Here’s another reason one might want to diversify their altruistic spending, which may or may not be distinct from the reasons in Jeff’s post: One’s budget may be too small to see significantly diminishing marginal returns, but one may belong to a population of people which as a whole sees diminishing marginal returns. Then one might want to diversify one’s spending so that other members of the population diversify their spending. Givewell used such reasoning when it suggested splitting one’s donation among Givewell’s top-rated charities so that other Givewell donors, following the same principle, would collectively cause Givewell’s money-moved metric to reflect this diversity, which would in turn give Givewell better access to more charities.
Agreed – the arguments against diversification don’t take into account the fact that individuals’ choices will be more highly correlated if they use similar decision processes. In order to make an optimal allocation decision you would need to estimate the total amount of money being donated by people who are using an algorithm similar to yours and decide as though you were allocating the entire amount. Sounds familiar, eh?
Here’s another reason one might want to diversify their altruistic spending, which may or may not be distinct from the reasons in Jeff’s post: One’s budget may be too small to see significantly diminishing marginal returns, but one may belong to a population of people which as a whole sees diminishing marginal returns. Then one might want to diversify one’s spending so that other members of the population diversify their spending. Givewell used such reasoning when it suggested splitting one’s donation among Givewell’s top-rated charities so that other Givewell donors, following the same principle, would collectively cause Givewell’s money-moved metric to reflect this diversity, which would in turn give Givewell better access to more charities.
Agreed – the arguments against diversification don’t take into account the fact that individuals’ choices will be more highly correlated if they use similar decision processes. In order to make an optimal allocation decision you would need to estimate the total amount of money being donated by people who are using an algorithm similar to yours and decide as though you were allocating the entire amount. Sounds familiar, eh?