TL;DR: Vote hereby 15 December. Voting should take 2-10 minutes and you’ll be able to edit your vote until the deadline. (The deadline for donating to the Donation Election Fund is 20 December.)
You can vote if you had a Forum account by October 22, 2023. If you didn’t, you can still share where you’re donating or make the case for voting for some candidates over others.
You’ll be able to edit your vote anytime before December 15.
The voting system is outlined here; your vote should basically just represent how you’d allocate funding between the candidates, and the voting portal will walk you through the process for that.
Should I actually vote in the Donation Election? (I haven’t read all the posts, I’m not that informed, I don’t think it matters that much…)
I think yes, you should vote (if your account was made before October 22 this year). Some reasons for my belief that you should vote (2-5 are most compelling to me):
You’ll influence how funds are distributed — probably in a positive way even if you don’t think you have that much expertise or context.
There are currently ~215 votes. The Donation Election Fund has ~$34,000. So (very very approximately) you’d be affecting ~$150 in donations in expectation.
Additionally, I think you should have a prior that more votes will lead to a better outcome. Aggregation mechanisms generally function better when there are more inputs, so the combined result should improve if you add to it even if you’re not super informed. See e.g. this analysis of Metaculus community predictions, which suggests that the Metaculus community prediction improves approximately logarithmically with the number of forecasters. (See also this post.)[2]
You’ll add useful information to the voting data.
I think the data we’ll get from the Donation Election could be extremely useful; we’ll have a sense for people’s priorities, and we might identify blind spots or important points of disagreement. But it’ll be a lot more useful if more people vote. (Moreover, if you’re not sure about voting, you might be part of a group that’s less likely to vote, which will be underrepresented in the information we’ll get.)
Voting will prompt you to think a bit more about your donation choices.
You might waste some time (but you can time-cap yourself). If you’re worried about mis-directing funds, you can view this as a fairly low-cost exercise. You’ll affect more than a trivial amount of money (unless we get tons of votes), but it won’t be that big; I think the second-order effects (2-5) will outweigh the effect of directing funds (1).
Some other common questions
Why isn’t charity X a candidate?
We restricted candidates to the charities on this list, largely for logistical reasons and vetting capacity, which stopped some charities from being candidates (we might change this next year if we run an election again). If the charity is on that list, it’s not a candidate because nobody nominated it on time. Consider sharing a post about why people should donate to the charity, though!
Why can’t people whose accounts are newer than October 22, 2023 vote?
We added this restriction to prevent vote manipulation (we’ll also be checking in other ways).
Quadratic voting seems good when you want to discourage people from voting entirely for their best bet/when you want to make sure that people fund collective goods.[3] This isn’t really relevant here (we’re doing something closer to aggregating beliefs), so there wasn’t a strong reason to go with quadratic voting as a ~default (especially since it’s not the easiest to use). Quadratic voting is also unfortunately susceptible to collusion.[4] I also wanted to make sure that votes for less-popular charities wouldn’t be wasted (so voters wouldn’t worry about this or avoid voting for charities they thought would be less popular), and we were worried about how this would interact with quadratic voting.
There are things I’d change,[5] but I don’t currently think we should switch to quadratic voting (or approval voting, for that matter [6]) if we do it again.
If you disagree, let me know! This estimate is based on the responses to my Quick Take question, loosely adjusted for the fact that people who vote fast are probably likelier to be the kind of people who would see and respond to that Quick Take.
Suppose there was a collective good that would generate some collective benefit X for some total cost Y, with X=Y*c (the value of the good scales ~linearly with the amount of funding the collective good gets). If the collective good were fully funded, every individual would get x = X/N benefit, and would have to pay y = Y/N (if the cost were distributed evenly).
If X > Y, then it is better for everyone if each individual pays the cost y, but there is often an individual incentive to defect and not pay your cost y; if you don’t pay your y, then you only lose x/N in benefit (instead of X/N; the value of the good without your support is Y*(N-1)=X-X/N = X-x, and that’s spread out across N people). So groups of people will often fail to pay anything towards collective goods. Quadratic voting fixes this by effectively boosting the value of the first few dollars you put towards something (so even if you are spending most of your dollars on other things, there is an incentive to put at least a small amount towards collective goods because this gets multiplied up).
I’m really grateful to @Will Howard for initially pointing out that this actually doesn’t really apply in our case (we’re not really paying for services, including collective benefit services, but rather acting as ~funders), and most of the explanation above is due to him (although errors are my own).
If I really want Charity A to win and my friend really wants Charity B to win, our votes will count for more if we agree to vote 50-50 each (and this is even worse at scale).
The main thing I’d change about the voting system is (probably) dropping the constraint that we’re designating funding to the top three charities specifically (mostly because (1) I think it would be nicer if people who disagree with the top three charities see that their vote for a different charity still led to some funding going to that charity, and (2) it was harder to find a voting system that didn’t have weird discontinuities with the 3-charities constraint). (We announced the fact that we’d restrict to 3 charities and opened the Donation Election Fund before realizing that this was somewhat inconvenient, and I didn’t want to change things like that after people had started donating — I was worried that donors would feel misled.) I’m not sure about this, though. One downside of designating funding to all candidates in proportion to the vote (instead of just the top N candidates) is that we’d probably end up transferring tiny amounts of funding to the lower-ranking charities, and the overhead costs might be relatively quite high.
I could see us changing to ranked-choice voting instead of our system (or at least giving people the option to use ranked-choice voting, which would be practically equivalent to them assigning their top-ranked charities points A, B, C, etc., such that A >> B >> C, etc. (e.g. 1000,000; 1,000; 1; 0.001; etc.). This would have the benefit of being clearer for people, since ranked choice voting is a familiar voting system (which is a nontrivial benefit!). But it would potentially reduce the information we’d get from people’s votes (I’m interested in seeing if lots of people think that Charity A and Charity B are basically similar, or the actual relative values people assign to donations to Charity A and B), and might be less useful for voters as an exercise, since voters wouldn’t have to think about the true relative value of donations to different charities. (But I think at least one person on the Online Team thinks we should switch to RCV if we do this again.)
I think I personally actually quite like our method (except that it’s confusing to some — I don’t have a strong sense of how strong a negative this is and will try to find out). I like that it basically encourages you to vote in a way that pretty closely represents how you’d actually allocate funding (except for also encouraging you to give non-zero scores to some charities you wouldn’t donate anything at all to, if you have opinions between them — although I think you should probably only score candidates if you’ve thought about them a decent amount, as otherwise you’re kind of adding noise).
Ultimately we’ll see what happens with this voting system, and do some analysis (like looking at how different the results would be if we had used ranked-choice, etc.) to see how robust and useful all of this is, and I think we’ll have better opinions afterwards.
We wanted people’s votes to matter even if they’re in a (predictable) minority group relative to other voters; approval voting would mean that if you’re pretty sure few would boost your favorite charity, there’s often no real point voting for that charity. Although I think approval voting might work decently well if we remove the 3-winning-candidates limit (and get a lot of votes).
Vote in the Donation Election by 15 December
TL;DR: Vote here by 15 December. Voting should take 2-10 minutes and you’ll be able to edit your vote until the deadline. (The deadline for donating to the Donation Election Fund is 20 December.)
You can vote if you had a Forum account by October 22, 2023. If you didn’t, you can still share where you’re donating or make the case for voting for some candidates over others.
🗳️ Vote in the Donation Election 2023
📖 Read about the candidates
More context:
The Donation Election Fund (currently around $34,000) will be designated for the top three candidates in the Donation Election (in proportion to the vote). You can read about the candidates here.
How long does voting take, and how does it work?
Voting should take around 2-10 minutes.[1]
You’ll be able to edit your vote anytime before December 15.
The voting system is outlined here; your vote should basically just represent how you’d allocate funding between the candidates, and the voting portal will walk you through the process for that.
Should I actually vote in the Donation Election? (I haven’t read all the posts, I’m not that informed, I don’t think it matters that much…)
I think yes, you should vote (if your account was made before October 22 this year). Some reasons for my belief that you should vote (2-5 are most compelling to me):
You’ll influence how funds are distributed — probably in a positive way even if you don’t think you have that much expertise or context.
There are currently ~215 votes. The Donation Election Fund has ~$34,000. So (very very approximately) you’d be affecting ~$150 in donations in expectation.
Additionally, I think you should have a prior that more votes will lead to a better outcome. Aggregation mechanisms generally function better when there are more inputs, so the combined result should improve if you add to it even if you’re not super informed. See e.g. this analysis of Metaculus community predictions, which suggests that the Metaculus community prediction improves approximately logarithmically with the number of forecasters. (See also this post.)[2]
You’ll add useful information to the voting data.
I think the data we’ll get from the Donation Election could be extremely useful; we’ll have a sense for people’s priorities, and we might identify blind spots or important points of disagreement. But it’ll be a lot more useful if more people vote. (Moreover, if you’re not sure about voting, you might be part of a group that’s less likely to vote, which will be underrepresented in the information we’ll get.)
Voting will prompt you to think a bit more about your donation choices.
It might be fun for you.
Voting is a public/collective good of sorts, and you might value being the kind of person who contributes to public goods.
The downside is limited.
You might waste some time (but you can time-cap yourself). If you’re worried about mis-directing funds, you can view this as a fairly low-cost exercise. You’ll affect more than a trivial amount of money (unless we get tons of votes), but it won’t be that big; I think the second-order effects (2-5) will outweigh the effect of directing funds (1).
Some other common questions
Why isn’t charity X a candidate?
We restricted candidates to the charities on this list, largely for logistical reasons and vetting capacity, which stopped some charities from being candidates (we might change this next year if we run an election again). If the charity is on that list, it’s not a candidate because nobody nominated it on time. Consider sharing a post about why people should donate to the charity, though!
Why can’t people whose accounts are newer than October 22, 2023 vote?
We added this restriction to prevent vote manipulation (we’ll also be checking in other ways).
Why wasn’t quadratic voting (/funding) chosen as the voting system?
Quadratic voting seems good when you want to discourage people from voting entirely for their best bet/when you want to make sure that people fund collective goods.[3] This isn’t really relevant here (we’re doing something closer to aggregating beliefs), so there wasn’t a strong reason to go with quadratic voting as a ~default (especially since it’s not the easiest to use). Quadratic voting is also unfortunately susceptible to collusion.[4] I also wanted to make sure that votes for less-popular charities wouldn’t be wasted (so voters wouldn’t worry about this or avoid voting for charities they thought would be less popular), and we were worried about how this would interact with quadratic voting.
There are things I’d change,[5] but I don’t currently think we should switch to quadratic voting (or approval voting, for that matter [6]) if we do it again.
What else can I do?[7]
Donate directly to different projects (explore recommendations from Giving What We Can, GiveWell, Animal Charity Evaluators, and more), share where you’re donating (and why),
Can I share feedback? (How?)
Yes, please! You can share feedback on the final page of the voting portal (right before you submit your vote), by commenting here, or via this form.
If you disagree, let me know! This estimate is based on the responses to my Quick Take question, loosely adjusted for the fact that people who vote fast are probably likelier to be the kind of people who would see and respond to that Quick Take.
(I’m interested in thoughts on how well this applies here; I think it’s fairly comparable but haven’t thought about it very carefully.)
Suppose there was a collective good that would generate some collective benefit X for some total cost Y, with X=Y*c (the value of the good scales ~linearly with the amount of funding the collective good gets). If the collective good were fully funded, every individual would get x = X/N benefit, and would have to pay y = Y/N (if the cost were distributed evenly).
If X > Y, then it is better for everyone if each individual pays the cost y, but there is often an individual incentive to defect and not pay your cost y; if you don’t pay your y, then you only lose x/N in benefit (instead of X/N; the value of the good without your support is Y*(N-1)=X-X/N = X-x, and that’s spread out across N people). So groups of people will often fail to pay anything towards collective goods. Quadratic voting fixes this by effectively boosting the value of the first few dollars you put towards something (so even if you are spending most of your dollars on other things, there is an incentive to put at least a small amount towards collective goods because this gets multiplied up).
I’m really grateful to @Will Howard for initially pointing out that this actually doesn’t really apply in our case (we’re not really paying for services, including collective benefit services, but rather acting as ~funders), and most of the explanation above is due to him (although errors are my own).
If I really want Charity A to win and my friend really wants Charity B to win, our votes will count for more if we agree to vote 50-50 each (and this is even worse at scale).
The main thing I’d change about the voting system is (probably) dropping the constraint that we’re designating funding to the top three charities specifically (mostly because (1) I think it would be nicer if people who disagree with the top three charities see that their vote for a different charity still led to some funding going to that charity, and (2) it was harder to find a voting system that didn’t have weird discontinuities with the 3-charities constraint). (We announced the fact that we’d restrict to 3 charities and opened the Donation Election Fund before realizing that this was somewhat inconvenient, and I didn’t want to change things like that after people had started donating — I was worried that donors would feel misled.) I’m not sure about this, though. One downside of designating funding to all candidates in proportion to the vote (instead of just the top N candidates) is that we’d probably end up transferring tiny amounts of funding to the lower-ranking charities, and the overhead costs might be relatively quite high.
I could see us changing to ranked-choice voting instead of our system (or at least giving people the option to use ranked-choice voting, which would be practically equivalent to them assigning their top-ranked charities points A, B, C, etc., such that A >> B >> C, etc. (e.g. 1000,000; 1,000; 1; 0.001; etc.). This would have the benefit of being clearer for people, since ranked choice voting is a familiar voting system (which is a nontrivial benefit!). But it would potentially reduce the information we’d get from people’s votes (I’m interested in seeing if lots of people think that Charity A and Charity B are basically similar, or the actual relative values people assign to donations to Charity A and B), and might be less useful for voters as an exercise, since voters wouldn’t have to think about the true relative value of donations to different charities. (But I think at least one person on the Online Team thinks we should switch to RCV if we do this again.)
I think I personally actually quite like our method (except that it’s confusing to some — I don’t have a strong sense of how strong a negative this is and will try to find out). I like that it basically encourages you to vote in a way that pretty closely represents how you’d actually allocate funding (except for also encouraging you to give non-zero scores to some charities you wouldn’t donate anything at all to, if you have opinions between them — although I think you should probably only score candidates if you’ve thought about them a decent amount, as otherwise you’re kind of adding noise).
Ultimately we’ll see what happens with this voting system, and do some analysis (like looking at how different the results would be if we had used ranked-choice, etc.) to see how robust and useful all of this is, and I think we’ll have better opinions afterwards.
We wanted people’s votes to matter even if they’re in a (predictable) minority group relative to other voters; approval voting would mean that if you’re pretty sure few would boost your favorite charity, there’s often no real point voting for that charity. Although I think approval voting might work decently well if we remove the 3-winning-candidates limit (and get a lot of votes).
This isn’t really a common question, but I wanted to include it anyway.