This philosophy seems at stark odds with 80k's recent hard shift into AI safety. The arguments for the latter, at least as an extinction risk, necessarily lack good evidence. If you're still reading this I'm curious whether you disagree with that assessment, or have shifted the view you espoused in the OP?

I'm really glad you're writing about this. I think this is an important criticism of the way the EA movement and a lot of individuals within it (myself very much included) often come off. I think I'd suggest a different focus for what to say in these situations (although it's compatible with many of your suggestions).

In particular, most of these suggestions seem fairly focused on stating/advocating for your own position while accurately expressing your uncertainty. Instead, when you don't have good evidence that your view is correct, I think the most important thing to do is to focus on asking questions. I think this is the most important bit of humility. It's also likely to lead to more learning. And, if you really don't know much about their alternative (or have good evidence that yours is better) you're not likely to convince anybody anyway.

I think a common mistake I make is to express humility by continuing to advocate for my position while making my uncertainty more explicit.(1) People often don't read this as humility, though, because I'm not acting as if I believe they might have evidence that they're right and I don't sound curious about their alternative and about whether I'm wrong.

When I'm emotionally invested in a topic (particularly value-laden issues like EA) I often struggle to remember to be in learning-mode instead of persuasion-mode -- even when I'm genuinely curious about what the other person has to say. FWIW, one strategy I've personally used in this situation is to try to mentally keep track of the amount of time I'm spending explaining my position versus listening to theirs. If i don't have good evidence about their position, hearing their take is usually more interesting (even if my system 1 sometimes forgets this).

A caveat

One failure mode with my approach, though, is that there can be a fine line between trying to learn about someone's position (which comes off as humble) and interrogating them (which does not). When I'm trying to make sure I'm not interrogating someone, the question I usually ask myself is:

• "Did I ask this question because I think they will have a good answer or because I think they will not have a good answer?"

(1)I had to learn the hard way that, at least for me, this doesn't actually come off as less confident. Instead it comes off as more confident AND better calibrated. Which is an improvement but doesn't lend itself to coming off as humble or to making others feel comfortable expressing disagreement.

I realize this is a higher-level discussion, but I am curious-by research into catastrophic risks do you mean AI specifically? Because I would be disheartened if you suspected that research on asteroid deflection, probabilities of high energy physics catastrophes, how to prevent global totalitarianism, how to prevent nuclear conflict, how to reduce nuclear stockpiles, how to ramp up conventional or alternative food supplies in a catastrophe, how to make global cooperation in a catastrophe more likely, prioritization within GCR, etc. are about as likely to have positive as negative effect.

Thanks for posting this. It is very easy to come off as arrogant when promoting Effective Altruism, but this article helps by providing specific examples of what you can reasonably say.

I agree with this kind of humility wholeheartedly. Although I think part of the problem is inseparable from what has to be called the righteous belief of most effective altruists that they are not propounding one way of doing good, but the single best way - one at which any rational reflection must conclude. Of course, they might disagree about which particular intervention has the greatest impact, but that disagreement occurs within the agreed framework of effective altruism.