I quit the Effective Altruism forum due to a new rule requiring all new posts and comments be basically put in the public domain without copyright, so anyone could e.g. sell a book of my posts without my consent (they’d just have to give attribution). More info. I had a bunch of draft posts, so I’m posting some of them here with minimal editing. In general, I’m not going to submit them as link posts at EA myself. If you think they should be shared with EA as link posts, please do it yourself. I’m happy for other people to share links to my work at EA or on social media. Please share stuff in whatever ways you think are good to do.
I haven’t studied EA much. There is plenty more about EA that I could read. But I don’t want to get involved much unless EA is rational.
By “rational”, I mean capable of (and good at) correcting errors. Rationality, in my mind, enables progress and improvement instead of stasis, being set in your ways, not listening to suggestions, etc. So a key aspect of rationality is being open to criticism, and having ways that changes will actually be made based on correct criticisms.
Is EA rational? In other words, if I study EA and find some errors, and then I write down those errors, and I’m right, will EA then make changes to fix those errors? I am doubtful.
That definitely could happen. EA does make changes and improvements sometimes. Is that not a demonstration of EA’s rationality? Partially, yes, sure. Which is why I’m saying this to EA instead of some other group. I think EA is better at that than most other groups.
But I think EA’s ability to listen to criticism and make changes is related to social status, bias, tribalism, and popularity. If I share a correct criticism and I’m perceived as high status, and I have friends in high places, and the criticism fits people’s biases, and the criticism makes me seem in-group not out-group, and the criticism gains popularity (gets shared and upvoted a bunch, gets talked about by many people), then I would have high confidence that EA would make changes. If all those factors are present, then EA is reliably willing to consider criticism and make changes.
If some of those factors are present, then it’s less reliable but EA might listen to criticism. If none of those factors are present, then I’m doubtful the criticism will be impactful. I don’t want to study EA to find flaws and also make friends with the right people, change my writing style to be a better culture fit with EA, form habits of acting in higher status ways, and focus on sharing criticisms that fit some pre-existing biases or tribal divisions.
What can be done as an alternative to listening to criticism based on popularity, status, culture-fit, biases, tribes, etc? One option is organized debate with written methodologies that make some guarantees. EA doesn’t do that. Does it do something else?
One thing I know EA does, which is much better than nothing (and is better than many other groups offer), is informal, disorganized debate following unwritten methodologies that vary some by the individuals you’re speaking with. I consider this option inadequately motivating to seriously research and critically scrutinize EA.
I could talk to EA people who have read essays about rationality and who are trying to be rational – individually, with no accountability, transparency, or particular responsibilities. I think that’s not good enough and makes it way too easy for social status hierarchies to matter. If EA offered more organized ways of sharing and debating criticism, with formal rules, then people would have to follow the rules and therefore not act based on status. Things like rules, flowcharted methods to follow, or step-by-step actions to take can all help fight against the people’s tendency to act based on status and other biases.
It’s good for informal options to exist but they rely on basically “everyone just personally tries to be rational” which I don’t think is good enough. So more formal options, with pro-rationality (and anti-status, anti-bias, etc.) design features should exist too.
The most common objection to such things is they’re too much work. On an individual level, it’s unclear to me that following a written methodology is more work than following an unwritten methodology. Whatever you do, you have some sort of methods or policies. Also, I don’t really think you can evaluate how much work a methodology is (and how much benefit it offers, since the cost/benefit ratio is what matters) without actually developing that methodology and writing it down first. I think rational debate methodologies which tries to reach conclusions about incoming criticisms are broadly untested empirically, so people shouldn’t assume they’d take too long or be ineffective when they can’t point to any examples of them being tried with that result. And EA has plenty of resources to e.g. task one full-time worker with engaging with community criticism and keeping organized documents that attempt to specify what arguments against EA exist, what counter-arguments there are, and otherwise map out the entire relevant debate as it exists today. Putting in less effort than that looks to me like not trying because the results are unwanted (some people prefer status hierarchies and irrationality, even if they say they like rationality) rather than because the results are highly prized but too expensive. There have been no research programs afaik to try to get these kinds of rational debate results more cheaply.
Also, suppose I research EA, come up with some criticisms, and I’m wrong. I informally share my criticisms on the forum and get some unsatisfactory, incomplete answers. I still think I’m right and I have no way to get my error corrected. The lack of access to debate symmetrically prevents whoever is wrong from learning better, whether that’s EA or me. So the outcome is bad either way. Either I’ve come up with a correct criticism but EA won’t change; or I’ve come up with any incorrect criticism but EA won’t explain to me why it’s incorrect in a way that’s adequate for me to change. Blocking conclusive rational debate blocks error correction regardless of which side is right. Should EA really explain to all their incorrect critics why those critics are wrong? Yes! I think EA should create public explanations, in writing, of why all objections to EA (that anyone actually raises) are wrong. Would that take ~infinite work? No because you can explain why some category of objection is wrong. You can respond to patterns in the objections instead of addressing every objection individually. This lets you re-use some answers. Doing this would persuade more people that EA is correct, make it much more rewarding to study EA and try to think critically about it, and turn up the minority of cases where EA lacks an adequate answer to a criticism, and also expose EA’s good answers to review (people might suggest even better answers or find that, although EA won the argument in some case, there is a weakness in EA’s argument and a better criticism of EA could be made).
In general, I think EA is more willing to listen to criticism that is based on a bunch of shared premises. The more you disagree with and question foundational premises, the less EA will listen and discuss. If you agree on a bunch of foundations then criticize some more detailed matters based on those foundations, then EA will listen more. This results in many critics having a reasonably good experience even though the system (or really lack of system) is IMO fundamentally broken/irrational.
I imagine EA people will broadly dislike and disagree with what I’ve said, in part because I’m challenging foundational issues rather than using shared premises to challenge other issues. I think a bunch of people trying to study rationality and do their best at it is … a lot better than not doing that. But I think it’s inadequate compared to having policies, methodologies, flowcharts, checklists, rules, written guarantees, transparency, accountability, etc., to enable rationality. If you don’t walk people step by step through what to do, you’re going to get a lot of social status behaviors and biases from people who are trying to be rational. Also, if EA has something else to solve the same problems I’m concerned about in a different way than how I suggest approaching them, what is the alternative solution?
Why does writing down step by step what to do help if the people writing the steps have biases and irrationalities of their own? Won’t the steps be flawed? Sure they may be, but putting them in writing allows critical analysis of the steps from many people. Improving the steps can be a group effort. Whereas many people separately following their own separate unwritten steps is hard to improve.
I do agree with the basic idea of EA: using reason and evidence to optimize charity. I agree that charity should be approached with a scientific and rational mindset rather than with whims, arbitrariness, social signaling or whatever else. I agree that cost/benefit ratios and math matter more than feelings about charity. But unfortunately I don’t think that’s enough agreement to get a positive response when I then challenge EA on what rationality is and how to pursue it. I think critics get much better responses from EA if they have major pre-existing agreement with EA about what rationality is and how to do it, but optimizing rationality itself is crucial to EA’s mission.
In other words, I think EA is optimized for optimizing which charitable interventions are good. It’s pretty good at discussing and changing its mind about cost/benefit ratios of charity options (though questioning the premises themselves behind some charity approaches is less welcome). But EA is not similarly good at discussing and changing its mind about how to discuss, change its mind, and be rational. It’s better at applying rationality to charity topics than to epistemology.
Does this matter? Suppose I couldn’t talk to EA about rational debate itself, but could talk to EA about the costs and benefits of any particular charity project. Is that good enough? I don’t think so. Besides potentially disagreeing with the premises of some charity projects, I also have disagreements regarding how to do multi-factor decision making itself.