Unless, of course, you believe that the decisions of other people donating to charity are correlated with your own. In this case, a decision to donate 100% of your money to SIAI would mean that all those people implementing a decision process sufficiently similar to your own would donate 100% of their money to SIAI. A decision to donate 50% of your money to SIAI and 50% to Charity Option B would imply a similar split for all those people as well.
If there are enough people like this, then the total amount of money involved may be large enough that the linear approximation does not hold. In that case, it seems natural to me to assume that, if both charity options are worthwhile, significantly increasing the successfulness of both charities is more important than increasing SIAI’s successfulness even more significantly. Thus, you would donate 50%/50%.
Overall, the argument you link to seems to me to parallel (though inexactly) the argument that voting is pointless considering how unlikely your vote is to swing the outcome.
Also your errors in choosing a charity won’t necessarily be random. For example, if you trust your reasoning to pick the best three charities, but suspect if you had to pick just one you’d end up influenced by deceptive marketing, bad arguments, or your biases you’d rather not act on, and the same applies to other people, you may be better off not choosing between them, and better off if other people don’t try to choose between them.
In this case, a decision to donate 100% of your money to SIAI would mean that all those people implementing a decision process sufficiently similar to your own would donate 100% of their money to SIAI. A decision to donate 50% of your money to SIAI and 50% to Charity Option B would imply a similar split for all those people as well.
This argument assumes that the people using a similar decision process are faced with the same evidence. In particular, if they made their decision significantly later then they would know about your donation (not directly, but if SIAI now had significantly more funds they could know about it).
If all decision makers were perfectly rational and omniscient, but didn’t have to make their decisions at the same time, then you wouldn’t expect to see the 50⁄50 splitting. You would expect everyone to donate to the charity for which the current marginal usefulness is greatest. In the situation you envision, the marginal usefulness would decrease over time, until eventually donors would notice that it was no longer the best option, and then start diverting their funding. Perhaps once this sort of equilibrium is reached splitting your money is advisable, but we are extremely unlikely to be anywhere near such an equilibrium (with respect to my personal values) unless there is an explicit mechanism pushing us towards it. This would probably require postulating a lot of brilliant rational donors with identical values.
Unless, of course, you believe that the decisions of other people donating to charity are correlated with your own. In this case, a decision to donate 100% of your money to SIAI would mean that all those people implementing a decision process sufficiently similar to your own would donate 100% of their money to SIAI. A decision to donate 50% of your money to SIAI and 50% to Charity Option B would imply a similar split for all those people as well.
If there are enough people like this, then the total amount of money involved may be large enough that the linear approximation does not hold. In that case, it seems natural to me to assume that, if both charity options are worthwhile, significantly increasing the successfulness of both charities is more important than increasing SIAI’s successfulness even more significantly. Thus, you would donate 50%/50%.
Overall, the argument you link to seems to me to parallel (though inexactly) the argument that voting is pointless considering how unlikely your vote is to swing the outcome.
Also your errors in choosing a charity won’t necessarily be random. For example, if you trust your reasoning to pick the best three charities, but suspect if you had to pick just one you’d end up influenced by deceptive marketing, bad arguments, or your biases you’d rather not act on, and the same applies to other people, you may be better off not choosing between them, and better off if other people don’t try to choose between them.
This only applies if people donate simultaneously, which I doubt is the case in practice.
I don’t understand. Could you please clarify?
This argument assumes that the people using a similar decision process are faced with the same evidence. In particular, if they made their decision significantly later then they would know about your donation (not directly, but if SIAI now had significantly more funds they could know about it).
If all decision makers were perfectly rational and omniscient, but didn’t have to make their decisions at the same time, then you wouldn’t expect to see the 50⁄50 splitting. You would expect everyone to donate to the charity for which the current marginal usefulness is greatest. In the situation you envision, the marginal usefulness would decrease over time, until eventually donors would notice that it was no longer the best option, and then start diverting their funding. Perhaps once this sort of equilibrium is reached splitting your money is advisable, but we are extremely unlikely to be anywhere near such an equilibrium (with respect to my personal values) unless there is an explicit mechanism pushing us towards it. This would probably require postulating a lot of brilliant rational donors with identical values.
I’m not keen on it myself, but I’ve seen it linked here (and pushed elsewhere by LessWrong regulars) quite a lot.