[epistemic effort: I had a lot of conversations with other EAs]
[see also: On Philosophers Against Malaria; GiveWell: A Case Study In Effective Altruism; Contra The Giving What We Can Pledge; What Is The Giving What We Can Pledge?; Why You Should Focus More On Talent Gaps, Not Funding Gaps]

There has recently been some discussion among effective altruists about whether we should be encouraging people to take the Giving What We Can pledge, which commits pledge-takers to give ten percent of their income to the charities they consider most effective. I recently wrote a blog post encouraging people to take the pledge, so I thought it would be a good idea to write up my reasoning for why I think this is a good idea.

Talent Gaps and Funding Gaps

As a vast oversimplification, there are two ways that a cause can be constrained. They might not have enough money to do the things they want (buy malaria nets, hire talented people, make grants, etc.), which means there’s a funding gap. Alternately, they might not have enough talented people to do the things they want: for instance, they might really want to hire someone to focus on biosecurity research, but there are only six people who have the level of expertise in biosecurity they want and all of those people are happily employed somewhere else and don’t want to change.

(This is just an example. I have no knowledge of whether there is actually a talent gap in biosecurity.)

Of course, funding can be turned into talent, and talent can be turned into funding. If you have lots of money, you can raise salaries to attract more talented people. If you have lots of talented people, you can convince some of them to retrain into a high-earning profession and donate lots of money to you. However, in both cases, it’s sometimes not possible. There simply might not be any, say, professional economists who care about and want to research animal rights, in which case offering higher salaries will not allow you to hire one, at least in the short run. In the nonprofit sector in particular, offering high salaries can be risky if you want to hire people who share your values: it incentivizes people to pretend to share your values so that they can get your money. And some problems can be solved only with talent and not with funding: for instance, if the best way to improve the world is founding a startup that helps people in the developing world or working for the IMF or the World Bank so that they adopt better policies, then you have a problem, because it’s kind of hard to turn additional nonprofit donations into startup founders and IMF employees. (Unless, I guess, the nonprofit in question is 80,000 Hours.)

On the turning-talent-into-money side, it’s perfectly possible that you have a lot of extremely talented philosophers, who might be very useful for convincing impressionable undergraduates and developing better arguments in favor of effective altruism but who might have a somewhat difficult time switching to becoming finance quants. So it makes sense to talk about talent gaps and funding gaps as being two different things.

How much you care about talent gaps versus funding gaps depends on what causes you think are most important. Some causes are better at soaking up money than others: for instance, Give Directly has remarkably low overhead, so a single talented person can direct huge amounts of money. If you support relatively unpopular charities, it’s also possible that they’re primarily lacking funding: for instance, the Effective Altruism Foundation’s charities (Sentience Politics, the Foundational Research Institute, and Raising for Effective Giving) and MIRI. Conversely, my understanding is that Open Phil’s primary problem is finding enough talented people to give away all of Dustin Moskovitz’s money. And the animal sector in general is kind of weird, because they seem to be primarily constrained by the fact that, while it seems plausible that there is a good way to transform money into animal welfare, no one has any idea what it is. That problem can be solved through either more money (for research grants, etc.) or more talent (for researchers, etc.).

It seems true to me that, on current margins and with a few exceptions, talent is more important than funding for most of the causes effective altruists care about. But that’s precisely because people keep taking the Giving What We Can pledge! Pledgers donate a lot of money. We can expect pledgers to donate a lot of money in the future, because many of the people taking it now are broke students. And the number of people pledging has grown and the growth shows no signs of diminishing. The Giving What We Can pledge is a success, and because it is a success the problem it’s trying to fix is a lot less bad than it used to be, and we instead have this different problem. But that’s somewhat dependent on us continuing to encourage pledgers to keep their pledges and new people to take the pledge; if we don’t, then those predictions about how much money the future effective altruist community will have will be wrong, and money will be a lot more important than it was.


Part of the problem with focusing on talent is that filling talent gaps is terrible fucking marketing.

Taking the Giving What We Can pledge might not be easy, but it is simple. There is exactly one thing that all pledgers need to do. It takes one sentence to explain (“give ten percent of your income for the rest of your life to the charity that you think will do the most good”). In theory, figuring out which charities do the most good should add more sentences to the pitch. In practice, people just say “donate to GiveWell top charities.” The people they’re pitching to believe them, because it seems like well-informed people think GiveWell is good and also GiveWell’s analyses have lots of numbers and citations in. (I don’t actually think this is an unreasonable use of heuristics.)

Conversely, it’s impossible to give one-size-fits-all advice about filling talent gaps. While my dollars are pretty much the same as everyone else’s dollars, my talents are very different from everyone else’s talents. In fact, it’s quite remarkable the number of potentially high-impact careers I’m ludicrously ill-suited for. (Tech startup founder! Party politics! Founding an effective non-profit! Product manager! AI risk researcher!) Funding gaps often tend to be fairly large, on the order of twenty million dollars or so. If I identify a good opportunity, I can tell thirty of my friends that they should donate to it, without making a huge dent in the funding gap. On the other hand, a charity might have one or two unfilled positions: if I tell two of my friends to work for a charity, I will have to find other opportunities for the other twenty-eight friends. In practice, people who wish to fill talent gaps usually have to find the specific talent gaps themselves, with guidance about general areas to look in from 80,000 Hours, rather than relying on outside experts for all their decision-making.

As I said, the Giving What We Can pledge takes one sentence to explain. The Life You Can Save pledge takes about three sentences to explain. The Giving What We Can pledge is far, far more popular, even though The Life You Can Save has a lot of other structural advantages (such as being more reasonable for most people and having been founded by a far more famous philosopher). Conversely, judging by the length of 80,000 Hours’s workshops, how to fill talent gaps takes something like four hours to explain. If the one sentence/three sentences thing makes such a difference, I shudder to think of the cost of switching to something that takes four hours.

Filling a talent gap is also a much bigger life change for many people. Giving What We Can’s primary audience is students who can expect to earn a lot of money in the future. For the prototypical Giving What We Can pledger, taking the Giving What We Can pledge looks like taking a $72,000/year job once they graduate rather than an $80,000/year job. It involves essentially no suffering or change of behavior. Even for non-student pledgers, Giving What We Can targets a relatively affluent group, for whom taking the pledge might mean eating out a bit less and not going on vacation this year, which is not a huge life difference. (If this doesn’t look like you, take the Life You Can Save pledge instead, seriously. That pledge involves different donations for people of different income levels, and in my experience manages to not be particularly burdensome for most people.)

Conversely, filling a talent gap involves completely changing your career. It could involve a shift in self-image, if you’ve seen yourself as being a doctor since you were sixteen and now you’re going off to become an economist instead. It could involve learning new skills. It could involve a significant cut in pay. It could involve moving to a different place. It could involve working a job that you’re not familiar with and none of your friends are familiar with either. It probably does involve becoming significantly more ambitious than you were before and learning to believe yourself capable of things you currently don’t think you’re capable of.

In my experience, it is extremely common for people’s effective altruist stories to begin “I was donating and getting involved in the community for a couple of years, when I decided to [work for a nonprofit/found a nonprofit/go back to school/work for Wave].” This makes perfect sense! When you’re part of the effective altruist community, you’re surrounded by people who think that trying to fill talent gaps is a totally reasonable thing to do. You’ve learned in a low-stakes way about what cause areas have a talent gap. You know people who work in high-impact careers. You have more of a network and better advice. Most importantly, a high-impact career starts to seem like a thing that normal people, people like you, do– not something that’s done by serious and terrifyingly competent people in fancy suits.

Probably most people who take the Giving What We Can pledge won’t fill any talent gap: many people, through no fault of their own, do not have the abilities necessary for a high-impact career, and many other people aren’t willing to. Maybe five out of a hundred pledgers will. But I think that promoting the pledge is a far more effective way of recruiting people to fill talent gaps than doing so directly is.

(This might be a weird and kind of off-putting message for potential pledgees. Two things. One, I also think what I said above: a big part of the reason that funding gaps are less important for many cause areas is that people keep taking the pledge. I expect that your donations will produce a lot of positive value even if you never fill a talent gap. Two, while I am less aggressive than some people in terms of “we should keep out the people who don’t think like EAs from effective altruism!”, I do care about it a non-zero amount. I would prefer to attract people who consider the above to be an example of refreshing honesty and transparency, because I suspect they are more likely to share effective altruist values on other issues as well.)

Commitment Mechanisms 

It has recently come to my attention that many effective altruists seem to disagree with each other about what the term “pledge” means. Some people seem to view it as a solemn promise that binds your future self to donate ten percent of their income even if your future self thinks it has much better things to do with the money. Other people seem to view it as a statement of intent: you plan to donate ten percent of your income for the rest of your life, but if something better comes along you will do that instead.

Consider a wedding vow: two people promise to love each other for as long as they both shall live. In general, the first group would not think that this promise is completely unbreakable: they would support a divorce if, say, you and your wife hate each other and you’ve gone to marital therapy and you can’t fix it and every time you interact with each other you have a screaming match. But you might, five years into the marriage, meet someone else and think that that person would be a much better wife than your wife, even though you continue to like and get along with your wife. If you then divorce your wife, they would think, you are doing something quite wrong. You are not allowed to get out of the promise just because you think you have a better option. The whole point is to get your future self to do things even if your future self thinks it is not in their best interest. If you wanted a promise that you can get out of whenever you like, then you should have pledged to be together as long as you both shall love instead.

The second group, however, views “as long as we both shall live” as a statement of intent. Right now, you intend to stay with your wife for the rest of your life, and you predict based on your best model of your future self that you will stay with her. But if it turns out that you actually have a better option than your wife, then it wouldn’t be doing anything wrong to leave her. At worst, it is an empirical mistake: you incorrectly modeled your future self in a way that led you to make wrong predictions.

In the case of the Giving What We Can pledge, the first group feels that the pledge binds you to donate ten percent of your income, period. Of course, there are serious circumstances in which it is justified to break the pledge: the pledge does not require you to donate ten percent of your welfare check. But even if future you thinks it would be better to spend the money on something else that is not a donation, you have still bound your future self to donate. Conversely, the second group feels that you should do the best thing. If you think that there’s something other than donating money that would lead to the best outcome, then you should do the thing that would lead to the best outcome, pledge or no pledge.

I do not here mean to discuss my opinions on which interpretation of the concept of “pledge” is correct. I personally favor the former, but I don’t think that there is any necessity for there to be a broad effective-altruism-wide consensus on the definition of the word “pledge”. I am happy to work with people who disagree with me about what promises bind you to do, even though I would be quite unlikely to e.g. marry them. (For what it’s worth, the Giving What We Can pledge appears to endorse the first interpretation, although I think it would be worth doing some empirical research about how most pledgers and the target audience of the pledge interpret it. This seems like it should affect their marketing.)

Under the second interpretation, I think taking the pledge is a straightforwardly good thing to do, except that if you don’t have very much money, consider taking The Life You Can Save pledge instead.

In the first interpretation, it’s more complicated, because of issues of flexibility. Imagine that I am living in the Bay Area and choosing between two jobs: a $100,000/year programming job and a $20,000/year job working for an animal rights nonprofit. I consider the pledge to bind me to donate 10% of my income.

There are two potential problems. First, there are a lot of ways to turn money into productive work. I might use money to save time: for instance, buying a dishwasher, taking Ubers instead of the train, or hiring a cleaner. I might use money to avoid burnout: for instance, purchasing enjoyable experiences that refresh me and allow me to return to my work happier. I might use money to avoid stress and worry: for instance, if I can easily pay my rent, I don’t have to spend time worrying about where I’m going to live. If my work at the animal rights nonprofit is sufficiently important, then spending money on those things can create more value than donating it. But if I’ve pledged to donate, then the $2000 has to go to charity– even if the better way to spend the money would be on not being homeless.

Second, most people want to spend a certain amount of resources on altruism and a certain amount of resources on selfishness. If I am already forgoing eighty thousand dollars of income to work at an animal rights nonprofit, then an additional two thousand dollars of donations might very well push me over the limit of resources I’m willing to spend on altruism. That means I’m far less likely to take the job, even if I believe the job will allow me to do the most good.

Of course, there are benefits to taking the pledge as well. Many people like the automatic nature of it: you donate ten percent of your income and then you never have to worry about altruism again. I personally find that making commitments helps make me be the sort of person who wants to keep my commitments: just as my vow to be married for life makes me want to be nice to my husband, my pledge to donate ten percent of my income makes me want to be a more compassionate and altruistic person.

I personally have not yet taken the Giving What We Can pledge, because it offends my aesthetics to take a lifelong pledge by filling out a form and pressing a button, and I have yet to come up with a suitably dramatic ritual. However, I do plan to take the pledge as soon as that’s sorted. My husband and I are sufficiently financially stable that it is unlikely either of the considerations I outlined above will ever apply: we will basically always have enough money to spend money on time-saving conveniences and not experience financial stress. The situations where it might happen (for instance, if my husband becomes disabled and cannot work or if some financial disaster wipes out all our savings) are situations in which I also think it would be reasonable to break the pledge.

I think Try Giving– in which a person agrees to donate ten percent for a certain period of time they decide, and then reassesses– is a great strategy for people who want to get the benefits of committing to give, but who also think they might want to switch to direct work in the future. It makes me sad that Try Giving appears to be marketed solely to people who want to, well, try giving. I think that for a lot of people, who aren’t sure about their future financial stability or whether they’ll switch into direct work, indefinitely planning to pledge for a few years at a time is the smart move. I wish there was a program (Flexible Giving?) for them.

Encouraging Long-Term Commitments 

I’d like to highlight this excellent comment of Justis Mills’s, which points out some ethical problems with encouraging people to take the pledge, instead of encouraging people to try giving. It seems wise that commitments should be done after careful thought, rather than in a moment of passion. To return to the marriage example, most people think it’s wise to date someone for a year or two before you marry them, and perhaps cohabitate; that way, when you make a lifelong commitment, you know that it’s informed.

Holly Elmore argues that getting students to take the pledge is good because they’ve never had a stable income and they don’t know what they’d be missing. It’s harder to give up three thousand dollars than it is to take a job that pays $27,000/year. Justis apparently changed their mind in response to Holly’s argument, but I think their original comment was right. The benefit Holly talks about can be easily obtained through doing Try Giving for two or three working years before taking the Giving What We Can pledge.

My objections are threefold. First, it seems to me that encouraging people to take the pledge in a less than fully informed way makes it more likely that they will break the pledge. That means that Giving What We Can has a less accurate estimate of how much money they can anticipate moving in the future. It also weakens the pledge for everyone else by turning pledge-breaking into a normal thing that lots of people do. If everyone’s breaking their pledges, there’s less social opprobrium for pledge-breaking.

Second, I’m concerned about flexibility. If supporting the Giving What We Can pledge is good both because donations are useful and because it helps us recruit more talented EAs, then we don’t want to make it harder for people to do direct work. Doing Try Giving for a few years gives the average new effective altruist a lot of time to familiarize themselves with arguments about talent gaps and figure out whether direct work is right for them, so they can decide whether to take the full pledge, do what I’m calling “flexible giving”, or not take the full pledge and do direct work instead.

Third, I do actually care about informed consent. I care about cooperating and living peacefully with people of other value systems, which implies that I should not use underhanded methods to get them to do things that are good by my lights and not by theirs. It seems to me to be disrespectful to get someone to promise something when they don’t really know what they’re promising. I would not like it if other people did such a thing to me, and therefore I am not going to do it to others. In practice, I think uninformed pledgers are not very useful and perhaps even harmful (see my first point), so this is not a very hard principle to abide by. I understand that some people have different values, but fortunately for them my desire to cooperate with people of different value systems extends to them.

In the future, when I write posts about Giving What We Can, I will encourage people to sign up for a year of Try Giving instead of taking the full pledge.