Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

[Link] GiveWell and the problem of partial funding

1 Post author: Benquo 14 February 2017 10:48AM

Comments (3)

Comment author: Raemon 15 February 2017 03:23:08AM 1 point [-]

Thank you for writing this up - don't know if I had formally asked for it but I was hoping for a more concise summary. :)

Comment author: Jiro 20 February 2017 05:25:17PM 0 points [-]

It also assumes that people are taking the cost-per-life-saved numbers at face value, and if so, then GiveWell already thinks they’ve been misled.

This is absurd. There are such things as degrees of misleading. They could believe that the numbers are already somewhat misleading, but that using them this way is more misleading.

Comment author: gwillen 19 February 2017 01:11:10AM 0 points [-]

Hm, my fox (https://en.wikipedia.org/wiki/The_Hedgehog_and_the_Fox) and satisficer instincts really really don't the recommendation to 'unwind partial funding'. (I feel like there's a lot of stuff mixed into this post, but I am only talking about partial funding issues.) I thought I had seen something similar to the rough argument I'm about to make in a GiveWell/OPP blogpost, but it's not in the one you're writing about, so I'm not sure whether I did or not. If I did, I am probably partly plagiarizing it, badly.

The argument basically goes this way: I think it's very often the case that a mixed strategy is a good idea in practice, even if you're totally sure that one of two pure strategies must be superior, but you can't tell with any confidence which one it is.

It seems to me that you're arguing it's better to pick whichever of the two pure strategies -- full funding or no funding -- seems more likely than not to be superior, rather than do some of each. (It seems like in reality you think fully funding is a clear winner, but in 'Unwind partial funding' you seem to allow that either is possible -- just not anything in between.) In fact, I see in a longer post you state the notion that "GiveWell thinks that its recommendation underperformed opportunity cost, and therefore did net harm." As far as I can tell from my sense of your meaning, this is a perfectly utilitarian position, but the idea that underperforming opportunity cost is a net harm implies that any possible action that GiveWell takes with imperfect information is doing vast, tremendous, incalculable amounts of net harm. No matter how much good they do relative to the state of the world if they didn't exist, they are doomed to do huge quantities of net harm, relative to the world where they have certainty about what the optimal choices actually are.

This doesn't seem like a very encouraging position to take, in a messy world where human beings with extremely limited knowledge and optimization capacity are slowly groping their way towards doing good.

So I bristle at the idea that, because GiveWell is concerned that Good Ventures fully-funding their charities might cause harm -- but is certain that Good Ventures not funding those charities at all would cause harm -- they should be subject to moral outrage for some sort of dishonesty because they chose to hedge their bets.