Does any of this sound familiar? You want to donate as effectively as possible, but you are worried about making the wrong choice. Should it be malaria nets, animal welfare, maybe x-risk? Also, how can you determine which charity evaluators are worthy of your trust? If you actually took the time to do an analysis, you could figure it out. But life gets in the way and you keep putting it off.
That is where I was at! But as it turns out, there already exists an extremely elegant solution, which is basically a simple but clever application of decision theory. I just gave $6,000 to a donor lottery and called it a day! In particular, I donated to this lottery (announced here), which is still open to entries until January 8th.
Under very conservative assumptions, donor lotteries allow you to minimize the amount of thinking required (in expectation) to give efficiently. In other words, it is a method to avoid overthinking it!
Why does this work?
If you are not already familiar with donor lotteries, here is the premise of what I did. By donating $6,000, there is a 97% chance that nothing will happen, and a 3% chance that I will get to direct $200,000 to be donated anywhere I choose. So, there is only a 3% chance I will need to carefully think about where my donation will be directed! But as long as impact is roughly linear in the amount of money donated, a 3% chance of donating $200,000 is equivalent to a 100% chance of donating $6,000 to the same cause.
Does is matter what the other donors might do if they win? Nope! If it weren’t for the fees and house advantage, I could have just gone to a casino with $6,000 trying to turn it into $200,000, and the expected utility would’ve worked out the same.
Anyways, the top GiveWell charitiesallegedly save a human life for about $5,000. So if I win and donate to one of those and GiveWell is correct, I save 1.2 lives in expectation. However, there is a small chance that (1) GiveWell is secretly tricking me and there is no way to save human life for $5,000 or (2) I decide that a donation for something other than saving human lives is more worthwhile (saving animal lives, improving human happiness, etc...). Give the unlikelihood of these, I think we can conservatively say that, apriori, the human lives saved in expectation is at least 1. (By conservative, I mean that I expect most readers to arrive at an expected value of at least 1, even if they have different evidence or priors than me.)
Who’s to say you can’t save lives for cheaper than $5,000 though? Would I then be making the wrong choice? Nope! The GiveWell value if just a lower bound. In the 3% of chances that I win, I will carefully think if I can do better. If so, I will save more than 1 life in expectation!
Even if I lose the lottery, my money is probably used to save a life
There is a 97% chance that I am giving $6,000 to another donor. Since it is pretty likely that the other donor is an honest EA, this money probably goes to some org that uses it to save someone’s life.
Note that this has no bearing on the expected utility calculation (since there is a 3% chance that I take $194,000 from other donors, which exactly cancels), but it is still nice to know.
Why not just use a charity evaluator ran by smarter people to think for me?
So, in practice if I win this is my strategy. However, it will still require thinking to choose the correct charity evaluator. In particular, I will need to think about how trustworthy they are. For example, is there a conflict of interest with the charities they are recommending?
I think GiveWell is fairly trustworthy, but (1) I still want to double check this and more importantly (2) they only compare health interventions. There are people presumably smarter than me who have tried comparing all charities, but I would need to do research to determine who is most trustworthy.
There is also some question of different values, such as weighing human welfare v.s. animal welfare. Although I can study what moral philosophers think, this might end up being a subjective call I get to make.
What about trusting the donor lottery itself? Well, the way it is setup requires much less trust. In particular, my lottery ticket is tied to the NIST randomness beacon.
So the only way I could get scammed is one of the following:
One of the other donors colludes with or hacks the U.S. government to compromise the randomness beacon.
I win the $200,000 according to the beacon, which is easily provable, but then Giving What We Can destroys its reputation and risks legal action by not following through.
Although this is possible, this is a much lower level of trust than “charity evaluator could give subtly biased evaluations, perhaps due to unconscious personal bias”.
In any case, donor lotteries have been endorsed by experts.
...we believe that a donor lottery is the most effective way for most smaller donors to give the majority of their donations, for those who feel comfortable with it. - “Why you should give to a donor lottery this Giving Season” EA Funds Blog
Although it isn’t as good of a headline, saving a fraction of a life in expectation would still be impressive. Saving 0.1 lives in expectation is the same as saving one person from a 10% chance of death, or ten people working together to save one person, for example. And smaller donations lead to a smaller probability of needing to think!
Here’s a fun calculation. I estimate that wearing your seatbelt for a year saves you about 150 micromorts from car accidents (averaged globally). If we go with my above estimate that donating $6,000 to a donor lottery equates to 1 human life = 1,000,000 micromorts, then donating $1 to a donor lottery equates to about 165 micromorts.
Thus, as a more honest version of a common charity cliché, we could say the following:
By donating literally just ONE DOLLAR to a donor lottery today, youmightsave 1 human life!(In the same sense that, by wearing your seatbelt for a year, you might save your own life.)
The lowest donation in the lottery I linked above (at the time of this writing) is a measly $9.48, which is about 1,600 micromorts if my calculations are correct. Best of luck to them!
How I saved 1 human life (in expectation) without overthinking it
Does any of this sound familiar? You want to donate as effectively as possible, but you are worried about making the wrong choice. Should it be malaria nets, animal welfare, maybe x-risk? Also, how can you determine which charity evaluators are worthy of your trust? If you actually took the time to do an analysis, you could figure it out. But life gets in the way and you keep putting it off.
That is where I was at! But as it turns out, there already exists an extremely elegant solution, which is basically a simple but clever application of decision theory. I just gave $6,000 to a donor lottery and called it a day! In particular, I donated to this lottery (announced here), which is still open to entries until January 8th.
Under very conservative assumptions, donor lotteries allow you to minimize the amount of thinking required (in expectation) to give efficiently. In other words, it is a method to avoid overthinking it!
Why does this work?
If you are not already familiar with donor lotteries, here is the premise of what I did. By donating $6,000, there is a 97% chance that nothing will happen, and a 3% chance that I will get to direct $200,000 to be donated anywhere I choose. So, there is only a 3% chance I will need to carefully think about where my donation will be directed! But as long as impact is roughly linear in the amount of money donated, a 3% chance of donating $200,000 is equivalent to a 100% chance of donating $6,000 to the same cause.
Does is matter what the other donors might do if they win? Nope! If it weren’t for the fees and house advantage, I could have just gone to a casino with $6,000 trying to turn it into $200,000, and the expected utility would’ve worked out the same.
Anyways, the top GiveWell charities allegedly save a human life for about $5,000. So if I win and donate to one of those and GiveWell is correct, I save 1.2 lives in expectation. However, there is a small chance that (1) GiveWell is secretly tricking me and there is no way to save human life for $5,000 or (2) I decide that a donation for something other than saving human lives is more worthwhile (saving animal lives, improving human happiness, etc...). Give the unlikelihood of these, I think we can conservatively say that, apriori, the human lives saved in expectation is at least 1. (By conservative, I mean that I expect most readers to arrive at an expected value of at least 1, even if they have different evidence or priors than me.)
Who’s to say you can’t save lives for cheaper than $5,000 though? Would I then be making the wrong choice? Nope! The GiveWell value if just a lower bound. In the 3% of chances that I win, I will carefully think if I can do better. If so, I will save more than 1 life in expectation!
Even if I lose the lottery, my money is probably used to save a life
There is a 97% chance that I am giving $6,000 to another donor. Since it is pretty likely that the other donor is an honest EA, this money probably goes to some org that uses it to save someone’s life.
Note that this has no bearing on the expected utility calculation (since there is a 3% chance that I take $194,000 from other donors, which exactly cancels), but it is still nice to know.
Why not just use a charity evaluator ran by smarter people to think for me?
So, in practice if I win this is my strategy. However, it will still require thinking to choose the correct charity evaluator. In particular, I will need to think about how trustworthy they are. For example, is there a conflict of interest with the charities they are recommending?
I think GiveWell is fairly trustworthy, but (1) I still want to double check this and more importantly (2) they only compare health interventions. There are people presumably smarter than me who have tried comparing all charities, but I would need to do research to determine who is most trustworthy.
There is also some question of different values, such as weighing human welfare v.s. animal welfare. Although I can study what moral philosophers think, this might end up being a subjective call I get to make.
What about trusting the donor lottery itself? Well, the way it is setup requires much less trust. In particular, my lottery ticket is tied to the NIST randomness beacon.
So the only way I could get scammed is one of the following:
One of the other donors colludes with or hacks the U.S. government to compromise the randomness beacon.
I win the $200,000 according to the beacon, which is easily provable, but then Giving What We Can destroys its reputation and risks legal action by not following through.
Although this is possible, this is a much lower level of trust than “charity evaluator could give subtly biased evaluations, perhaps due to unconscious personal bias”.
In any case, donor lotteries have been endorsed by experts.
Also see Deference Culture in EA.
Saving a fraction of life
Although it isn’t as good of a headline, saving a fraction of a life in expectation would still be impressive. Saving 0.1 lives in expectation is the same as saving one person from a 10% chance of death, or ten people working together to save one person, for example. And smaller donations lead to a smaller probability of needing to think!
Here’s a fun calculation. I estimate that wearing your seatbelt for a year saves you about 150 micromorts from car accidents (averaged globally). If we go with my above estimate that donating $6,000 to a donor lottery equates to 1 human life = 1,000,000 micromorts, then donating $1 to a donor lottery equates to about 165 micromorts.
Thus, as a more honest version of a common charity cliché, we could say the following:
The lowest donation in the lottery I linked above (at the time of this writing) is a measly $9.48, which is about 1,600 micromorts if my calculations are correct. Best of luck to them!