I think it could go either way. If you think the limiting factor on auditing is the number of people you have to audit, then having the same N people be the board of trustees of all EA charities is maximally efficient. If you think it’s the number of connections between sources of potential CoI, then having completely nonoverlapping groups seem optimal.
I guess I’m imagining from either Open Phil’s perspective, or that of other large funders, the risk of value misalignment or incompetence of Open Phil staff is already priced in, and they’ve already paid the cost of evaluating Claire.
It’s hard to imagine that (purely from the perspective of reducing costs of auditing)Holden or Cari or Dustin preferring an unknown quantity to Claire. There might be other good reasons to prefer having a more decentralized board[1], but this particular reason seems wrong.
Likewise, from the perspective of future employees or donors to EVF, the risk of value misalignment or incompetence of EVF’s largest donor is already a cost they necessarily have to pay if they want to work for or fund EVF. So adding a board member (and another source of COI) that’s not associated with Open Phil can only increase the number of COIs, not decrease it.
for example, a) you want a diversity of perspectives, b) you want to reduce the risks of being beholden to specific entities c) you want to increase the number of potential whistleblowers
Nonoverlapping doesn’t mean no COI. Imagine Org1 grants to Org2, which grants to many orgs:
Org1 board: A, B, C
Org2 board: D, E, F
Grantees: lots of people
Imagine there are COIs between each person with probably X.
Each grantee org has a 3X chance of COI with Org2, and 3X with Org1. You care more about the ones with Org2, but Org1 matters too.
Now imagine C and F are the same person. The chance of COI is unchanged: 3X with each org.
And if you’re worried about auditing you have 6 connections to audit per grantee in the first case, and five in the second. It’s true that auditing a single C=F board member is more work then auditing just C or F, but it’s probably less work than auditing both C and F. (And there definitely isn’t exponentially more total work.)
Sure, that’s one way of modelling it. Another is something like COIs between entities in which you’re invested. That means a trustee of a single org has one possible COI vector—looking after themselves when they should be looking after their org. Two trustees of one org each have two vectors between them. One trustee of two orgs has four vectors—looking after themselves when they should be looking after either org, or looking after either org at the expense of the other.
Generalising, it seems like on this way of thinking you’d get (n^2 - n) more COI vectors, where n is number of affiliations, than you would with single-affiliation trustees doing the same set of roles. I have the sense that senior EAs can be affiliated with maybe as many as 4 or 5 legally distinct organisations—and if you count the subsidiaries of EVF as effectively different organisations with shared trustees, those numbers get pretty big.
I think this way of thinking about it more congruent to the concerns many (including me) had around eg EVF buying Wytham Abbey, or grantmakers from one nonprofit giving money to another with which they’re closely involved.
I see, you are thinking about conflicts between the interest of being “person X in their role at org Y” vs “person X in their role at org Z”. I agree that this goes up quadratically as people have roles at more organizations: in the extreme where everyone had at most one organizational role you couldn’t have any of that kind of CoI.
I had thought, though, we were talking about interpersonal conflicts of interest. Things like “person X in their role at org Y” vs “person X in their role as someone dating person W at org Z”. This is the kind of CoI that I think doesn’t increase if you have fewer people across the same number of organization roles.
I think this way of thinking about it more congruent to the concerns many (including me) had around eg EVF buying Wytham Abbey
I thought the issue with the purchase was people thinking it was not cost effective, or should have been better disclosed? What’s the CoI argument here?
Let’s distinguish some legal conflict of interest moral conflict of interest
I think that what I described was not a legal conflict of interest, which would be where a senior staff member is faced with a decision that could benefit them.[1] IMO EAs should consider it necessary but not sufficient to avoid these.
I would call a moral conflict of interest any situation in which a senior member is faced with a decision where they have an incentive to do something that isn’t in the best interest of the organisation for which they work. IMO these should also be avoided, or at least recognised as a serious cost to weigh against the upsides. Such situations put the staff in situations where it’s extremely difficult to recognise from the outside whether the decision process was really for the greater good, or a rationalisation. They seem at high risk of biasing support strongly towards the pet projects of the decisionmakers at the expense of many smaller projects that continually struggle for funding.
Wytham Abbey seems like an example of this, if as I understand members of the EVF board were involved in the decision making on OP’s behalf. Asterisk is perhaps a similar example. If anyone else in the community had suggested ‘an EA magazine sponsored webzine’ I suspect it would have really struggled to get traction.
faced with a decision where they have an incentive to do something that isn’t in the best interest of the organisation for which they work.
That’s already how I was using the term, fwiw.
Wytham Abbey seems like an example of this, if as I understand members of the EVF board were involved in the decision making on OP’s behalf.
It sounds like maybe you’re saying you think there’s fundamentally a conflict of interest in having someone from OP on the EVF board, because OP is the
funds a lot of things at EVF and what’s best for EVF could be different from what OP wants?
This misses why OP would want to have one of their employees on the EVF board: they are providing so much funding to EVF I expect they want someone to formally represent them internally. It would be bad for OP if they granted money to EVF which then went to do things that OP didn’t want to be supporting, and having board representation is a way to make that less likely.
It sounds like maybe you’re saying you think there’s fundamentally a conflict of interest in having someone from OP on the EVF board, because OP is the funds a lot of things at EVF and what’s best for EVF could be different from what OP wants?
Yeah, that sounds right. When you’re a senior staff member at an organisation you are, in some sense, supposed to be optimising for the health of that org. On this understanding you’re the senior staff member at multiple orgs you’re then supposed to be optimising for multiple variables, which isn’t logically possible.
Obviously you can optimise for some function of the two, but this involves a lot more subjective judgement, so a) could more easily go wrong without any bad faith, and b) it can obfuscate genuinely bad faith decisions (such as prioritising support for an org with which you’re associated because it gives you greater social status) - even to the people making them.
I don’t assert (or think) anyone at EVF is acting in seriously bad faith, but as I’ve said elsewhere, I think we should assume a) non-0 probability that they sometimes do so in minor ways and b) that if they continue to stay in a position that incentivises them to do so the risk will continue to increase.
This misses why OP would want to have one of their employees on the EVF board: they are providing so much funding to EVF I expect they want someone to formally represent them internally.
It doesn’t ‘miss’ it. I understand there are other considerations—they’re discussed everywhere; I’m just stating that there exists this downside which isn’t discussed nearly as much, and doesn’t seem to be acknowledged by the people it applies to.
I guess I could see it that way, but this is a pretty non-central CoI. My understanding is EVF took on an OP staff member as a board member because both EVF and OP wanted to be more closely aligned. They understand that the orgs have somewhat different interests, and by putting someone in a position to embody both they’re pulling the two orgs closer together.
Your argument here cuts against your prior comment.
I think it could go either way. If you think the limiting factor on auditing is the number of people you have to audit, then having the same N people be the board of trustees of all EA charities is maximally efficient. If you think it’s the number of connections between sources of potential CoI, then having completely nonoverlapping groups seem optimal.
I guess I’m imagining from either Open Phil’s perspective, or that of other large funders, the risk of value misalignment or incompetence of Open Phil staff is already priced in, and they’ve already paid the cost of evaluating Claire.
It’s hard to imagine that (purely from the perspective of reducing costs of auditing)Holden or Cari or Dustin preferring an unknown quantity to Claire. There might be other good reasons to prefer having a more decentralized board[1], but this particular reason seems wrong.
Likewise, from the perspective of future employees or donors to EVF, the risk of value misalignment or incompetence of EVF’s largest donor is already a cost they necessarily have to pay if they want to work for or fund EVF. So adding a board member (and another source of COI) that’s not associated with Open Phil can only increase the number of COIs, not decrease it.
for example, a) you want a diversity of perspectives, b) you want to reduce the risks of being beholden to specific entities c) you want to increase the number of potential whistleblowers
Nonoverlapping doesn’t mean no COI. Imagine Org1 grants to Org2, which grants to many orgs:
Org1 board: A, B, C
Org2 board: D, E, F
Grantees: lots of people
Imagine there are COIs between each person with probably X.
Each grantee org has a 3X chance of COI with Org2, and 3X with Org1. You care more about the ones with Org2, but Org1 matters too.
Now imagine C and F are the same person. The chance of COI is unchanged: 3X with each org.
And if you’re worried about auditing you have 6 connections to audit per grantee in the first case, and five in the second. It’s true that auditing a single C=F board member is more work then auditing just C or F, but it’s probably less work than auditing both C and F. (And there definitely isn’t exponentially more total work.)
Sure, that’s one way of modelling it. Another is something like COIs between entities in which you’re invested. That means a trustee of a single org has one possible COI vector—looking after themselves when they should be looking after their org. Two trustees of one org each have two vectors between them. One trustee of two orgs has four vectors—looking after themselves when they should be looking after either org, or looking after either org at the expense of the other.
Generalising, it seems like on this way of thinking you’d get (n^2 - n) more COI vectors, where n is number of affiliations, than you would with single-affiliation trustees doing the same set of roles. I have the sense that senior EAs can be affiliated with maybe as many as 4 or 5 legally distinct organisations—and if you count the subsidiaries of EVF as effectively different organisations with shared trustees, those numbers get pretty big.
I think this way of thinking about it more congruent to the concerns many (including me) had around eg EVF buying Wytham Abbey, or grantmakers from one nonprofit giving money to another with which they’re closely involved.
I see, you are thinking about conflicts between the interest of being “person X in their role at org Y” vs “person X in their role at org Z”. I agree that this goes up quadratically as people have roles at more organizations: in the extreme where everyone had at most one organizational role you couldn’t have any of that kind of CoI.
I had thought, though, we were talking about interpersonal conflicts of interest. Things like “person X in their role at org Y” vs “person X in their role as someone dating person W at org Z”. This is the kind of CoI that I think doesn’t increase if you have fewer people across the same number of organization roles.
I thought the issue with the purchase was people thinking it was not cost effective, or should have been better disclosed? What’s the CoI argument here?
Let’s distinguish some legal conflict of interest moral conflict of interest
I think that what I described was not a legal conflict of interest, which would be where a senior staff member is faced with a decision that could benefit them.[1] IMO EAs should consider it necessary but not sufficient to avoid these.
I would call a moral conflict of interest any situation in which a senior member is faced with a decision where they have an incentive to do something that isn’t in the best interest of the organisation for which they work. IMO these should also be avoided, or at least recognised as a serious cost to weigh against the upsides. Such situations put the staff in situations where it’s extremely difficult to recognise from the outside whether the decision process was really for the greater good, or a rationalisation. They seem at high risk of biasing support strongly towards the pet projects of the decisionmakers at the expense of many smaller projects that continually struggle for funding.
Wytham Abbey seems like an example of this, if as I understand members of the EVF board were involved in the decision making on OP’s behalf. Asterisk is perhaps a similar example. If anyone else in the community had suggested ‘an EA magazine sponsored webzine’ I suspect it would have really struggled to get traction.
Possibly this is a ‘conflict of loyalty’? I haven’t found much around the legal term.
That’s already how I was using the term, fwiw.
It sounds like maybe you’re saying you think there’s fundamentally a conflict of interest in having someone from OP on the EVF board, because OP is the funds a lot of things at EVF and what’s best for EVF could be different from what OP wants?
This misses why OP would want to have one of their employees on the EVF board: they are providing so much funding to EVF I expect they want someone to formally represent them internally. It would be bad for OP if they granted money to EVF which then went to do things that OP didn’t want to be supporting, and having board representation is a way to make that less likely.
Yeah, that sounds right. When you’re a senior staff member at an organisation you are, in some sense, supposed to be optimising for the health of that org. On this understanding you’re the senior staff member at multiple orgs you’re then supposed to be optimising for multiple variables, which isn’t logically possible.
Obviously you can optimise for some function of the two, but this involves a lot more subjective judgement, so a) could more easily go wrong without any bad faith, and b) it can obfuscate genuinely bad faith decisions (such as prioritising support for an org with which you’re associated because it gives you greater social status) - even to the people making them.
I don’t assert (or think) anyone at EVF is acting in seriously bad faith, but as I’ve said elsewhere, I think we should assume a) non-0 probability that they sometimes do so in minor ways and b) that if they continue to stay in a position that incentivises them to do so the risk will continue to increase.
It doesn’t ‘miss’ it. I understand there are other considerations—they’re discussed everywhere; I’m just stating that there exists this downside which isn’t discussed nearly as much, and doesn’t seem to be acknowledged by the people it applies to.
I guess I could see it that way, but this is a pretty non-central CoI. My understanding is EVF took on an OP staff member as a board member because both EVF and OP wanted to be more closely aligned. They understand that the orgs have somewhat different interests, and by putting someone in a position to embody both they’re pulling the two orgs closer together.