Don't Split EA In Two
Against the imaginary movement you made up
In the wake of various effective-altruism-related scandals and kerfluffles, I’ve seen people say that the effective altruism movement should split in two: there should be one movement that focuses on effective spending in global health and development, and one movement that focuses on all of the other weird stuff.1
That split already happened. The movement to spend global health and development money more effectively is called “development economics.” It is very popular and successful. The founders of the randomista movement won a Nobel Prize. A leading development economist is one of the leaders of USAID. Charities like Evidence Action were not founded by effective altruists; they were founded by development economists.
The big effective altruist contribution to development economics—other than donations—is GiveWell. At its core, GiveWell takes the development economics research (that was already happening) and identifies charities for individual people to donate to, unlike other organizations, which influence large foundations or lobby governments. GiveWell is an excellent charity and I myself donate to their All Grants fund. I’m glad that individuals concerned about global health don’t have to just donate to Médecins Sans Frontières and hope that its prioritization is sensible. But GiveWell didn’t in any way invent the concept of “do studies to check whether your spending on global poverty works.”
You could imagine a mass activist movement for effectiveness in aid spending. But aid effectiveness is a terrible issue for mass activism: wonky, detail-oriented, and immensely dull. Normal people simply don’t care about foreign aid. The solution here is for boring technocrats (subtype academic) to reach out to boring technocrats (subtype civil service) and get them to make a bunch of boring technical changes no one cares about which incidentally save hundreds of thousands of lives. People are doing that, it’s noble work, and as far as I can tell it’s going pretty well.2
The only other service I can think the Hypothetical Effective Global Health And Development Spending Movement might provide is support in continuing to donate to charity. I agree it’s good for people to donate to global-health charities and most people will require some light social pressure to do so. But “you should donate to charity” does not a subculture make. Imagine:
Alice: So, still donating?
Bob: Yep. It’s automatically deducted from my paycheck. You?
Alice: Yep. Same here.
Bob: …
Alice: …
Bob: So, have you seen the new season of Severance?
Alice: We actually share no common interests because we’re the Charity Donation Support Group.
Bob: Right, right. Uh. Same time next week?
Alice: Don’t call me, I’ll call you.
If you want to donate 10% of your income to global-poverty charities, want social support in continuing to do this, and find effective altruism too weird, join a religion. I’m serious! You might have to shop around a bit—a lot of religions already have strong opinions about what you should be donating to—but we have this entire societal infrastructure for people holding each other to demanding moral standards. Maybe you can even tell your coreligionists about GiveWell.
Effective altruism is not and has never been a global-poverty-donation support group for people too precious to become Quakers. Animal advocacy, global catastrophic risks, and (yes) artificial intelligence have been part of the DNA of the movement since the beginning. Effective altruism is the attempt to answer the question of how to best improve the world, following a set of assumptions including but not limited to:
Morally relevant traits include things like “the capacity for complex experiences” and not things like “lives near me” or “shares my species.”
The primary thing that matters in ethical decision-making is the effects of your actions on the world.
It is possible and helpful to make decisions using math even if you are very uncertain about many of the numbers.
The world is rapidly changing and we shouldn’t assume that it will look the way it did when we grew up.
You should not dismiss a potential way of doing good out of hand because it sounds kind of goofy.
“Effective altruism but just the global-poverty parts” isn’t effective altruism, because you’ve already picked what the best way to improve the world is. Similarly, “physics but just Newtonian mechanics” isn’t physics, “history but China is the most important country in the world” isn’t history, and “economics but the invisible hand of the market generally leads to the best outcomes for everyone” isn’t economics. If you’re trying to answer a question, you don’t get to decide what the answer is ahead of time. This is true even if your answer is right!
There is not a discrete set of “global poverty effective altruists” that can be severed from everyone else. We talk about cause prioritization, but many highly involved effective altruists are messy: wild-animal-welfare advocates who donate to global poverty charities; technical AI safety researchers who are vegan; GiveWell employees who are mulling over a switch into biosecurity. In my anecdotal experience, to the extent that there are people who are like “my thing is overwhelmingly more important than everyone else’s things and no one else should be working on anything else,” they are nearly universally doing AI safety work. Sorry, everyone, the global poverty donors are also freaking out about large language models and the wellbeing of tuna caught in nets.
I think that this is what you’d expect. There are multiple plausible answers to the question “what is the best thing to do, given effective altruist assumptions?”3 Nearly everyone will be sympathetic to more than one answer, if there’s not some overwhelming consideration like “this thing will probably destroy the world in fifteen years.” Some answers cluster—if you don’t care about farmed animals, you are unlikely to care about wild animals—but nothing clusters based on “how appealing does this sound to people totally uninvolved in the movement” because… why would it do that?
If you don’t like the thing effective altruism is, that’s fine. If you want the intellectual movement that’s trying to use evidence to alleviate global poverty, I wish you the best of luck in getting into development economics.4 If you want support in keeping to your donation goals, I wish you luck in that too. But there is no sense in making up an imaginary version of effective altruism and being mad that it won’t split off from the effective altruism that actually exists.
Come on, can’t being worried about factory farming be normal? No? We’re with the AI people? Fine.
Maybe you disagree with me and you think that a social movement in favor of effective foreign aid spending would work very well, in which case I encourage you to try to start one.
Traditionally divided into three groups, but I’d count more like a dozen: global health, policymaking in the developing world, scientific research aimed at helping people in the developing world, the whole progress studies cluster, farmed animal welfare, wild animal welfare, AI safety, biosecurity, nuclear war prevention, climate change, global priorities research, and a twelfth for whatever I’m forgetting.
Poor Economics is very readable!
I think the framing of this post goes for a strawmanned version of your opponent. As I understand the discourse, it is not about "spin out specifically global health and development" (which, as you point out, would be weird). As I understand it, it's "spin out specifically AI risk" (see e.g. Will MacAskill here https://forum.effectivealtruism.org/posts/euzDpFvbLqPdwCnXF/university-ea-groups-need-fixing?commentId=Bi3cPKt27bF9GNMJf).
A chief reason for this is that there is a tendency in recent EA for AI-risk and related worries to take over a disproportional amount of attention, particularly among highly-engaged EAs. Two examples which I think were bad:
1. Germany has one of the largest EA communities globally, partly through the work of the Effektiver Altruismus organization being professionalized early on. A few years ago, the organization quite abruptly dissolved without real replacement because the people involved wanted to work on S-risks. This left a pretty substantive gap in the German EA infrastructure up untill somewhat recently.
2. 80.000, one of the most promising EA organizations, has deprioritized global health & development, as well as animal welfare issues, and now seems to focus almost exclusively on longtermist causes (most prominent of which is, of course, AI risk). This strikes me as quite the loss, as it no longer seems sensible to refer people willing to change careers to the famous "EA switching careers" advisory organization, unless you want them to be nudged strongly towards longtermist causes.
You may think that this switch is fine - maybe longtermism, and AI risk as a specific cause, should really dominate the EA movement, because they rank best on utility calculations. But I think there are significant, unaddressed criticisms of those calculations - see e.g. Thorstadts article here https://globalprioritiesinstitute.org/david-thorstad-three-mistakes-in-the-moral-mathematics-of-existential-risk/, which pushes back on the few examples of explicit calculations favouring existential risk reduction, which he believes rest on deeply implausible assumptions. This is if they are underpinned by calculations at all - Thorstadt had to go to great lengths to even find any concrete numbers being provided for the cost-efficiency of risk reduction being greater than that of other shorttermist interventions. I suspect that there is a good chance that other, less epistemically kosher dynamics (https://forum.effectivealtruism.org/posts/euzDpFvbLqPdwCnXF/university-ea-groups-need-fixing) may drive highly engaged EA's to have significantly above-average rates of buying hard into global risk reduction than more "traditional" and boring areas - some of which you outline here: it is certainly more exciting to be doing research on a fancy new technological risk than to have to engage with the drudgery of modern development economics.
I think your post misunderstands the reasons why people wonder if a part of EA should be split into its own movement, and it did not make me update on it. I am unsure if splitting EA into a future risk and a "all other things" branch would be good. I think it is important that EA works on global catastrophic risk reduction, and have encouraged and helped people enter that space. But I do think there is a real tendency for that specific sector to eat up other sectors in attention, funding, talent and infrastructure, I believe that tendency is stronger than currently justified by the positive arguments for risk reduction, and I believe that trend actively harms the Effective Altruist movement currently.
IME most people who want a split want anti-factory-farming on the normal side with the bednets, not on the weird AI side. I'm one of the people you describe, but I seem to be in a minority.
Okay, so what I want isn't the Effective Altruism movement (except for the company, y'all nerds are great). But the places you suggest are even worse fits.
The movement I want is evangelistic, at least to the extent of being, y'know, a movement, not a social club. I'd like every single person in the world to give 1% of their income, and everyone in rich countries who's not particularly poor to give 10%. I'd like a world where that's seen as basic decency, like holding doors open for people.
It doesn't look like quiet work to direct government spending: Sure, that's part of it, directing foreign aid is great, but private donors save a bunch of lives. Wonkery is necessary — I want charity money to actually help, not be wasted — but most donors aren't wonks, we're just paying the wonks to tell us where to send the cash.
It doesn't look like the religion option: I want a big tent. It's perfectly fine if charitypilled people share no common interests, just like the goal of suffragism was votes for all women, not pleasant conversation between suffragists.
The Effective Altruist movement as it exists is a decent fit for this: GiveWell is moving about half a billion USD per year from private donors to excellent charities. Between EA (big social movement, lots of stuff I disapprove of and find annoying to be associated with) and randomistas (very good, but purely academic), I stick with EA for now.
But I'd love a big social movement whose heroes are, like, Melinda Gates and Saloni Dattani. It's not obvious to me whether this is a harder sell, or an easier sell, than EA as it currently exists.