I guess my main disagreement with this piece is that I think core EAs do a pretty good job:
Decisions generally seem pretty good to me (though I guess I am pretty close to the white-male etc artchitype, but still). I think that that community decisionmaking wouldn’t have fixed FTX, the FLI issue or the bostrom email. In fact the criticism of the CEA statement is more from the white-male EA, I guess
You want people empowered to take quick decisions who can be trusted based on their track record.
I wish there was more understanding of community opinion and discussing with the community. I have long argued for a kind of virtual copy of many EA orgs that the community can discuss and criticise.
For the FLI issue, I think we can confidently say more democratic decision making would have helped. Most EAs would have probably thought we should avoid touching a neo Nazi newspaper with a 10 foot pole.
On the other hand, more decentralized grantmaking, i.e. giving money to more individuals to regrant, increases the risks of individuals funding really bad things (unilateralist curse). I suppose we could do something like give regranters money to regrant and allow a group of people (maybe all grantmakers and regranters together) to vote to veto grants, say with a limited number of veto votes per round, and possibly after flagging specific grants and debating them. However, this will increase the required grantmaking-related work, possibly significantly.
I think some sort of community-involved decisionmaking could have reduced the risk of FLI. The community involvement could be on the process side instead of, or in addition to, the substantive side. Although there hasn’t been any answer on how much of a role the family member of FLI’s president played in the grant, the community could have pushed for adoption of strong rules surrounding conflict of interest.
Another model would be a veto jury empowered to nix tentatively approved grants that seemed poorly justified, too risky for expected benefit, or otherwise problematic. Even if the veto jury had missed the whole neo-nazi business, it very likely would have thrown this grant out for being terrible for other reasons.
I don’t think that is necessarily correct. We know the FLI grant was “approved” by September 7 and walked back sometime in November, so there was time for a veto-jury process without delaying FLI’s normal business process.
I am ordinarily envisioning a fairly limited role for a veto jury—the review would basically be for what US lawyers call “abuse of discretion” and might be called in less technical jargon a somewhat enhanced sanity check. Cf. Harman v. Apfel, 211 F.3d 1172, 1175 (9th Cir. 2000) (noting reversal under abuse of discretion standard is possible only “when the appellate court is convinced firmly that the reviewed decision lies beyond the pale of reasonable justification under the circumstances”).
It would not be an opportunity for the jury to comprehensively re-balance the case for and against funding the grant, or merely substitute its judgment for that of the grantmaker. Perhaps there would be defined circumstances in which veto juries would work with a less deferential standard of review, but it ordinarily should not take much time to determine that there was no abuse of discretion.
The “ceteris paribus” is the key part here, and I think in real life fast processes for deciding on huge sums of money tend to do much worse than slower ones.
If someone says that A is worse than B because it has a certain property C, you shouldn’t ask “Why is C bad?” if you are not disputing the badness of C. It would be much clearer to say, “I agree C is bad, but A has other properties that make it better than B on balance.”
Re: FTX—I’m not sure what would have “fixed FTX”, but I did think there were decisions here that made an impact. I don’t think EA is to blame for the criminal activities, but we did platform SBF a lot (and listened to him talk about things he had no idea about), and sort of had a personality cult around him (like we still do aroung Will, for example). You can see this from comments of people saying how disappointed they were and how much they had looked up to him.
So there are some cultural changes that would’ve had a chance of making it better—like putting less emphasis on individuals and more on communities; like leaving hard domain-knowledge questions to experts; like being wary of crypto, as a technology that involves a lot of immoral actions in practice from all big actors.
Re: Bostrom—Part of the reason that EA was so susceptible to backlash from his writings was that the movement made him a central figure despite his already fringe and alarming views, which IMO should have been a red flag even before we knew about the outright racism. And this, I think, is a result of lack of diversity in the movement, and even more so in decisionmaking circles.
I guess my main disagreement with this piece is that I think core EAs do a pretty good job:
Decisions generally seem pretty good to me (though I guess I am pretty close to the white-male etc artchitype, but still). I think that that community decisionmaking wouldn’t have fixed FTX, the FLI issue or the bostrom email. In fact the criticism of the CEA statement is more from the white-male EA, I guess
You want people empowered to take quick decisions who can be trusted based on their track record.
I wish there was more understanding of community opinion and discussing with the community. I have long argued for a kind of virtual copy of many EA orgs that the community can discuss and criticise.
For the FLI issue, I think we can confidently say more democratic decision making would have helped. Most EAs would have probably thought we should avoid touching a neo Nazi newspaper with a 10 foot pole.
Oh okay, but grant level democratisation is a huge step.
On the other hand, more decentralized grantmaking, i.e. giving money to more individuals to regrant, increases the risks of individuals funding really bad things (unilateralist curse). I suppose we could do something like give regranters money to regrant and allow a group of people (maybe all grantmakers and regranters together) to vote to veto grants, say with a limited number of veto votes per round, and possibly after flagging specific grants and debating them. However, this will increase the required grantmaking-related work, possibly significantly.
I think some sort of community-involved decisionmaking could have reduced the risk of FLI. The community involvement could be on the process side instead of, or in addition to, the substantive side. Although there hasn’t been any answer on how much of a role the family member of FLI’s president played in the grant, the community could have pushed for adoption of strong rules surrounding conflict of interest.
Another model would be a veto jury empowered to nix tentatively approved grants that seemed poorly justified, too risky for expected benefit, or otherwise problematic. Even if the veto jury had missed the whole neo-nazi business, it very likely would have thrown this grant out for being terrible for other reasons.
I think that would hugely slow down the process.
I don’t think that is necessarily correct. We know the FLI grant was “approved” by September 7 and walked back sometime in November, so there was time for a veto-jury process without delaying FLI’s normal business process.
I am ordinarily envisioning a fairly limited role for a veto jury—the review would basically be for what US lawyers call “abuse of discretion” and might be called in less technical jargon a somewhat enhanced sanity check. Cf. Harman v. Apfel, 211 F.3d 1172, 1175 (9th Cir. 2000) (noting reversal under abuse of discretion standard is possible only “when the appellate court is convinced firmly that the reviewed decision lies beyond the pale of reasonable justification under the circumstances”).
It would not be an opportunity for the jury to comprehensively re-balance the case for and against funding the grant, or merely substitute its judgment for that of the grantmaker. Perhaps there would be defined circumstances in which veto juries would work with a less deferential standard of review, but it ordinarily should not take much time to determine that there was no abuse of discretion.
Why’s that bad?
Is it really that hard to think of reasons why a faster process may be better, ceteris paribus, than a slower process?
The “ceteris paribus” is the key part here, and I think in real life fast processes for deciding on huge sums of money tend to do much worse than slower ones.
If someone says that A is worse than B because it has a certain property C, you shouldn’t ask “Why is C bad?” if you are not disputing the badness of C. It would be much clearer to say, “I agree C is bad, but A has other properties that make it better than B on balance.”
Re: FTX—I’m not sure what would have “fixed FTX”, but I did think there were decisions here that made an impact. I don’t think EA is to blame for the criminal activities, but we did platform SBF a lot (and listened to him talk about things he had no idea about), and sort of had a personality cult around him (like we still do aroung Will, for example). You can see this from comments of people saying how disappointed they were and how much they had looked up to him.
So there are some cultural changes that would’ve had a chance of making it better—like putting less emphasis on individuals and more on communities; like leaving hard domain-knowledge questions to experts; like being wary of crypto, as a technology that involves a lot of immoral actions in practice from all big actors.
Re: Bostrom—Part of the reason that EA was so susceptible to backlash from his writings was that the movement made him a central figure despite his already fringe and alarming views, which IMO should have been a red flag even before we knew about the outright racism. And this, I think, is a result of lack of diversity in the movement, and even more so in decisionmaking circles.