Moving Toward More Concrete Proposals for Reform

By Jason @ 2023-01-27T00:36 (+88)

Epistemic status: uncertain, unsure of the extent that I mean this as a practical proposal vs. a thought experiment to make certain observations, aware I could improve this but also aware that it should be timely

Timing note: Most of this was written before Doing EA Better was released, and is largely based on public events on the Forum. It is not specifically in response to that post, and is not really about possible epistemic reforms at all.

Summary

This post suggests that discussions of possible reforms relating to governance, deconcentration of power, and transparency have often been relatively unhelpful due to a lack of specificity. It suggests raising modest funding for the creation of more developed proposals, with the broader EA community taking the lead on the project. The post presents various reasons I think this proposal has value even if one assumes the underlying reform proposals lack merit. It explains how the proposal would provide information for and tie into further reform initiatives, and suggests some mechanics for funding and selecting proposals. 

Introduction

I find the very-high level discussion of possible reform to be generally unhelpful. At that level, it's too easy for proponents to gloss over implementation costs and downsides, and too easy for skeptics to attack strawmen. There needs to be more substance for anyone to seriously evaluate most of the proposals that are floating around as possibilities -- either in the abstract or in terms of whether investing resources in a proposal specific to their organization makes sense.

I've also noted, as have others, that it's unreasonable to place the burden of producing more detailed proposals on reform advocates without either a strong reason to believe those reforms would be successful or providing reasonable compensation for their work.  (Of course, no one should ever feel pressure to write even if compensation is offered!)

For what it's worth, I think some of the reform proposals going around are likely correct, some are worthwhile in carefully selected contexts, and some are very unlikely to be viable. But I also think, for many proposals, that my assessments could update significantly if I read more detailed versions of the proposal.

I'm a newcomer who has never accepted anything from any EA source other than one free book. This gives me certain advantages and limitations. I don't have to worry about damaging effects on my career, nor do I personally have anything to personally lose or gain from reform. On the other hand, my outsider status means I am probably ignorant of some important information. And it certainly means I have no right to make demands on anyone, only suggestions.

Background Assumptions

Most of the money in the ecosystem comes from Open Phil and a few other large donors. However, there is still a respectably-sized slice of independent funding from small to mid-size donors. I submit that certain functions within EA are in particular need of a diversified funding base, rather than one that is dependent on Open Phil or other megadonors. In particular, the development of proposals about power, governance, and transparency is one of those functions. Not only are people not particularly good at deciding whether they need to give up power or expose their decisions to transparency, they also have an obvious conflict of interest on the question. One can doubtless think of other functions for which independence from major funders is important or even critical.

For various reasons, I identify with a desire for moderate reform at this time. I'll define that (imprecisely) as a set of reforms that seeks to make EA a better version of itself, but still clearly identifiable as EA and not radically different in organization or operations. The proposal here comes from that perspective, but certainly can be adapted for more extensive/fundamental reform.

Taking the Next Incremental Approach on Possible Reforms Would Be Inexpensive.

I don't think it would take that long to bring individual reform proposals to the next step. By the next step, I do not mean something particularly polished, shovel-ready, or (in most cases) targeted to a specific organization. Rather, I mean something that briefly describes each proposed reform, identifies expected costs and benefits at a moderately high level, lists the organizations to which it most likely applies, and addresses any issues about feasibility that are fairly obvious on their face. I am imagining that would be 2000-3000 words and maybe 12-18 hours of work for most reforms, but have only moderate confidence in that estimate. Optionally, one could invite two critics to spend one hour each preparing a comment on the writeup for concurrent publication.

Of course, writers should be compensated fairly, but not handsomely.  I confess to knowing relatively little about compensation norms here, being in full-time non-EA work. But skimming past grant writeups from EAIF, it sounds like $60 per hour may not be too far off.^[1]

So let's do a BOTEC: 12 writeups * 17 hours * $60/hour would be $12,240 -- let's add 25% for contingencies, and things I'm not thinking of, to get to $15,300. If you wanted to see a broader range of moderate and more extensive reforms, maybe it should be 18 writeups. The right amount of overhead cost to budget for ops/backend work is unclear to me and would depend on specific implementation details. Of course, one could adjust the scope and depth based on how much one had to spend.

I envision that one proposal would be posted to the Forum each week or so, to give the community a chance to comment on each without overload. It would be helpful to have an associated poll.^[2]

Perhaps a Governance, Transparency, and Reform Contest would be the orthodox EA way to do this, but my assumption is that the prize fund would need to be somewhat higher. Moreover, I sense that some people think that they might not get a fair shake from a judging panel that may not share their specific views on reform. Guaranteeing funding up front based on a short proposal avoids having to ask people to invest lots of hours of work based on trust in the judging panel.

Why Even Reform Skeptics Should Support At Least Some Further Investigation.

In my view, this should be a helpful exercise even if your priors are strongly against most reforms and you assume none of the reforms would (or should) ultimately be adopted.

Suppose you think, for example, that the suggestions for whistleblowing reforms lack merit. It's clear the idea has surface-level appeal to a lot of people inside and especially outside EA, especially given events of the past few months. If you think the idea is weak, then commissioning a next-step proposal should only expose the idea's weaknesses. 

Conditioned on more whistleblower protections being a bad idea, I think it is better for EA to be able to show that it encouraged (and compensated) someone selected by the rank-and-file community to take their best shot, but the community decided after that exploration that the idea does not actually work in the EA context. Justifications offered after the next major scandal will come off as post hoc rationalizations driven by motivated thinking.

The current public narrative seems to be that EA is summarily dismissing these ideas without giving them a hearing -- and one could point to certain Forum comments in support of that characterization. Likewise, there are portions of the public narrative that imply that the people in positions of power are the people who decide whether to fund the writing of proposals to reduce their power. That is not good.

Presumably, most people from outside EA who are considering becoming large donors will do their own research. They will find the governance/transparency criticisms quite easily, and I think a significant number of them will find those criticisms to have some validity on their face. A summary dismissal will not help EA's chances with that funder. Even slightly decreasing the risk that not-yet-EA donors will be turned off by EA's response to governance/transparency criticisms has a meaningful expected value.

There are also recruiting and retention arguments here. I recognize that some people here don't care about governance-type issues, but some people do -- I see this pol.is data as some evidence that a sizable number of people attracted to EA think these types of issues are important.  It's clear to me that some people are experiencing disillusionment surrounding the issue of reform. Moving specific proposals to a next step of research would help signal to those people that their concerns are being taken seriously -- which is generally helpful for morale, even when one disagrees with the ultimate decision that was made. More generally, lots of people are making great sacrifices to work in EA -- they could get paid far more, and enjoy greater social standing, in other domains. It's not unreasonable for them to expect that their ideas get more of a public hearing than they have.

To the extent that you may be concerned about the risk of distracting people from more important work (especially to the extent you think AGI is near at hand): I think we have to trust individual EAs to evaluate the benefits of working on reform proposals against the counterfactual use of their time. Also, the same pol.is data suggest that the reform-minded population may be somewhat less into longtermism than the EA population as a whole, so the time expenditure here may come disproportionately from EAs not as focused on longtermism.

Finally, to the extent that you think senior-leader bandwidth is a limiting factor, that might be an argument against certain reforms at this time. However, at some point, leaders have to trust their lieutenants with moderate-impact decisions (and even their ensigns with lower-impact decisions) that they do not have time to make.

To More Extensive Reform Advocates

I think seeing how some more concrete/specific reform proposals fare would help you figure out your next steps. As noted earlier, I don't think community reaction (either on the Forum or elsewhere) to a list of summarily-stated proposals is necessarily a valid indicator of reaction to more detailed proposals -- in either direction. I think you're at some risk of spreading yourself too thin, and getting better information about what the demos actually wants would be beneficial. Although I don't agree with everything you have proposed, I do think your ideas have value and would hate to see little come out of your passion and commitment.

The proposal/thought experiment is in a sense a possible prelude to an independent reform organization. I think many of your proposals -- including some I would likely support -- will need one or more institutional homes that are legally, financially, and organizationally independent. I think it's somewhere between very important and essential that (1) at least a significant portion of those organizations' funding does not come directly or indirectly through megadonors; and (2) a reasonable cross-section of the EA community would be willing to devote some time to the organizations (with reasonable compensation).

Not only do I think taking too much megadonor money would undermine the reform organization, I think requiring the majority of the support to come from rank-and-file EAs would be an important means of keeping the organization(s) accountable to and grounded in the community's interest. For instance, GiveWell doesn't allow anyone to contribute more than 20% toward its operating budget. So part of the proposal/thought experiment is seeing how much interest ordinary EAs have in concretely supporting this kind of work.

Who Should Fund Work Like This?

Of course, one could write all this up and apply for some money from a funder like the EAIF -- and I'd probably be pleased to see one of the usual funders pick up a significant fraction of the costs of a project. Anonymous and untraceable crypto donations could also work! But I submit that it is healthy for interested rank-and-file EAs to directly contribute at least a significant fraction of the costs for several reasons:

First, I don't think it is healthy for major funders -- who have significant power in the ecosystem -- to have (or even appear to have) any influence over what reform-related proposals get written up. To accrue some of the benefits I described above, this kind of work needs to be seen as controlled by the broader community (and actually be so controlled). 

One reason I suggested a major funder should have a role here at all is that it gives them an opportunity to signal some openness to reform. To be honest, the largest value-add from an established organization for early-stage efforts might be some sort of ops support / fiscal sponsorship.

Second, I think rank-and-file EA members specifically routing a modest percentage of their financial giving to fund reform-minded work has an important signaling / information value. Upvotes -- even strong upvotes -- are cheap and can be performative. Choosing to fund independent reform evaluations conveys that the community member's preference is at least somewhat strong and considered. 

On the other hand, if there isn't interest from the community in providing even modest funding for this sort of work, I think that itself is a valuable piece of information. To send a non-cheap (but not too costly) signal myself: I would be willing to devote up to 2% of my expected charitable spend for the year (about $350) on this, conditioned on enough tangible interest from the community.

Who Should Make Decisions About What Proposals to Fund?

Ideally, decisions about what proposed topics to fund, and who to fund, should be made by the individuals who donated to the effort who are not affiliated with powerful EA organizations. This is for two reasons: first, the legitimacy of the project depends on confidence that the strongest reform proposals were selected. Limiting the voting pool to those willing to donate, while having some downsides, helps guard against possible tactical voting by reform opponents. Second, I think that, in general, decisions are better made when the decisionmaker has some "skin in the game." How to operationalize that, particularly from an IT/ops perspective, is not a topic on which I claim any real insight.^[3]

The program would probably solicit unpaid 300-500 word summaries from people interested in writing up ideas, and then perform a preliminary ranking by approval voting. Ideally, a small committee would make the final decisions on the set after considering balance and overlap, and would explain any deviations from the rank order. E.g., the sixth- and eighth-ranked proposals might be too similar to proceed with both, but the sixteenth-ranked proposal might justify a spot based on balance criteria.

In reality, given time and ops constraints, it might be more realistic to ask some Forum members who have posted in favor of moderate reforms to choose the funded moderate proposals, and to invite the ConcernedEAs to pick most of the more extensive ones (with a few selections by others).

A Broader Approach

If an initial project is a success, the next step for community-influenced reform efforts might be to establish a small community-funded and community-run grantmaker focused on reform-minded efforts. For instance, this grantmaker could help fund more specific proposals, or even implementation costs, for organizations that were interested in undertaking reforms. 

One could argue that it would be better to just give unrestricted donations to those organizations and defer to their leadership's priorities. However, I think donors should be much less willing to defer to organizational leadership's judgment when the funding priority in question would do things like deconcentrate power, improve transparency, or bring in new perspectives. Also, I assume that the priorities of Open Phil and other major grantmakers have a significant influence on the direction of various organizations they fund (whether intentionally or otherwise), so it's difficult to say that the community-run grantmaker should defer to the organizations and not seek to have any influence.

Conclusion

Hopefully, we can move forward to researching and discussing more specific proposals soon. I'll get off my soapbox now. Thanks for listening to my first post!

1. ^{^}

   This was based mainly on a grant of $1700 for 30 hours of work here. 

2. ^{^}

   Polls as a measure of community support have their problems. But so does watching the reaction on the Forum, which is not a representative sample. The idea is that having multiple ways to gauge community reaction would be helpful, so that the weaknesses of one method are at least partially covered by a second method.

3. ^{^}

   Theoretically, I'd suggest that donation size has a correlation to investment in the issue but is a noisy measure. Thus, my tentative theoretical approach would be that decisionmaking influence (DI) should be equal to the square root of donation size, subject to a floor and a ceiling. The floor could be about $49 (7 DI) for full-time workers in high-income countries making at least $49,000, $9 for anyone in a low-income country or students in middle-income countries, and $25 for everyone else. The cap could be $1600 (40 DI). But this is probably too complicated for the scope of the proposal/thought experiment.