If the community is paying back, I think it makes more sense for this to become primarily the responsibility of major funders instead of individual grantees, for practical reasons. I don’t think individual EAs need to take on hardship to pay it back, and we can consider large funders paying on their behalf to be community support work or grants that they would have made anyway. Also, having major funders pay is more efficient than having individual grantees each worry about this.
Still, it might make more sense for this to come from cause-specific budgets, so that longtermist funding from FTX and associates is paid back from the longtermist EA budget, and global health and development funding from FTX and associates is paid back from the GHD EA budget, and so on. You shouldn’t be able to shift EA funding between causes by committing fraud.
Ehhh, maybe? I feel like giving money back to people EAs took advantage of isn’t part of the “longtermist cause”; it’s part of the “do right by people who EAs harmed” cause.
I don’t particularly mind if that money comes from the longtermist budget, but I wouldn’t want to delay the “doing right” part on that account.
Like, you could claim that it’s only longtermists who are responsible, because SBF was a longtermist. But you could also claim that it’s only a narrower group that’s responsible, because SBF had lots of specific views that aren’t universal among longtermists. I don’t want to go too far down that road, because carving things up too finely eventually means that no one in EA takes responsibility, since we can all claim to have different views from SBF on some dimension.
I think the main (only?) reason for doing it by cause would be to deter such future harms in the name of specific views, not about righting wrongs to others or attributing responsibility to specific views or groups. Otherwise EAs are more likely to think the expected benefits of unethical behaviour outweigh the expected costs, because they get to shift the expected costs onto other causes they care less about.
I don’t think there needs to be delay here. If Open Phil primarily takes this on, they could pay first and then rebudget later. OTOH, if other longtermist funders take some of this on, it’ll be more clear to the public that longtermists are contributing specifically to paying off their “debts” (even if funding is pretty fungible and this shouldn’t really matter). I agree that carving things up can be pretty complicated, in principle. I guess I don’t expect this to matter too much, but I could see others disagreeing, since I’m less familiar with the disagreements within longtermism. Also, it’s not about attributing responsibility to specific views, but tracking benefits. We know who got grants from FTX and associates, and we want to simulate them paying it back.
(I made some edits to this comment within the first 18 minutes of posting this comment, in case you were already reading or replying.)
EDIT: I guess we know less about where those benefits would have gone otherwise if not for FTX, which is still a problem, and can depend on differences in views within longtermism. This also applies to grantees paying back instead of funders.
Also, I guess another reason is that it’s just more fair to other causes that benefited less that they should pay less.
Not sure why someone decided to (strong) downvote (and not just disagreevote) your comment here. I’ve upvoted your comment, since I think it has useful considerations, and neither agree nor disagree overall, since I don’t actually have a strong enough opinion here.
I guess it’s more complicated than this, because FTX entering as a funder and disproportionately funding longtermism may have freed up funding for other cause areas. Still, I’d guess the counterfactual was disproportionately to the benefit of longtermism. We could try to simulate what would have happened without FTX and check where the extra money went. If only direct beneficiaries pay up, this is unfair to them, and indirect beneficies get away without paying.
Someone raised the point that if EAs try to offset the harm SBF caused, this creates a moral hazard of the form “people may be more willing to cause harm in the name of EA in the future, expecting other EAs to offset that harm”.
I think that’s a stronger objection to “offset all of SBF’s harms” (which I don’t endorse anyway) than to “collectively (at the community level, not the individual org level) give back the amount EA received”, but maybe it will shift my view once I’ve chewed on it a bit more. At a glance, I don’t think I’d expect this concern to be a dominant factor?
Our offsetting their harms out of the budget for things they care most about should be bad by their own lights, though, unless they’re naive consequentialists (which they may disproportionately be). They should care more about the harms this way, since it’s clear it’s counterproductive by their own lights, whereas it’s easy to discount harms to FTX customers relative to longtermist (or generally EA) donations.
Agreed, though I think the primary reason the EA community should collectively give back money that was stolen and given to us is “it’s the right thing to do”.
This is related to incentives, and there are complicated ways in which being a high-integrity, broadly honorable community finds a lot of its justification in game-theoretic LDT-ish arguments, but I think EAs empirically are better at reasoning about “what’s the right thing to do?” than at explicitly reasoning about LDT.
I think that’s fair, but even when we consider the “right thing to do”, there’s still a question of on whom this burden is supposed to fall and how to do this in a fair way. Even for cross-cause funders, if they model themselves as multiple cause-specific agents or representing sections of the EA community, there’s still an issue of being fair to those agents or community sections. I think tracking the counterfactual without FTX/Alameda (or just their bad actions) would be the most accurate way to capture this + maybe some more pooled EA community fund thing.
One way to think about it is to ask who benefited counterfactually, whether directly and indirectly (through counterfactually shifting budgets) from FTX funds, and treat those individuals as owing debt equal to the counterfactual funding received. Some individuals who actually received FTX funding might not have even benefited counterfactually. Then funders can decide which debts to take on, which might happen by cause or worldview, e.g. global health people might not feel like they should take on the debts of longtermists, although they might anyway towards a community pool.
(Also, I don’t think LDT really needs to come into it specifically. Are the different decision theories going to disagree dramatically here? Or, at least, I think they should all recognize some deterence value here, but maybe give it different relative weight.)
If the community is paying back, I think it makes more sense for this to become primarily the responsibility of major funders instead of individual grantees, for practical reasons. I don’t think individual EAs need to take on hardship to pay it back, and we can consider large funders paying on their behalf to be community support work or grants that they would have made anyway. Also, having major funders pay is more efficient than having individual grantees each worry about this.
Still, it might make more sense for this to come from cause-specific budgets, so that longtermist funding from FTX and associates is paid back from the longtermist EA budget, and global health and development funding from FTX and associates is paid back from the GHD EA budget, and so on. You shouldn’t be able to shift EA funding between causes by committing fraud.
Ehhh, maybe? I feel like giving money back to people EAs took advantage of isn’t part of the “longtermist cause”; it’s part of the “do right by people who EAs harmed” cause.
I don’t particularly mind if that money comes from the longtermist budget, but I wouldn’t want to delay the “doing right” part on that account.
Like, you could claim that it’s only longtermists who are responsible, because SBF was a longtermist. But you could also claim that it’s only a narrower group that’s responsible, because SBF had lots of specific views that aren’t universal among longtermists. I don’t want to go too far down that road, because carving things up too finely eventually means that no one in EA takes responsibility, since we can all claim to have different views from SBF on some dimension.
I think the main (only?) reason for doing it by cause would be to deter such future harms in the name of specific views, not about righting wrongs to others or attributing responsibility to specific views or groups. Otherwise EAs are more likely to think the expected benefits of unethical behaviour outweigh the expected costs, because they get to shift the expected costs onto other causes they care less about.
I don’t think there needs to be delay here. If Open Phil primarily takes this on, they could pay first and then rebudget later. OTOH, if other longtermist funders take some of this on, it’ll be more clear to the public that longtermists are contributing specifically to paying off their “debts” (even if funding is pretty fungible and this shouldn’t really matter). I agree that carving things up can be pretty complicated, in principle. I guess I don’t expect this to matter too much, but I could see others disagreeing, since I’m less familiar with the disagreements within longtermism. Also, it’s not about attributing responsibility to specific views, but tracking benefits. We know who got grants from FTX and associates, and we want to simulate them paying it back.
(I made some edits to this comment within the first 18 minutes of posting this comment, in case you were already reading or replying.)
EDIT: I guess we know less about where those benefits would have gone otherwise if not for FTX, which is still a problem, and can depend on differences in views within longtermism. This also applies to grantees paying back instead of funders.
Also, I guess another reason is that it’s just more fair to other causes that benefited less that they should pay less.
Not sure why someone decided to (strong) downvote (and not just disagreevote) your comment here. I’ve upvoted your comment, since I think it has useful considerations, and neither agree nor disagree overall, since I don’t actually have a strong enough opinion here.
I guess it’s more complicated than this, because FTX entering as a funder and disproportionately funding longtermism may have freed up funding for other cause areas. Still, I’d guess the counterfactual was disproportionately to the benefit of longtermism. We could try to simulate what would have happened without FTX and check where the extra money went. If only direct beneficiaries pay up, this is unfair to them, and indirect beneficies get away without paying.
Agreed.
I argued for major funders doing this (at least if some conditions hold) here, and argued against the grant recipients doing this here.
Someone raised the point that if EAs try to offset the harm SBF caused, this creates a moral hazard of the form “people may be more willing to cause harm in the name of EA in the future, expecting other EAs to offset that harm”.
I think that’s a stronger objection to “offset all of SBF’s harms” (which I don’t endorse anyway) than to “collectively (at the community level, not the individual org level) give back the amount EA received”, but maybe it will shift my view once I’ve chewed on it a bit more. At a glance, I don’t think I’d expect this concern to be a dominant factor?
Our offsetting their harms out of the budget for things they care most about should be bad by their own lights, though, unless they’re naive consequentialists (which they may disproportionately be). They should care more about the harms this way, since it’s clear it’s counterproductive by their own lights, whereas it’s easy to discount harms to FTX customers relative to longtermist (or generally EA) donations.
Agreed, though I think the primary reason the EA community should collectively give back money that was stolen and given to us is “it’s the right thing to do”.
This is related to incentives, and there are complicated ways in which being a high-integrity, broadly honorable community finds a lot of its justification in game-theoretic LDT-ish arguments, but I think EAs empirically are better at reasoning about “what’s the right thing to do?” than at explicitly reasoning about LDT.
I think that’s fair, but even when we consider the “right thing to do”, there’s still a question of on whom this burden is supposed to fall and how to do this in a fair way. Even for cross-cause funders, if they model themselves as multiple cause-specific agents or representing sections of the EA community, there’s still an issue of being fair to those agents or community sections. I think tracking the counterfactual without FTX/Alameda (or just their bad actions) would be the most accurate way to capture this + maybe some more pooled EA community fund thing.
One way to think about it is to ask who benefited counterfactually, whether directly and indirectly (through counterfactually shifting budgets) from FTX funds, and treat those individuals as owing debt equal to the counterfactual funding received. Some individuals who actually received FTX funding might not have even benefited counterfactually. Then funders can decide which debts to take on, which might happen by cause or worldview, e.g. global health people might not feel like they should take on the debts of longtermists, although they might anyway towards a community pool.
(Also, I don’t think LDT really needs to come into it specifically. Are the different decision theories going to disagree dramatically here? Or, at least, I think they should all recognize some deterence value here, but maybe give it different relative weight.)