The Siren Song of Meta-Altruism

Roughly, the Effective Altruism movement asks the question “how can we do the most good with the resources we have?”. As you can imagine, solving this kind of tough optimization problem in a vast search space that isn’t fully quantifiable is pretty hard. There’s a lot of room for error and pre-existing bias to creep in despite our best efforts to think as clearly as possible.

Really tough problems tend to have answers that take one of two forms. The first is sort of like a mathematical proof that is discovered after many years of research but is itself compact, elegant, and everything we want our proofs to be. The toughness of the problem is that we weren’t sure where to even begin looking for the answer, but once we have an answer it’s often (relatively) easy to verify, and can become the starting point for new research in the next generation.

The other kind of answer to tough problems is messy and complicated. Think of the last political discussion you had with someone you disagreed with—often these problems are so big, and have so many aspects or stakeholders that simple, elegant answers just don’t exist. Instead, we’re forced to find compromises that don’t really satisfy anyone’s preferences completely, which tends to leave the exact boundaries of the compromise open for debate forever. These answers aren’t final, pretty, or elegant, but sometimes they’re necessary.

I believe the problem of “how do we do the most good” has the latter form, which tends to make me skeptical of simplistic or elegant answers. I’d like to talk about one answer in particular, which I’ll call “meta-altruism”.

If altruism is giving to charity, meta-altruism is giving to people who themselves look into various charities and pick which to give to. Like any investment fund manager, they have stakeholders (the original donors), but instead of maximizing a financial return their goal is to maximize the good that is done with what’s in the fund.

In theory, because these fund managers’ occupation is to research charities and allocate funds to them, they will know more than the average charitable donor, and so will make better decisions about what to do with the funds, and produce more good in the world. I largely agree with this theory! If it makes it easier to give, then by all means, give to a meta-charity and have them do the hard work of allocating your donations. I just happen to think this answer is a bit too simplistic for such a complicated problem, and has a few undesirable knock-on effects.

A typical extremely basic investment strategy usually involves some amount of diversification. Even in metacharitable giving, some of this happens naturally because of the hardness of the problem space—there are multiple meta-charities with differing views on what the most important causes are. However, you might expect each individual fund manager not to diversify as much as they would if merely picking stocks. Because the amount of impact to be had from the best charities (i.e. those with a focus area that tends to produce the most good) is so much greater than from others, there’s a tendency to select a few “pet causes” and stand by them.

There’s nothing wrong with this in and of itself. If the expected value from these causes is higher than from a diversified charitable portfolio, then a rational fund manager will allocate funds accordingly. I think this tends to run into trouble at around this level though. What are the most effective causes? Well, certainly being high upstream has a chance to multiply your impact. And what could be higher upstream than…metacharities?

I don’t think this is actually happening to the extent I’m making it sound like, but if we always give to the highest-value causes, and those tend to be meta-causes, then there’s a chance we get stuck in a loop that never bottoms out in actually helping!

This also looks kind of bad to outsiders. When I’m deciding who I should give my donation to, it’s hard to distinguish whether cries of “give it to us, we’ll use it most effectively!” are coming from an extremely effective meta-charity, or an extremely ineffective one. This might cause me to instead give to a less-effective bottom-level charity, even though the effective meta-charity would have made better decisions and amplified my impact immensely.

There’s also the straightforward danger of looking scammy. Saying “give to the most effective organization ever…which happens to be us” doesn’t engender much trust. I think this particular problem is best solved by an ecosystem of truly independent charity evaluators, and I suppose meta-charity evaluators. Coming up with objective groups of people out of thin air is pretty hard though.

I don’t think this overmetaness is a huge problem, and clearly I’d rather people did some amount of good rather than be scared off by having to be the most effective giver ever, and doing nothing. But going one level too meta has some pitfalls that seem to crop up especially often in Effective Altruism. You’d hope EA could avoid these pitfalls, and instead just help people who are struggling with deciding among all the possible ways to give in the first place.


❮ Manifesting Manifolds
Order and chaos at different structural scales
Elm Everywhere ❯
Building Android, iOS, and web apps with shared front-end logic