Thread by @EpistemicHope

I'm less confident about the cost-effectiveness of AMF than I was a few years ago.

To my knowledge, no one who I know personally (whose epistemics I've evaluated) and who is not under distorting incentives, has double-checked that the oft-cited claims are correct.

https://twitter.com/1240344633013501952/status/1735732887620874322

(I'm ready for you all to come out of the woodwork and show me that I'm wrong about that : P

@slatestarcodex seems like he has maybe done the work here, and I've read enough of his argumentation to trust his epistemics.)

I know of a few folks who have done a more-or-less-deep dive into GiveWell's results.

At least two were unimpressed, but there's a selection effect where I'm more likely to hear from the people who think the results don't hold up than those who do.

But I haven't checked myself.

Again, my guess is that the claims are basically correct. But I don't think that I, personally, have that strong of a reason to think one way or the other.

Mainly a lot of people around me seem to think that malaria nets are among the most effective life-saving interventions.

https://twitter.com/1240344633013501952/status/1735732894101016800

And crucially, I now think it is a mistake to defer to GiveWell because other EAs say GiveWell is careful, and that I should trust EAs because they are the people who do their due diligence to maximize impact when giving to charity.

That's basically an epistemic pyramid-scheme.

Lest anyone think that's a strawman, this was basically my outlook in 2016.

I hadn't assessed GiveWell's research, but I confidently recommended that others give to GW recommended charities, because I trusted EA's milieu, because we were the thoughtful altruistic people.

Indeed, I still think that EAs are more thoughtful and more altruistic than the average person, by a lot!

But that only counts for so much if most of us are deferring to social cues and only a small fraction are doing difficult original thinking or verification.

My claim is not that EAs are particularly bad. It's that they give undue trust to the EA blob.

My impression from talking with lots of young EAs is that they have a very similar outlook towards EA as the one I had a few years ago.

They often talk about getting more resources "for EA", in a way that reifies the community, and takes for granted that the things it does are good, because we're the careful altruistic thinkers that can be trusted.

That EA having more power and more resources is obviously good.

They _might_ speculate about how EA could somehow do harm on-net for fun at parties, but my impression is that they don't really believe this is realistic in their bones.

(It's fun because it feels exotic and speculative. If it felt realistic, it would be scary instead.)

` Maybe EAs will make mistakes, but obviously EA having power is better than any of the alternatives! `

My probability of EA doing massive harm on net is probably my main crux with mainstream EAs.

The strongest example of this kind of "deferring to reified EA" is people donating to / campaigning for Carrick Flynn without knowing much about his character or his policies, on the argument that it was "good to have an EA in congress."

But I see it all over the place.

- Donating to EA funds
- Finding an "EA job" doing ops at an EA org
- Doing EA movement building
- Doing productivity / mental health / rationality / etc coaching for EAs.

All of these have impact-models that depend on a reified EA blob that does good things, such that non-specifically boosting the blob is good.

I feel pretty suspicious of people reasoning in that way, in general.

And this strategy seems a lot less like a good bet for positive impact if you think it's a live possibility that the EA blob will have negative impacts on net.

If so, then your own small boosting of EA is making the world worse, not better.

I think many EAs clearly trust reified EA in a way that amounts to "We are the good group!"

I think they don't really find it credible that EA as a whole could really lead to disastrous outcomes...which is one way "We are the good group!" can feel, from the inside.

(A an intuition pump: consider a social justice activist thinking that of course some activists will fail or make mistakes but that it's ridiculous that the crusade for anti-racism as a whole could lead to WORSE outcomes for black people, on net.

Or the 19th century communist disbelieving that communism would lead to 7 million deaths.

If you get some foundational assumptions wrong, you can totally have an impact far into the negatives. Not just doing nothing, but causing the OPPOSITE of your intended effect.

This volley of reference class tennis doesn't prove anything, but it might give my reader an intuition for how my view differs from theirs.

Especially since I've seen many iterations of people doing stuff to help with AI risk that probably made things worse.)

https://twitter.com/1240344633013501952/status/1734884684390416431

Now, there's an important quantitative question of what fraction of EAs (however you draw the boundary) are reasoning by the "EA is good, therefore boosting EA is good" pattern.

That's hard for me to estimate because of selection effects in which people I end up talking with.

My GUESS is that more than half of the attendees of EAGs over the past 2 years basically defer to EA in this way?

And the more so for the more involved EAs. Though less so for the EA thought-leaders.

But that's just a guess.

Ultimately, I'm reporting my sense of the vibe from when I talk with EAs, the sense that decreases my interest in EA, even though I share their principles, and am actively trying to figure out how to make the world much better.

I haven't done a careful anthropological analysis.

This whole dynamic is ironically diagonalizing. The more EAs believe that "EA is the community that does careful due diligence" the less careful due diligence EAs tend to do.

The more EAs take for granted that EA is good and does good things, the less likely that is to be true.

I have basically no objection to AMF and to people donating to AMF.

But I DO object to the distributed organism that defers to GiveWell on the basis of social cues, out of a social trust which is founded on "EAs being the good people who do the research."

Of course, some other people might have done the legwork to assess a lot of these object-level questions. And from their epistemic vantage point, some of the EA-endorsed actions might be slam dunks.

That I haven't done my due diligence no GiveWell doesn't mean NO ONE has.

But I roll to disbelieve that this is true of the typical young EA enthusiastically trying to get more people involved.

@threadreaderapp @Twtextapp @unrollthread unroll