Which cause area is best?

Ranking the effective altruist cause areas

I think this is an important article, so I’d appreciate if you’d like and restack it.

Effective altruists advocate giving your money to where it does most good. The leading areas EAs recommend donating to are making the long-term future go better, animal welfare, global health, and growing the effective altruism movement. Yet there’s been surprisingly little comparison between these cause areas. Here, I’ll explain my ranking: why I think it’s best to donate to Longtermist organizations or organizations focused on growing EA. I’ll also explain why I think animal welfare is better to donate to than global health.

Comparing cause areas is important. I sometimes find there’s a weird tendency among EAs to throw out cost-effectiveness when analyzing different cause areas—to treat all causes as equal, and only look at effectiveness within a cause area. This is wrongheaded. If we care about doing the most good, then we should primarily give to the best cause area. The moral imperative towards cost-effectiveness doesn’t go away just because all the cause areas have a generic EA stamp of approval.

In general, I’ll assume you’re broadly on board with the basic assumptions behind each cause area. So, for instance, I won’t bother arguing for Longtermism or against the idea that animal torture isn’t a big deal because they don’t have a rational nature and can’t solve differential equations. That’s because this article is primarily about trying to evaluate what those convinced by the central EA ideas should do, rather than to convince people that these ideas are right. I will, however, link to places where I’ve defended the cause areas in more detail and provide brief summaries of the essential arguments.

TLDR: my ranking for donations: Longtermist political donations>growing EA>other Longtermist donations>animal welfare>global health.

My ranking for careers: Growing EA=Longtermist careers>animal welfare>global health.

1 Longtermism

1.1 The case for Longtermist interventions being best

The case for Longtermist organizations being best is pretty straightforward. The future could be huge. One back-of-the-envelope calculation guessed there could be around 10^58 future well-off people—and another calculation was 20 orders of magnitude greater (see here and here for additional reasons to think this is conservative). There are a number of ways we could screw up the future. Experts consistently think there’s a several percent chance that misaligned AI kills everyone. Despite this, there are only about 1,000 people working on it. Some other crucial cause areas, like preventing AI-enabled coups and improving AI character, have just a few people working on them.

If this is right, Longtermist activities surpass everything else in value. To see this, let’s be generous and say the expected number of future well-off people added by preventing an existential catastrophe is only 10^40—meaning odds are only 1/10^18 that anything like the Bostrom calculation is right. Now, let’s say that ten billion dollars would reduce existential risks by .01%. This strikes me as very conservative. Doing so is enough to hire 10,000 people, which would massively increase the number of people working on the biggest issues.

Note: this money likely wouldn’t directly go into hiring 10,000 extra people. Longtermism is mostly talent-constrained—there aren’t tons of projects that Longtermist organizations want to fund but can’t for lack of funds. But still, my guess is that across time, this is a conservative lower bound on the effectiveness of Longtermist donations (especially given that it’s likely that there will soon be projects that Longtermists can productively pour lots of money into). If you find these numbers too high, shave off a few zeroes however you see fit. Though remember that we already shaved off eighteen zeroes just for funzies, and that is surely enough to cover!

But even by these very conservative estimates, each dollar then lowers the odds of extinction by one in one hundred trillion. Thus, each dollar increases the number of well-off future people in expectation by 10^26. That’s more than all the people who have ever lived. And remember, overall these assumptions are outrageously conservative.

This might seem surprising. But when you take note of how big the future might be, even tiny effects on it might be enormously impactful. See here for my broader defense of Longtermism.

1.2 Longtermist jobs>>Longtermist donations

Thus, it seems pretty clear that Longtermist donations surpass everything else in expectation. Even with pretty aggressive fudging of the numbers, it easily outweighs other things. Getting Longtermist jobs is especially impactful, because talent is a much bigger constraint than money. See here for some especially impactful project ideas that an individual could work on, and check out the 80,000 Hours job board for high-impact Longtermist careers. 80,000 Hours also has more general career advising.

I want to emphasize: getting one of these jobs is much much much much more impactful than donating. Some very wealthy longtermists are already willing to fund nearly any sufficiently promising project. The greater need is therefore to identify and build additional high-impact projects, rather than merely to raise more money. Donations are still good, but they’re a lot less good.

1.3 Where Longtermists should donate

For Longtermist donations, some good places to donate are the EA Long-Term Future Fund, Longview,¹ and Forethought.² The first two of these provide grants for impactful Longtermist projects. If you think Longtermism is right, then it’s probably a pretty good bet to give your money to smart people who spend their days thinking about how best to spend money longtermistly. Impactful political donations are likely many times more impactful than this (and harder for Anthropic donors to saturate). Note: this isn’t a particularly idiosyncratic view of mine; my sense is that most Longtermist grantmakers would agree with this.

If you are convinced that Longtermism is this important, then you really should try to make an impact in Longtermist areas. Demandingness concerns do not apply when we’re dealing with this kind of positive impact. If the impact in expectation that one person can have by working on Longtermist cause areas is equivalent to preventing more suffering than all the suffering so far in human history, then it’s obviously obligatory. Even if morality doesn’t always demand doing what’s best, it does when the stakes are this high.

2 Growing effective altruism

2.1 Why grow EA?

There are some organizations that are focused on growing the effective altruism movement. Some examples include the EA Infrastructure Fund, Effektiv Spenden, regional EA groups, Giving What We Can, and 80,000 Hours. These tend to be super effective: dollars given directly to Giving What We Can return about 6 dollars to effective charities and dollars given to Effektiv Spenden return about 13 dollars to effective charities.

The case for this being good is pretty straightforward. Many of the people doing the best work on the world’s biggest problems are EAs. EA is generally more talent constrained than money constrained, so getting productive people involved in the movement is valuable. And growing EA pretty straightforwardly dominates global health and animal welfare, because growing EA returns more money to global health and animal welfare. I don’t think it’s clear whether this is higher or lower impact than Longtermist donations. But it seems clear it beats global health, say, when each dollar given to GWWC returns many dollars to global health charities.

2.2 Comparing growing EA to direct Longtermist interventions

Two factors support growing EA over traditional Longtermist organizations:

1. Longtermism is more talent constrained than money constrained. In other words, there’s a bigger problem with a limited supply of talented people than a limited supply of funds. This supports converting money into talent by growing EA.

2. In my view, most future value is lost in worlds where we survive but fail to get a near-optimal future. If this is a bigger threat than specific catastrophes, then growing the influence of those motivated by doing what’s best is hugely important.

One factor that points in the opposite direction: we might be at an unusually important time given very rapid AI technological advancement. It may be that work in the next few years will be more important than work in the years after that. Thus, growing EA—which often takes years to pay off—won’t be worth it in time. On the other hand, we should have reasonable uncertainty about when AI will arrive, so there are still decent odds that superintelligent AI arrives after growing EA pays off.

Another advantage of promoting EA is that even if you’re wrong about what’s best, you still can contribute to whichever area is most important. So even if I’m wrong that Longtermism is right, I can still know that I’m preventing people and animals from suffering and dying. I suspect that this moves me more than it should.

My general guess, if I had to rank these things gun to my head, is that the ranking for donations would be something like:

Longtermist political donations>EA promotion>Other Longtermist donations.

And for careers, it’s too close to call, though I’d tentatively prioritize specifically Longtermist careers over EA promotion careers that don’t focus mostly on Longtermist stuff. Though things are close enough that the main things you should consider are personal aptitude and fit. While my guess is that pretty much any net-positive Longtermist job will beat the best global health job in expectation, the same isn’t true here.

3 Animal welfare and global health

3.1 The basic case for animal welfare

The basic case for animal welfare over global health is that it is ridiculously cheap to help animals. Charities helping chickens prevent about 10 years in a cage per dollar they raise. This pretty clearly dominates global health’s direct impacts on humans. It costs about 4,500 dollars to save one life. Let’s assume the saved person would have lived an extra 40 years.

Thus, for 4,500 dollars you can either add 14,600 days of human life or spare chickens from 45,000 years in cages. Even if we assume that the ten-year estimate is inflated by one order of magnitude, then it still comes out to sparing a chicken from a year in a cage or adding about three days of extra human life. Surely sparing the chicken from a year of nightmarish torture is better?

And note: many animal welfare charities are even better than this. Given the near total neglect of invertebrates and wild animals, donating to help them can be even more impactful. And when one takes into account the long-term effects that neglect of animals—farmed and especially wild—would have, this seems quite good from a Longtermist perspective. We might spread factory farming throughout the galaxy, and we’re even likelier to spread wild animal suffering.

3.2 Longtermism and growing effective altruism>animal welfare

Let’s first compare animal welfare donations to Longtermist donations. Animal welfare, I’d guess, is lower stakes than Longtermism. Nearly all expected future welfare will be experienced by digital beings. This is both because of the amazing possible efficiency of compute³ and the ease of converting large amounts of energy into well-off digital minds. Given how many digital minds there could be, this holds even if you have a low credence in the possibility of digital sentience.

Thus, if you’re interested in steering the long-term future, you should focus more on digital welfare than animal welfare. This favors Longtermist interventions over animal welfare interventions, because they have a much more direct route to benefitting digital minds.

The case for growing EA over animal welfare is similar.

1. Its indirect effects are greater. Getting hard-working people on board with animal welfare probably has better second-order effects than direct work on animal welfare. It leads to more people who are on board with reducing animal suffering, which is important for eventually abolishing factory farming and seriously mitigating wild animal suffering.

2. Growing EA also grows animal welfare—my guess is that each dollar spent growing EA returns multiple dollars to effective animal charities.

3. Growing EA has similar benefits to Longtermist donations—benefitting digital minds, in expectation, which likely swamps the benefits of animal welfare charities. Thus, if Longtermism beats animal welfare, probably growing EA does too.

3.3 Animal welfare>global health

Now let’s compare animal welfare to global health. The best argument for global health over animal charities is that global health charities lower wild animal populations. I’ll first explain this argument and then say why I don’t buy it.

There are strong reasons to think that animals in nature live bad lives. Most animals, especially the small and simple ones that are most numerous, have very short lives that end in a painful death. For this reason, I think it’s more likely than not that it is good for wild animal populations to be lowered. Humans lower wild animal populations to a staggering degree, especially populations of insects. Brian Tomasik estimated that the average person prevents about 14 million years of insect life per year they’re alive.

Similarly, Tomasik estimated that the Against Malaria Foundation prevents about 14,000 insect life-years per dollar. Let’s assume chickens are 10 times more conscious in expectation than the average insect. Let’s additionally assume that insects in nature on average live about 2 weeks. This means that each dollar given to the Against Malaria Foundation prevents about 392,000 insect deaths, which are mostly pretty painful. Adjusting for sentience, this ends up about as good as preventing around 39,200 painful chicken deaths. Let’s cut this by a factor of ten to adjust for uncertainty about the net effect of global health charities on insects.

Which is worse: ten years in a cage or 3,900 painful deaths? Well, that’s about one death per day. It’s not obvious how a day in a cage compares to a death from starvation or predation. So for this reason, I’m pretty uncertain about the direct comparisons.

However, the expected second-order effects of animal welfare donations are a lot more positive than those of GiveWell donations. Here is why: nearly all animals, in expectation, will live in the far future. More importantly, almost none of them will live on earth. There are about 10^21 stars in the reachable universe. Even if you think it’s pretty unlikely that we’ll settle many of those planets, one-in-a-million odds that we’ll settle 1% that many planets still means that we’ll settle 10^13 planets in expectation.

Compared to this, whatever long-term environmental effects GiveWell charities have on earth is an insignificant blip. If we’re concerned about benefitting animals, what matters most is ensuring that their lives are good outside of earth. It’s easy to see how changing legal standards for animals and shifting cultural attitudes might do that. It’s hard to see how GiveWell charities would do that.

I’ll write an article about this at some point, but just as nearly everyone who will live, in expectation, will reside in the far future, almost everyone in expectation will reside in distant star systems. The universe is big (source???). Thus, when trying to make the world better, just as we should care a lot about making sure that the vast distant temporal regions go well, we should also mostly focus on trying to make the vast distant spatial regions go well. That matters a lot more than what happens on earth. Thanks to Toby Ord for pointing this out. When you take this seriously, animal welfare impact-mogs global health.

4 Conclusion

Comparing cause areas is important and difficult. The thing I’m most sure of is that EA promotion and Longtermism beat global health and animal welfare. I’m also pretty sure that animal welfare beats global health in expectation. In any case, all the areas are hugely important, and people who give to one tend to give to others as well, so just generally giving EA charities more money is very good!

It is a tragedy that we have to make these sorts of tradeoffs: that there is some chance that future life is extinguished, that there’s a good chance that it doesn’t reach its full potential, that animals suffer by the billions in ghastly torture chambers of human construction, that more suffer in the wild, and that children die every day. But when faced with many horrible things, one has to make tradeoffs; one should do it sensibly, after reflecting carefully, instead of just giving to the charity that they like best.

1. Conflict of interest alert: they’ve given grants to Forethought which is where I work. I think it’s a good place to give because it’s one of the only places seriously tackling many of the biggest issues during the intelligence explosion—especially trying hard to get a near-best future, rather than just an okay one.

2. Other conflict of interest alert: I work for Forethought for another month.

3. @Noah Birnbaum (context for people not named Noah: Noah and I have a dumb running joke that consists of us simply saying the word “compute.”)