working unpaid for over two years = essentially 0% overhead
Operating in areas where international responders cannot reach = extremely high counterfactual impact
Community-based accountability without expensive M&E infrastructure
Direct cash to services: when funding goes through intermediaries, ERRs complain that 10% administrative fees could instead fund more relief supplies or volunteer stipends
The epistemic challenge: You’re right that we lack the kind of rigorous cost-per-life-saved calculations GiveWell produces. But I want to distinguish between two scenarios:
“We studied this and it’s not cost-effective”
“We can’t study this during an active war, so we’ll assume it’s not cost-effective”
The independent research from SSHAP (London School of Hygiene & Tropical Medicine) and ACAPS confirms ERRs are operating at massive scale. Between 2023-2024, they provided first aid, delivered medicines, mapped evacuation routes, supported IDPs, ran communal kitchens, distributed food, and operated hospitals. We just can’t quantify it the way we’d like.
A question for the community: Should EA develop any framework for responding to acute crises where traditional cost-effectiveness analysis isn’t possible? Or is our position that if we can’t measure it with near-certainty, we won’t fund it—even during famines?
I’m genuinely asking, because I think this gets at something important about what EA is for. If the answer is “we focus exclusively on interventions we can measure precisely over long timeframes,” that’s a legitimate choice, but it means explicitly ceding all emergency response to non-EA actors. I’m not sure that’s the right call, especially when volunteer networks with exceptional efficiency characteristics are already operating at scale.
What do you think?
Those are all great questions. i think with option 2 i wouldn’t assume it’s not cost effective, its often just that we don’t know. I for one would be surprised if there weren’t really cost effective places to donate in Sudan, to just that it’s hard to know which ones.
With acute crisis as well i think there is often an assumption that they are getting relatively well funded anyway, but like you say that might not be the case any more.
i think if you made a decent case for one particular situation that was cost effective and needed funding you night be able to convince folks here.
“Should EA develop any framework for responding to acute crises where traditional cost-effectiveness analysis isn’t possible? Or is our position that if we can’t measure it with near-certainty, we won’t fund it—even during famines?”
This is tricky. I think that most[1] of EA is outside of global health/welfare, and much of this is incredibly speculative. AI safety is pretty wild, and even animal welfare work can be more speculative.
GiveWell has historically represented much of the EA-aligned global welfare work. They’ve also seemed to cater to particularly risk-averse donors, from what I can tell.
So an intervention like this is in a tricky middle-ground, where it’s much less speculative than AI risk, but more speculative than much of the GiveWell spend. This is about the point where you can’t really think of “EA” as one unified thing with one utility function. The funding works much more as a bunch of different buckets with fairly different criteria.
Bigger-picture, EAs have a very small sliver of philanthropic spending, which itself is a small sliver of global spending. In my preferred world we wouldn’t need to be so incredibly ruthless with charity choices, because there would just be much more available.
[1] In terms of respected EA discussions/researchers.