Summary
- I think EA is under centralised
- There are few ‘large’ EA organisations but most EA opportunities are 1-2 person projects
- This is setting up most projects to fail without proper organisational support and does not provide good incentives for experienced professionals to work on EA projects
- EA organisations with good operations could incubate smaller projects before spinning them out
Levels of Centralisation
We could imagine different levels of centralisation for a movement ranging from fully decentralised to fully centralised.
- Fully decentralised, everyone works on their own project, no organisations bigger than 1 person
- Fully centralised, everyone works inside the same organisation (e.g. the civil service)
It seems that EA tends more towards the decentralised model, there are relatively few larger organisations with ~50 or more people (Open Phil, GiveWell, Rethink Priorities, EVF), there are some with ~5-20 people and a lot of 1-2 person projects.
I think EA would be much worse if it was one large organisation but there is probably a better balance found between the two extremes then we have at the moment.
I think being overly decentralised may be setting up most people to fail.
Why would being overly decentralised be setting people up to fail?
- Being an independent researcher/organiser is harder without support systems in place, and trying to coordinate this outside of an organisation is more complicated
- These support systems include
- Having a manager
- Having colleagues to bounce ideas off/moral support
- Having professional HR/operations support
- Health insurance
- Being an employee rather than a contractor/grant recipient that has to worry about receiving future funding (although there are similar concerns about being fired)
- When people are setting up their own projects it can take up a large proportion of their time in the first year just doing operations to run that project, unrelated to the actual work they want to do. This can include spending a lot of the first year just fundraising for the second year
How a lack of centralisation might affect EA overall
- Being a movement with lots of small project work will appeal more to those with a higher risk tolerance, potentially pushing away more experienced people who would want to work on these projects, but within a larger organisation
- Having a lot of small organisations will lead to a lot of duplication of operation/administration work
- It will be harder to have good governance for lots of smaller organisations, some choose to not have any governance structures at all unless they grow
- There is less competition for employees if the choice is between 3 or 4 operationally strong organisations or being in a small org
What can change?
- Organisations with good operations and governance could support more projects internally - One example of this already is the Rethink Priorities Special Projects Program
- These projects can be supported until they have enough experience and internal operations to survive and thrive independently
- Programs that are mainly around giving money to individuals could be converted into internal programs, something more similar to the Research Scholars Program, or Charity Entrepreneurship’s Incubation Program
Mergers and Acquisitions seems underexplored in EA.
I agree. I've explored some of this. I think it definitely could make sense in a lot of cases.
I think a lot of the stuff Deena touches on in "3 Basic Steps to Reduce Personal Liability as an Org Leader" are important here too. I think de-centralization has lead to a lot of people doing work independently and then being really under-resourced to handle the pressures of that (and this went 100x in particular response to FTX). I think grantees-as-individuals need to be very careful about not co-mingling funds, making sure taxes are in order, etc. and our current community plan of having a lot of individual grantees may involve getting people to be taking on a lot of legal risk that bigger organizations are in a better position to handle.
I think especially during the FTX era but still now, I have been a bit surprised to see a bias towards wanting to fund many smaller orgs rather than one bigger org. FTX had an explicit preference for newer and less established orgs / individuals and I think that clearly backfired. Some of this makes sense as you want to avoid having "all of your eggs in one basket" / "hedge your bets" but I think big orgs have a lot of great advantages that are underrated by EA funders and others.
~
Disclaimer: Obviously I would say this though given that I run a "big org" (Rethink Priorities). I'm speaking for just myself personally here though, not some RP position (there's a lot of diversity of perspective at RP). Also I am complaining about "funders" here but I am on the EA Infrastructure Fund so maybe I'm part of the problem?
Meta: Just wanted to quickly say that I really appreciate how you communicate on the Forum, Peter. You're thoughtful, ask productive questions and lay out biases which I find very helpful. Thank you for taking the time to engage with the content on the Forum, and sharing your thoughts.
Something worth clarifying:
I agree that this is super confusing. However, I do think that some claims about EA being too centralised have been about there being too few big orgs. I think that's just not true, and I think the confusion between these two points has probably made it difficult to discuss in the past.
All of which is to say: we should probably prefer to use more specific words than just "centralisation" unqualified.
It seems like the main point of the original post and the comments are about how more centralization is helpful. For balance, I want to argue against myself and while I think there are clear benefits to net centralization, there are also some reasons/ways net centralization may be harmful:
You are consolidating legal risk into fewer entities, meaning that one high-level mistake can take a lot of things down (in the wake of FTX this seems extra important... I think this is the biggest drawback to more centralization)
Mainly due to the above but also other factors, larger organizations are much more risk averse and can just do fewer things
Smaller orgs/individuals who don't have to care about their reputation as much can take bolder risks (this is both good and bad)
Smaller organizations are quicker to act, require less stakeholder sign-off to get things done (this is both good and bad)
Smaller organizations I guess on the margin are more able to shut down with fewer politics if things don't work out rather than continuing to do something not effective (but on the other hand maybe shutting down more clearly means you are fired and don't get money whereas in a bigger org maybe you can be moved to a new project?)
I want to illustrate the “larger organizations are much more risk averse” point. When I worked at Rethink Priorities, I felt less freedom to publicly share unpolished and controversial thoughts because that could hurt Rethink Priorities reputation. And the bigger Rethink Priorities grew, the more there was to lose, the more this became a problem. Because of this, articles that I didn’t think were very promising but worth publishing (e.g., aquatic noise) took more time to finish as every claim went through more scrutiny than it would be optimal if I was an independent researcher. Also, I never publicly shared my preliminary findings on things that I didn’t think were promising because it would’ve taken too much time. Finally, it was more difficult to express some bold opinions like Wild Animal Welfare is not very promising because it could hurt Rethink Priorities' Wild Animal Welfare funding and relationships with some other organizations.
Not just high-level mistakes, either. Imagine you are a religious movement with a dozen congregations, each of which has considerable but not total autonomy. It's totally plausible that one rogue clergyperson (employee) or congregation (project) could incur enough liability to overwhelm a single corporation's insurance and consume the assets of the other 11 congregations and the central office.
There are ways to mitigate that risk -- for instance, if RP Special Projects grew and/or took on riskier projects, I would at least consider spinning it off into a wholly owned subsidiary of RP. If done correctly (which can be a bit of a pain as I understand it), that should reduce main RP's exposure to any meltdowns coming out of Special Projects. In many jurisdictions, you can also get substantial protection for donor-restricted funds if you keep them separate and manage your accounting well. So that could help manage certain types of contagion risk within a sponsoring organization like the hypothetical spun-off Special Projects itself.
Of course, I can't give anyone legal advice but am merely observing that large-org legal risk can be mitigated to some degree if the payoff from being with a large org is big enough.
Yeah I think this is a really plausible risk to centralization.
~
This is definitely something we are considering doing.
On the first point, if FTX had happened and there were more large EA organisations, it would have been easier to handle the fall out from that, with more places for smaller organisations and individuals to go to for support.
On the last point it seems like that was a part of why DARPA had success, they had lots of projects and were focused on the best succeeding rather than maintaining failing ideas.
I'd argue that the universe in which FTX exposure was at its lowest is the world in which all grantees were small . . . but were properly incorporated and managed like corporations (no comingling of money, observance of corporate formalities, etc.)
It would be easier to manage a scenario with few big entities, but it'd be a lot easier for John Ray & Co. to manage too -- especially because those orgs would be more likely to have non-FTX-derived assets that could be taken to satisfy a judgment.
With proper corporate status, the individuals should have been safe, as a corp-to-its-individual employee transfer seems to be beyond the reach of section 550's power over subsequent transferees. And the individual grantee corporations would make relatively unattractive targets for several reasons (mainly having little money to net per lawsuit).
In other words, I think the core defect was not structuring grantees appropriately rather than their size per se.
Right, I totally agree with that. I think the way FTX Future Fund pushed their risk onto individuals who were not well equipped to understand and/or handle that was a huge failing and is very sad.
~
Agreed. That's part of what I'm trying to do with Rethink Priorities. But I think there are many ways in which we fail to live up to that, and my guess is that other big orgs maintain cost-ineffective projects due to inertia or political pressure because it's really hard to be ruthless about these sorts of things.
I strongly agree that as we mature we should push towards a more centralised model. Two specific reasons which expand on your post:
There’s a point for funders here. If you’re minded to fund certain work in a certain cause area – you might do more good by approaching established organisations that already have a “back end” and offering to fund them to find more talent and expand in certain directions (rather than giving a new grant and spinning off yet another person or small organisation).
Other comments joke about mergers and acquisitions. But in seriousness, my vote on boards of the EA-related organisations I’m on has been, and will continue to be, in favour of sensible centralisation.
Thanks for posting this! I agree, and one thing I've noticed while community building is that it's very easy to give career direction to students and very early-career professionals, but much more challenging to mid/late-career professionals. Early-career people seem more willing to experiment/try out a project that doesn't have great support systems, whereas mid/late-career people have much more specific ideas about what they want out of a job.
Entrepreneurship is not for everyone, and being advised to start your own project with unclear parameters and outcomes often has low appeal to people who have been working for 10+ years in professions with meaningful structure, support, and reliable pay. (It often has low appeal to students/early-career professionals too, but younger people seem more willing to try.) I would love to see EA orgs implement some of the suggestions you mentioned.
Hey David, love you, but as one of the people who has struggled as an independent EA worker and in small orgs I very strongly disagree. The overarching theme of my disagreement is that EA desperately needs more competition in general and more more incentives for its staff to do a good job.
All of the downsides you mention could be addressed by (hopefully multiple) support organisations that would prioritise according to their sense of counterfactual value. Such orgs could directly compete or lightly compete, by eg having distinct focus areas, but that one could turn to if they felt underserved by the one nominally in their domain. This would give a recourse for people who felt like a particular org wasn't doing a good job to turn to an alternative one, and for effectiveness-minded funders to respond by raising the relative funding the more popular org received.
The downsides of centralisation are numerous: we've seen accusation after accusation of how EA people have run orgs very badly. One needn't believe all of these (and I'm certainly sceptical of some parts of them) to think they provide evidence if it were needed that more centralisation massively increases the chance that bad or incompetent actors end up steering the movement, or just that EA becomes/remains overly homogenous and cliquey, such that valuable projects are less likely to get funding because they're outside the funders' social circle.
I think EA would improve with more competition as well, what I'm suggesting is more competition in the 'larger' orgs category. If someone has a disagreement with how things are run at one of the bigger EA orgs, there are very few other places for them to go within EA.
I don't think that the issues you linked to are because of centralisation, there are lots of small organisations badly run, and without larger organisations there often isn't a way to hold bad actors accountable.
I'm talking about increasing the number of large organisations. I don't think I can do much about how many different types of funders we have, which is a separate question.
I'm confused what you are suggesting exactly. When reading post I assumed that you suggested more centralisation in general. If there where a compotator to CEA, I would not call that "more centralisation". Although maybe it depends on how we get there form here?
If several small orgs join together to form a new big org, that would seem like going towards more centralisation. But if someone starts a new org that grows into a large org, which competes with an existing large org, that would look like de-centralisation to me. Maybe (as someone pointed out in another comment) "centralisation" is not a useful word here, since it's meaning is too ambiguous.
I think it would be useful if you could clarify your suggestion.
I think it would be better to have 20 organisations with about 50 people each than 3 organisations with 50 people and then everyone else working as individuals. One organisation with 1000 people would probably be the worst option.
Thanks, that clarifies things.
I'm still not sure what you mean by org. Do you count CEA as an org, or EVF as an org?
I think in terms of projects and people and funding. Legal orgs are just another part of the infrastructure that supports funding and people.
I think it would be great if AI Safety Support, where given enough funding to hire 50 people, and used that funding to provide financial security to lots of existing projects. Although that is heavily biased by the fact that I personally know and trust the people running AISS, and that their work style and attitude is aligned with mine. I might feel very different about someone I did not trust as much getting that power.
In my example I was more referring to orgs like EVF, but I imagine if EA was more centralised there would be a range of larger orgs, some more like EVF and others more like Open Phil, who aren't incubating projects.
I still strongly disagree if, as it seems from the OP, you mean 'we want more larger orgs' rather than 'we want to make large orgs smaller'. The larger the org, the less fidelity you get from any competition.
If Org 1 and Org 2 each do only one activity, they can compete perfectly, in that funders can express an unambiguous preference between those two activities by donating a marginal $ to one or other of them.
But for each further activity an org does, ignoring restricted donations (which on my understanding are counterfactually not very effective), they approximately multiply the number of preference-orderings consistent with donating to either org by the total number of activities. If I'm thinking about this correctly, the number of preference orderings consistent with a donation to one of two orgs is (mn)! - m!n! where m and n are the number of activities each org does.
For example, if Org 1 does activities A, B, and and Org 2 does C and D, then giving to Org 1 is consistent with any combination of preferences such that at least one activity at Org 1 is preferred to at least one activity at Org 2. This gives 4! - 2!*2! = 20 consistent orderings:
A > B > C > D
A > B > D > C
A > C > B > D
A > C > D > B
A > D > B > C
A > D > C > B
B > A > C > D
B > A > D > C
B > C > A > D
B > C > D > A
B > D > A > C
B > D > C > A
C > B > A > D
C > B > D > A
C > A > B > D
C > A > D > B
C > D > B > AC > D > A > BD > B > C > A
D > B > A > C
D > C > B > AD > C > A > BD > A > B > C
D > A > C > B
Obviously this is a bit abstract/oversimplified, but if we think it's a reasonable approximation of reality, it implies that the difference in fidelity of competition between small and large organisations is colossal. Depending on how you think about CEA (the sub-org, not EVF) alone, their activities might include
Arguably software development and any other in-house services they provide. Even by comparison to a hypothetical org who was only doing one of those things, that would mean a donation to either org would be consistent with 8! preference orderings (or 9! if you include software): that's 40320 or 362880, respectively. Obviously take these numbers with a large grain of salt in the real world, but I think the picture isn't totally unreasonable, and the numbers are large enough that they could outweigh a very great number of inside view considerations, to make any organisation provided an alternative to have close to zero value in the sense of incentivising either org to perform better.
This is before you're even looking at the umbrella orgs, where the picture gets even murkier - in theory their sub-orgs seem to be expected to arrange their own funding independently, but in principle their sources of funding seem to be extremely correlated.
My thoughts here are something like:
I directionally agree with this, but think there's an important distinction to be made between various models. On the one hand, you have a fiscal sponsorship-like model with some degree of operations support. On the other hand, you have an organization like GiveWell, where all funding is (as far as I know) centralized.
The former model is likely going to be easier for existing organizations to offer, but it lacks some of the advantages you mention in the post. Particularly, most fiscal sponsors expect a project to bring in its own funding (plus often a percentage for full-organization overhead). So that type of arrangement isn't likely to address concerns about being a "contractor/grant recipient that has to worry about receiving future funding" or time spent fundraising very effectively. If the sponsoring organization is going to commit to paying the program staff whether or not the program can attract grants, that is a much higher level of commitment that will likely involve a significantly higher bar for acceptance.
I like that (at least my part of) EA is a network of small projects. It makes it easier to take initiative and get things done. Also, the there are more support structure that you seem to think.
For context: I'm an AI Safety organiser, and independent researcher. I mainly work for AI Safety Camp but I'm also involved to a lesser degree in ~3 other field building projects.
I have most of these, at least the ones I want.
There are many types of centralisation, and they don't necessarily have to go together. If you count the number or legal orgs in EA, there will be fewer than you would naively think, since a lot or orgs are sharing their legal and administrative back-end, while outward operating as separate project, with separate names, logos and webpages.
I'm against power centralisation, i.e. fewer bosses are telling more people what to do. But I'm in favour of legal/admin centralisation, i.e. fewer set of legal/administrative entities are serving more projects.
Grater power centralisation would lead to less duplication of projects, but it would also lead to much fewer things getting done in general. I don't know exactly how to argue for this, because this is mainly a empirical observation. But in my experience, a pear network, where everyone is fee and encourage to take initiative and start new projects with out having to ask their boss for permission, is just waaaaay more efficient, than something more centralised.
I am not arguing for maximal decentralisation, where everyone is working on their own. I think small teams of around 3-7 people are often ideal. That's also large enough groups for people who want to have a boss can have one.
A note on independent researches, I think many of them do work closely with other researchers in practice, but I this is mostly my guess. If a lot of people are doing research alone and would like colleges, I think helping people find collaborators would be great. Bu especially in research, there is no need to work for the same org, in order to collaborate.
Lack of financial security is a real problem. I would not personally want to trade more security for less freedom, and there are potential solutions where I don't have too, if the funders are on board. I would love there to be a system where if you've done good work for some length of time you get a grant lasting for 3 years to just "keep up the good work". I do understand why funders would hesitate do give such a grant, but I think it would be worth it.
It's also important to recognise that a fully decentralised model favours people with greater privilege.
Small projects with uncertain funding are much easier to join for people who are financially secure and don't care about things like maternity leave.
Strong upvote. I think there is economic research showing that people are more productive in large companies than in small ones, and that management practices significantly affect worker productivity. I haven't looked much into this, but it suggests that small organizations could be failing to realize their potential.
I'd be suppressed if this is true in general. Can you find some of this research? I'd be interested in finding out what specifically their results are.
Would you say more organizational support would make for a broader equilibrium for EA-organizational needs...
Perhaps a 'Nash-Equilibrium'?
Does anyone have a handy typology of small EA organizations? The potential arguments seem to be at least somewhat distinct for (e.g.) people doing community-building work vis-a-vis independent researchers. The weight assigned to various advantages and disadvantages will vary based on type of work, legal riskiness, etc.