(There’s a lot more I might want to say about this, and also don't take the precise 80% too seriously, but FWIW:)[1]
When we do cause prioritization, we’re judging whether one cause is better than another under our (extreme) uncertainty. To do that, we need to clarify what kind of uncertainty we have, and what it means to do “better” given that uncertainty. To do that, we need to reflect on questions like:
- “Should we endorse classical Bayesian epistemology (even as an ‘ideal’)?” or
- “How do we compare actions’ ‘expected’ consequences, when we can’t conceive of all the possible consequences?”
You might defer to others who’ve reflected a lot on these questions. But to me it seems there are surprisingly few people who’ve (legibly) done so. E.g., take the theorems that supposedly tell us to be (/“approximate”?) classical Bayesians. I’ve seen very little work carefully spelling out why & how these theorems tell either ideal or bounded agents what to believe, and how to make decisions. (See also this post.)
I’ve also often seen people who are highly deferred-to in EA/rationalism make claims in these domains that, AFAICT, are straightforwardly confused or question-begging. Like “precise credences lead to ‘better decisions’ than imprecise credences” — when the whole question of what makes decisions “better” depends on our credences.
Even if someone has legibly thought a lot about this stuff, their basic philosophical attitudes might be very different from yours-upon-reflection. So I think you should only defer to them as far as you have reason to think that’s not a problem.
- ^
Much of what I write here is inspired by discussions with Jesse Clifton.
I personally dealt with this (in part) by referencing Jack Malde's excellent guided cause prio flowchart (this was a first draft to gauge forum receptivity). Sadly, when asked about updates, he replied that "Interest seemed to be somewhat limited."
I'm glad you found this useful Mo!
Thanks for sharing, Mo! I do not think humans need to have special status for one to prioritise interventions targeting humans. I estimate GiveWell's top charities may well be more cost-effective than interventions targeting animals due to effects on soil nematodes, mites, and springtails.
Thanks Vasco, I really appreciate your work to incorporate the wellbeing of wild animals into cost-effectiveness analyses.
In your piece, you focus on evaluating existing interventions. But I wonder whether there might be more direct ways to reduce the living time of soil nematodes, mites, and springtails that could outperform any human life-saving intervention.
On priors it seems unlikely that optimizing for saving human lives would be the most effective strategy to reduce wild animal suffering.
Thanks, Jack! I agree it is unlikely that the most cost-effective ways of increasing human-years are the most cost-effective ways of increasing agricultural-land-years. Brian Tomasik may have listed some of these. However, buying beef decreases arthropod-years the most cost-effectively among the interventions for which Brian estimated the cost-effectiveness, and I estimated GiveWell's top charities are 2.65 (= 1.69/0.638) times as cost-effective as buying beef.