Thanks for posting these updates, I’m quite excited about the project!
Have you considered incentive problems stemming from the fact that you require fractions of impact to be allocated among participants so that they add up to 1? My understanding is that this way of allocating credit doesn’t produce the desired results in cases where the project wouldn’t have happened without all participants (see e.g. 5 mistakes of moral reasoning).
If you’ve already answered this, I’d appreciate a link—I know you’ve thought about this quite a bit.
Impact purchases incorrectly value good deeds based on the highest bidder, instead of summing over all people. They are supposed to work correctly on the supply side, but not the demand side. This is a complicated issue I may discuss later. In order to get a correct protocol you need to combine them with another idea. For the rest of the post I am going to assume that everyone has the same values, which makes this issue go away. Note that this issue is similar for normal donations.
If there are increasing or diminishing returns to scale, the sum of people’s marginal contributions don’t add up to 1. The simplest case is when output = (sum of inputs)^x for some x other than 1.
If there are decreasing returns to scale, then there are rents: the sum of the marginal outputs adds up to less than the total output, and so there is some extra value to be captured. Certificate purchases work fine in that case—each contributor can unilaterally claim their impact, or the group can claim its impact and decide how to split the rent.
Increasing returns are more complicated. It’s still fine to pay a project for its total impact—there is guaranteed to be some way of assigning that impact which would incentivize people to do the project (othertise they should have all done their second-best option, and they would have produced more value that way). Our approach is to group tightly complementary contributions together and let them negotiate a solution. This is the same thing that we normally do in the broader market. Philanthropy normally dodges the issue by just not thinking about it.
Note that the naive strategy of just paying each person for their marginal contribution would also go wrong.
For example, suppose that there are two people, who can each contribute up to 1 unit of effort in a project that creates E^2 value, where E is the amount of effort. Each person can also use their unit of effort to create 2.5 units of value.
It’s easy to see that each person’s marginal contribution is 4-1 = 3, but in fact they would be better served by both doing the 2.5 thing.
I think that a sufficiently careful accounting of marginal impacts would work, but that is a more complicated issue and I don’t have a link to a good account.
Thanks! This reply makes sense to me, and the refutation of the marginal-contribution strategy is interesting. I can see why you’ve chosen to group tightly complementary contributions.
Thanks for posting these updates, I’m quite excited about the project!
Have you considered incentive problems stemming from the fact that you require fractions of impact to be allocated among participants so that they add up to 1? My understanding is that this way of allocating credit doesn’t produce the desired results in cases where the project wouldn’t have happened without all participants (see e.g. 5 mistakes of moral reasoning).
If you’ve already answered this, I’d appreciate a link—I know you’ve thought about this quite a bit.
Discussed briefly here.
There are two things going on in these cases:
Impact purchases incorrectly value good deeds based on the highest bidder, instead of summing over all people. They are supposed to work correctly on the supply side, but not the demand side. This is a complicated issue I may discuss later. In order to get a correct protocol you need to combine them with another idea. For the rest of the post I am going to assume that everyone has the same values, which makes this issue go away. Note that this issue is similar for normal donations.
If there are increasing or diminishing returns to scale, the sum of people’s marginal contributions don’t add up to 1. The simplest case is when output = (sum of inputs)^x for some x other than 1.
If there are decreasing returns to scale, then there are rents: the sum of the marginal outputs adds up to less than the total output, and so there is some extra value to be captured. Certificate purchases work fine in that case—each contributor can unilaterally claim their impact, or the group can claim its impact and decide how to split the rent.
Increasing returns are more complicated. It’s still fine to pay a project for its total impact—there is guaranteed to be some way of assigning that impact which would incentivize people to do the project (othertise they should have all done their second-best option, and they would have produced more value that way). Our approach is to group tightly complementary contributions together and let them negotiate a solution. This is the same thing that we normally do in the broader market. Philanthropy normally dodges the issue by just not thinking about it.
Note that the naive strategy of just paying each person for their marginal contribution would also go wrong.
For example, suppose that there are two people, who can each contribute up to 1 unit of effort in a project that creates E^2 value, where E is the amount of effort. Each person can also use their unit of effort to create 2.5 units of value.
It’s easy to see that each person’s marginal contribution is 4-1 = 3, but in fact they would be better served by both doing the 2.5 thing.
I think that a sufficiently careful accounting of marginal impacts would work, but that is a more complicated issue and I don’t have a link to a good account.
Thanks! This reply makes sense to me, and the refutation of the marginal-contribution strategy is interesting. I can see why you’ve chosen to group tightly complementary contributions.