by banx
4 min read

# 7

How is the world different if I give X dollars to the Against Malaria Foundation, where X < 100,000 USD? (In reality, giving directly to GiveWell is probably the better choice. But using AMF works well for this question.)

Here's the answer that would make me very motivated to give:

If I don't give, AMF will purchase and cause to be distributed N bed nets in the next 5 years. If I do give, AMF will purchase and cause to be distributed N+K bed nets in the next 5 years, where K is roughly X / (cost of a bed net, taking into account administrative costs, etc).

In reality, I think the answer is more like:

If I give, AMF will have X more dollars in its bank account than if I don't give. It will make decisions on whether or not to move forward with various large net distribution projects based on the amount of money it has available at the time of each decision. For each decision there will be some amount Z such that AMF will choose to go forward with the project if and only if it has at least Z dollars in the bank, holding all other factors equal. (Even if they claim there's such a cut-off point Y, but would still go forward with the project if they had slightly less than Y, there's an actual amount Z that is the real cut-off point.)

I can think of two kinds of these distribution decisions, which interact differently with this extra money. First, if AMF is presented a unique, time-sensitive distribution opportunity, the extra money slightly increases the chance that they have enough to move forward with the distribution. Second, suppose there are recurring opportunities that are mostly the same. Assuming that AMF has a steady income of donations, there is a small chance this money will cause AMF to have enough money to fund the K-1th distribution instead of the Kth distribution, thereby getting bed nets out to people who need them sooner. This one-time time advantage would not "use up" the extra money. Assuming the K-1th distribution occurred but the Kth did not, at the time of the K+1th, AMF would still have X more dollars than if I hadn't given. So the chance of the time-advantage happening again (and the chance of a unique opportunity being funded) still exists.

I'm fairly confused about the above, and haven't succeeded in rigorously defining a satisfying model from which I can compute the expected utility of a donation. My intuition is that the expected value of a donation should work out to the good done by the number of nets that can be purchased and distributed with that amount of money, but I can't justify it. What is important is that I don't think you can just say that AMF will eventually spend that money on nets in that way, and so that is the good done by the donation. It seems that that argument only works if AMF eventually spends all of its money on distributing nets and then ceases to exist. While this is an ideal end to AMF -- that they've solved the problem they were created to solve -- it doesn't seem like it's definitely going to happen. Even if it did, it's very possible that by the time they are spending their last dollars the marginal cost of saving lives will have gone up significantly compared to what it is today (either because the cost of purchase and distribution has increased or, probably more likely, because the strongest opportunities for doing good with bed nets are no longer available).

In the past, I was very motivated to give because I had the impression that for every five-ish dollars I gave to AMF, an extra bed net would be distributed. It seems that this isn't correct, and this year I'm not nearly as motivated. I will likely be mailing a check to GiveWell tomorrow anyway (I have an employer match that expires at the end of the year), but I would like to have a better understanding of what I'm doing and if it makes sense. I have a sense that I'm not really accomplishing anything, and could spend my money better by giving to a small organization that spends money much less efficiently but would definitely get more done if it had the money, or, alternatively, by buying myself a car. I think this sense is confused, but the fact that I can't articulate exactly how makes it hard to dismiss. I think the ideas in this post more-or-less apply whether I'm considering a gift to AMF, GiveWell, or MIRI. AMF makes things easier to think about because they're mostly focused on only one thing.

Along with any other feedback, I'd love some help with these questions:

1. Does it make sense to be asking this question (How is the world in which I give different from the world in which I don't)? Is the answer to that question not important for some reason? Should I be satisfied with just doing my share in causing AMF to have the money it needs?

2. Can the model of how an organization like AMF actually benefits from small amounts of marginal dollars be made more precise, to the point that I can roughly compute the expected utility (in terms of whatever) of a donation?

3. Is it at all reasonable to prefer using my money on something that has a reasonably high likelihood of succeeding compared to something that has a very small chance of resulting in any desired outcome, when the thing with the tiny likelihood of success has (based on calculations using my assessments of utility) a higher expected utility? It feels really bad to throw tens of thousands of dollars at something, when I know that there's a very tiny chance that it will accomplish anything. This problem is a lot easier to deal with when I expect to iterate enough times that I expect to eventually hit the low-chance outcome. But when I don't expect to iterate nearly enough times for that to happen, it's hard to feel good about the choice. I think the answer may be that the community of givers iterates sufficiently many times that the low-chance outcomes are hit. What if that weren't true? Does it matter? I've asked a similar question in an open thread once, and I got some helpful answers. I think I need to work on truly accepting and internalizing that tiny probabilities of really good things are worth caring about.

Thanks for reading.

# 7

New Comment
7 comments, sorted by Click to highlight new comments since:

If you gave AMF a million dollars, they would distribute an extra 200000 bed nets. If you gave them five dollars 200000 times, the total effect would be an extra 200000 bed nets. The average effect would be one bed net. If you don't know AMF's budget, and you can only guess to within a million dollars, then the expected effect of a five-dollar donation would be the average of all those possible five-dollar donations, which is one bed net.

I'm fairly confused about the above, and haven't succeeded in rigorously defining a satisfying model from which I can compute the expected utility of a donation. My intuition is that the expected value of a donation should work out to the good done by the number of nets that can be purchased and distributed with that amount of money, but I can't justify it.

Suppose they distribute bed nets in batches of D, and your donation can buy K < D nets. There is a probability of K/D that your donation triggers another batch of D nets, hence expected number of nets = (K/D)*D = K. (By a slightly longer argument, this is still true even when K >= D.)

Note that by the same reasoning, cutting 1 minute off your journey to catch a train, when you have no knowledge of the train timetable, has an expected saving of 1 minute on your whole journey, regardless of how frequently the trains run.

But your calculation changes if you have even a small amount of information about the timetables. Can we get this information from the organizations in question?

You can figure out things in the short term that way, but if you want to know if the marginal money will be used in a year, you'll have to know every project they run in the mean time and how much each will cost. This will include projects that haven't even been planned yet.

This isn't a train that doesn't hold close to its time table. This is a train that leaves when it seems like a good idea. And you're trying to predict the time to within a few minutes weeks ahead of time. You won't even be able to predict which train you'd catch.

[-]Elo00

Consider alternative forms of donations. Giving time; volunteering locally can mean that you personally see the effects and needs of people. I believe that money going into someone's big bank account is reasonably less provably effective to me than personally seeing the need and attempting to provide for it.

Consider visiting a local shelter and seeing how they are doing. They could probably use an extra \$100k; but even if you gave it away; you couldn't see how it was used unless you stayed to watch it be used.

[-]lmm00

1/2. It makes sense to ask the question, though as others have said, the naive answer comes out correct. Money is fungible, that's the whole point of money, so if you assume the organization's internal operations are efficient then your intuitive answer will always be the right one (if "your" \$5 was spent on repairing the employee coffee machine, then that will be because the organization believed that repairing the coffee machine would improve their working efficiency enough to do \$5 worth of extra net-distribution).

3|. Open question. Many on this site would say it's worthwhile; my instinct is it isn't. We can't judge moral questions by anything other than our own moral instincts (see The Moral Void). It smells Pascal's-mugging-ey to me.

[-][anonymous]-10

Their goal is not the distribution of net beds. Their goal is the distribution of net beds which are then used. So obvious that you didn't mention it, but without mentioning it any contribution yields a net bed and nobody in it. Thus worse than no contribution at all. Since you must consider time frames longer than 'distribution of net bed' you now have to ask when you draw the line. Distribution then use then stop? Distribution then use then research into outcome of use then stop? Where you draw the line is your choice. If you draw the line at the right place you can have any outcome you like in an infinite universe. Distribution of net bed then use then 100,000 generations later a distant relative of a sleeper in a net bed did a great deed. Or an ill deed.

Which is a longer way to say it matters as much as and in the way you want it to who you donate to in an infinite universe.

Further, you are in 'the world' that will change if you make a donation. If you are very moved by a donation then the donation size might matter less. If you donate much but don't care then 'the world' is changed in that way.