I still think you’re overstating the extent to which the things described aren’t about institutional reform
I think they kind of are—but it should be understood that the purpose of the reforms is more about improving legibility than about fundamentally reducing the risk of similar occurrences
So, I feel like you are using “improving legibility” as kind of a euphemism here. I don’t currently believe that you think that the changes that CEA has made would make it easier for you or me or the external world to discover whether something like FTX was likely to happen again. At least I can’t think of any changes that would help with that.
To the contrary, my sense is the changes that have happened are much more directly optimizing for “making people believe that something like FTX won’t happen again”, but not in a way that is actually sensitive to the underlying facts. Like, I don’t think CEA has made it easier for external parties to judge whether EA will have good or bad consequences for the world.
I think in order to use the words “improving legibility” here, the key thing that needs to be the case is that CEA is doing things that would cause people to believe that future things like FTX are likely to happen, if they are likely to happen, and to cause people to believe that future things like FTX are unlikely to happen, if they are indeed are unlikely to happen. I don’t currently see such things, and indeed my guess is the public’s ability to judge what is going in inside of EA, and the consequences of our actions, has gotten worse, not better, as a result of recent changes.
I don’t think the legal investigation helps very much here, and I am not sure whether it overall would make people’s beliefs here more or less accurate (as I also think the similar OpenAI investigation has demonstrated) and the primary reason people do these kinds of legal investigations is to get a voice of authority on their side that makes people more hesitant to criticize them. I don’t think it’s completely useless, but it’s pretty close to it (especially when realizing how much OpenAI seems to have been able to manipulate a similar investigation in whatever way they desired).
(Sorry you’re getting downvoted, this seems like a productive conversation to me.)
There’s a big question about who things are being legible to. I don’t think this would make things meaningfully more legible to you or people in your reference class of basically being high context on the org and having a lot of intersections in social circles.
I think there’s a large crowd who are closer to “the general public” who only know about things via stuff that’s written publicly, and will only ever read a small fraction of that. I think that these kinds of reform help to make legible to that crowd that the org isn’t on some broad index-of-corruption where there is a culture of doing things for your mates, following the path of convenience (especially if that means getting more money), etc. That’s not directly about “will there be another SBF?” (base rates are surely low), but pretty informative anyway for a bunch of questions about how excited people will be to interact with the org/community.
I think if EV had had a high corruption culture (even if they didn’t literally know about any crimes), it would have been much more costly to have an external investigation (and such an investigation would likely have shaken out differently). This doesn’t exclude the possibility that there would have been better/cheaper ways to signal this. And perhaps there are other effects I’m not tracking; I’d be interested if you wanted to elaborate on the view that it makes it harder for the public to tell what’s going on inside EA.
I think if EV had had a high corruption culture (even if they didn’t literally know about any crimes), it would have been much more costly to have an external investigation (and such an investigation would likely have shaken out differently).
This is an interesting argument, thanks for making it. Would you mind explaining in a bit more depth? I would have thought the structure of the investigation, especially including the fact that the results were not published, would limit the cost of the investigation. As far as I’m aware the only thing outside observers have to go on is this paragraph, which appears to be from this guy:
Finally, I want to add this point. Over the last year, my team and I had the chance to meet dozens of people affiliated with the various EV projects. We were consistently impressed by everyone’s genuine commitment and dedication to the EA mission, particularly during a very challenging year. I wish you all the best in your future endeavors.
Is there some significant difference you see between the EVF investigation and the OpenAI investigation?
In the world with high corruption, I’d have expected more individuals to be scared of what an investigation would turn up.
If the investigation found major issues which weren’t publicly reported on, it would increase the fragility of EV’s position, since anyone who had seen the investigation could become a whistleblower with a relatively incontrovertible case.
These don’t totally exclude “there’s a big collusion not to share the issues”, but “major internal collusion” without also “was colluding with SBF” is getting to be a narrower target, and I think it would still feel fragile (such that the colluders might rationally prefer to avoid an investigation).
(Not commenting on the comparison with the OpenAI investigation because I haven’t understood it thoroughly enough to feel I have good takes.)
Mintz found no evidence that anyone at EV (including employees, leaders of EV-sponsored projects, and trustees) was aware of the criminal fraud of which Sam Bankman-Fried has now been convicted.
While we are not publishing any additional details regarding the investigation because doing so could reveal information from people who have not consented to their confidences being publicized and could waive important legal privileges that we do not intend to waive, we recognize that knowledge of criminal activity isn’t the only concern. I plan to share other non-privileged information on lessons learned in the aftermath of FTX and encourage others to share their reflections as well.
The information released from the investigation in this statement is exceedingly narrow—we didn’t have actual knowledge of SBF’s criminal fraud. That’s an exceedingly low bar to meet! That EVF won’t be joining SBF on his Club Fed vacation is an awfully slender reed to talk about “justified trust.”
General Reaction to Scope of Disclosure (and Likelihood of Possible Inferences)
Is there anything else that has been “credibly signal[ed]” through the investigation process? (quoting Owen’s earlier comment). Although I am aware of the reasons to maintain legal privileges, the choice to do so does little to persuade that this is a world in which EVF has “nothing to hide compared to [one in which it] wanted to hide something” (quoting same comment). Indeed, worlds in which the investigation sponsor chooses to hide behind privilege as to all but one (or a few) of the report’s ultimate conclusions seem very likely to be worlds in which there the sponsor “wanted to hide something.” And the choice to exclude certain important questions from the investigation’s scope would update me toward the view that the investigation sponsor didn’t want to find out the answers or strongly suspected they would be unflattering.
If Mintz had concluded that there was no negligence or recklessness in relation to EVF’s interactions with SBF, I am guessing that conclusion would have been disclosed! Likewise, if I wanted to convey to the CC that I was a trustworthy public charity, I would release a whole lot more than “no actual knowledge of a massive fraud” if I could do so.
Reacting to the Positive View of the Investigation
In the world with high corruption, I’d have expected more individuals to be scared of what an investigation would turn up.
But fear based on what an investigation would turn up only makes sense if you were concerned that the negative results would get out. So I think this is only correct insofar as those conducting the investigation were confident that it would find no actual knowledge of the criminal fraud.
If the investigation found major issues which weren’t publicly reported on, it would increase the fragility of EV’s position, since anyone who had seen the investigation could become a whistleblower with a relatively incontrovertible case.
But the number of people who would see the investigation report could be extremely limited. Presumably access to the report itself is extremely limited, given the rationale about not waiving privileges and protecting sources who spoke to the investigative firm. Outside counsel can’t leak without counsel surrendering his law license, incurring massive liability, and watching his business crumble. It’s plausible to me that no more than 10-15 people at EVF have seen the full report, and that these people are almost all insiders who are rather unlikely to leak.
How The Investigation Could Have Actually Rebuilt Lost Trust and Confidence
There was a more transparent / credible way to do this. EVF could have released, in advance, an appropriate range of specific questions upon which the external investigator was being asked to make findings of fact—as well as a set of possible responses (on a scale of “investigation rules this out with very high confidence” to “investigation shows this is almost certain”). For example—and these would probably have several subquestions each—one could announce in advance that the following questions were in scope and that the investigator had committed to providing specific answers:
Did anyone associated with EVF ever raise concerns about SBF being engaged in fraudulent activity? Did they ever receive any such concerns?
Did anyone associated with EVF discourage, threaten, or seek to silence any person who had concerns about illegal, unethical, or fraudulent conduct by SBF? (cf. the “Will basically threatened Tara” report).
When viewed against the generally-accepted norms for donor vetting in nonprofits, was anyone associated with EVF negligent, grossly negligent, or reckless in evaluating SBF’s suitability as a donor, failing to raise concerns about his suitability, or choosing not to conduct further investigation?
That kind of pre-commitment would have updated my faith in the process, and my confidence that the investigation reached all important topics. If EVF chose not to release the answers to those questions, it would have known that we could easily draw the appropriate inferences. Under those circumstances—but not the actual circumstances—I would view willingness to investigate as a valuable signal.
Thanks, this makes sense. I hadn’t thought about the possibility of committing to have the external investigation provide public answers on specific questions. (And the fact that I hadn’t thought of it gives me something to reflect on about how to ideally act during crises.)
Yeah, I think you are pointing towards something real here.
Like, I do think a thing that drove my reaction to this was a perspective in which it was obvious that most people in EA didn’t literally actively participate in the FTX fraud. I have encountered very extreme and obviously wrong opinions about this in the public (the comment section of the WaPo article provides many examples of this), and there is some value in engaging with that.
But I do think that is engaging with a position that is extremely shallow, and the mechanism of it seems like its trying to attack shallowness with more shallowness, which I don’t think is necessarily a strategy to rule out, but one that I am glad EA has avoided so far.
The way I would like Zack to have done it would have been to make a well-scoped statement that is honest and says “we might still be substantially responsible for SBF, but if so, we were because of indirect effects, not because we directly helped with the fraud. We really didn’t know how bad the fraud was, and we did not directly participate in it, and here is an investigation that shows that”
I’d be interested if you wanted to elaborate on the view that it makes it harder for the public to tell what’s going on inside EA.
I mean, as I mentioned as a concrete example, Will was told by CEA board members and leadership to not write anything detailed publicly about his FTX experiences due to the ongoing investigation. This seems like it has been extremely costly in terms of my ability and others ability to understand what happened, and I don’t see the corresponding payoff in the investigation results. To quote from Will 9 months ago:
The independent investigation commissioned by EV is still ongoing, and the firm running it strongly preferred me not to publish posts on backwards-looking topics around FTX while the investigation is still in-progress. I don’t know when it’ll be finished, or what the situation will be like for communicating on these topics even after it’s done.
Will publishing backwards-looking posts on FTX seems like the obvious and most central thing that should happen as part of FTX reform. We have clear documentation that the investigation prevented that from happening (and as far as I can tell the reflections are still not published, so it seems like the investigations completely prevented this kind of information from entering the public record).
So, I feel like you are using “improving legibility” as kind of a euphemism here. I don’t currently believe that you think that the changes that CEA has made would make it easier for you or me or the external world to discover whether something like FTX was likely to happen again. At least I can’t think of any changes that would help with that.
To the contrary, my sense is the changes that have happened are much more directly optimizing for “making people believe that something like FTX won’t happen again”, but not in a way that is actually sensitive to the underlying facts. Like, I don’t think CEA has made it easier for external parties to judge whether EA will have good or bad consequences for the world.
I think in order to use the words “improving legibility” here, the key thing that needs to be the case is that CEA is doing things that would cause people to believe that future things like FTX are likely to happen, if they are likely to happen, and to cause people to believe that future things like FTX are unlikely to happen, if they are indeed are unlikely to happen. I don’t currently see such things, and indeed my guess is the public’s ability to judge what is going in inside of EA, and the consequences of our actions, has gotten worse, not better, as a result of recent changes.
I don’t think the legal investigation helps very much here, and I am not sure whether it overall would make people’s beliefs here more or less accurate (as I also think the similar OpenAI investigation has demonstrated) and the primary reason people do these kinds of legal investigations is to get a voice of authority on their side that makes people more hesitant to criticize them. I don’t think it’s completely useless, but it’s pretty close to it (especially when realizing how much OpenAI seems to have been able to manipulate a similar investigation in whatever way they desired).
(Sorry you’re getting downvoted, this seems like a productive conversation to me.)
There’s a big question about who things are being legible to. I don’t think this would make things meaningfully more legible to you or people in your reference class of basically being high context on the org and having a lot of intersections in social circles.
I think there’s a large crowd who are closer to “the general public” who only know about things via stuff that’s written publicly, and will only ever read a small fraction of that. I think that these kinds of reform help to make legible to that crowd that the org isn’t on some broad index-of-corruption where there is a culture of doing things for your mates, following the path of convenience (especially if that means getting more money), etc. That’s not directly about “will there be another SBF?” (base rates are surely low), but pretty informative anyway for a bunch of questions about how excited people will be to interact with the org/community.
I think if EV had had a high corruption culture (even if they didn’t literally know about any crimes), it would have been much more costly to have an external investigation (and such an investigation would likely have shaken out differently). This doesn’t exclude the possibility that there would have been better/cheaper ways to signal this. And perhaps there are other effects I’m not tracking; I’d be interested if you wanted to elaborate on the view that it makes it harder for the public to tell what’s going on inside EA.
This is an interesting argument, thanks for making it. Would you mind explaining in a bit more depth? I would have thought the structure of the investigation, especially including the fact that the results were not published, would limit the cost of the investigation. As far as I’m aware the only thing outside observers have to go on is this paragraph, which appears to be from this guy:
Is there some significant difference you see between the EVF investigation and the OpenAI investigation?
In the world with high corruption, I’d have expected more individuals to be scared of what an investigation would turn up.
If the investigation found major issues which weren’t publicly reported on, it would increase the fragility of EV’s position, since anyone who had seen the investigation could become a whistleblower with a relatively incontrovertible case.
These don’t totally exclude “there’s a big collusion not to share the issues”, but “major internal collusion” without also “was colluding with SBF” is getting to be a narrower target, and I think it would still feel fragile (such that the colluders might rationally prefer to avoid an investigation).
(Not commenting on the comparison with the OpenAI investigation because I haven’t understood it thoroughly enough to feel I have good takes.)
I have the opposite reaction.
Background
To recap, Zach wrote about four months ago:
The information released from the investigation in this statement is exceedingly narrow—we didn’t have actual knowledge of SBF’s criminal fraud. That’s an exceedingly low bar to meet! That EVF won’t be joining SBF on his Club Fed vacation is an awfully slender reed to talk about “justified trust.”
General Reaction to Scope of Disclosure (and Likelihood of Possible Inferences)
Is there anything else that has been “credibly signal[ed]” through the investigation process? (quoting Owen’s earlier comment). Although I am aware of the reasons to maintain legal privileges, the choice to do so does little to persuade that this is a world in which EVF has “nothing to hide compared to [one in which it] wanted to hide something” (quoting same comment). Indeed, worlds in which the investigation sponsor chooses to hide behind privilege as to all but one (or a few) of the report’s ultimate conclusions seem very likely to be worlds in which there the sponsor “wanted to hide something.” And the choice to exclude certain important questions from the investigation’s scope would update me toward the view that the investigation sponsor didn’t want to find out the answers or strongly suspected they would be unflattering.
If Mintz had concluded that there was no negligence or recklessness in relation to EVF’s interactions with SBF, I am guessing that conclusion would have been disclosed! Likewise, if I wanted to convey to the CC that I was a trustworthy public charity, I would release a whole lot more than “no actual knowledge of a massive fraud” if I could do so.
Reacting to the Positive View of the Investigation
But fear based on what an investigation would turn up only makes sense if you were concerned that the negative results would get out. So I think this is only correct insofar as those conducting the investigation were confident that it would find no actual knowledge of the criminal fraud.
But the number of people who would see the investigation report could be extremely limited. Presumably access to the report itself is extremely limited, given the rationale about not waiving privileges and protecting sources who spoke to the investigative firm. Outside counsel can’t leak without counsel surrendering his law license, incurring massive liability, and watching his business crumble. It’s plausible to me that no more than 10-15 people at EVF have seen the full report, and that these people are almost all insiders who are rather unlikely to leak.
How The Investigation Could Have Actually Rebuilt Lost Trust and Confidence
There was a more transparent / credible way to do this. EVF could have released, in advance, an appropriate range of specific questions upon which the external investigator was being asked to make findings of fact—as well as a set of possible responses (on a scale of “investigation rules this out with very high confidence” to “investigation shows this is almost certain”). For example—and these would probably have several subquestions each—one could announce in advance that the following questions were in scope and that the investigator had committed to providing specific answers:
Did anyone associated with EVF ever raise concerns about SBF being engaged in fraudulent activity? Did they ever receive any such concerns?
Did anyone associated with EVF discourage, threaten, or seek to silence any person who had concerns about illegal, unethical, or fraudulent conduct by SBF? (cf. the “Will basically threatened Tara” report).
When viewed against the generally-accepted norms for donor vetting in nonprofits, was anyone associated with EVF negligent, grossly negligent, or reckless in evaluating SBF’s suitability as a donor, failing to raise concerns about his suitability, or choosing not to conduct further investigation?
That kind of pre-commitment would have updated my faith in the process, and my confidence that the investigation reached all important topics. If EVF chose not to release the answers to those questions, it would have known that we could easily draw the appropriate inferences. Under those circumstances—but not the actual circumstances—I would view willingness to investigate as a valuable signal.
Thanks, this makes sense. I hadn’t thought about the possibility of committing to have the external investigation provide public answers on specific questions. (And the fact that I hadn’t thought of it gives me something to reflect on about how to ideally act during crises.)
Yeah, I think you are pointing towards something real here.
Like, I do think a thing that drove my reaction to this was a perspective in which it was obvious that most people in EA didn’t literally actively participate in the FTX fraud. I have encountered very extreme and obviously wrong opinions about this in the public (the comment section of the WaPo article provides many examples of this), and there is some value in engaging with that.
But I do think that is engaging with a position that is extremely shallow, and the mechanism of it seems like its trying to attack shallowness with more shallowness, which I don’t think is necessarily a strategy to rule out, but one that I am glad EA has avoided so far.
The way I would like Zack to have done it would have been to make a well-scoped statement that is honest and says “we might still be substantially responsible for SBF, but if so, we were because of indirect effects, not because we directly helped with the fraud. We really didn’t know how bad the fraud was, and we did not directly participate in it, and here is an investigation that shows that”
I mean, as I mentioned as a concrete example, Will was told by CEA board members and leadership to not write anything detailed publicly about his FTX experiences due to the ongoing investigation. This seems like it has been extremely costly in terms of my ability and others ability to understand what happened, and I don’t see the corresponding payoff in the investigation results. To quote from Will 9 months ago:
Will publishing backwards-looking posts on FTX seems like the obvious and most central thing that should happen as part of FTX reform. We have clear documentation that the investigation prevented that from happening (and as far as I can tell the reflections are still not published, so it seems like the investigations completely prevented this kind of information from entering the public record).