I'm glad to see more of this criticism as I think it's important for reflection and moving things forward. However, I'm not really sure who you're critiquing or why. My response would be that your critique (a) appears to misrepresent what the "EA mainstream" is, (b) ignores comparative advantage, or (c) says things I just outright disagree with.
~
The EA Mainstream
Perhaps the biggest example of this is the prevalence of “earning to give”. While this is certainly an admirable option, it should be considered as a baseline to improve upon, not a definitive answer.
I imagine we know different people, even within the effective altruist community. So I'll believe you if you say you know a decent amount of people who think "earning to give" is the best instead of a baseline.
However, 80,000 Hours, the career advice organization that basically started earning to give have themselves written an article called "Why Earning to Give is Often Not the Best Option" and say "A common misconception is that 80,000 Hours thinks Earning to Give is typically the way to have the most impact. We’ve never said that in any of our materials.".
Additionally, the earn...
GiveWell, for example, has explicitly distanced themselves from numerical calculations (albeit recently) and several EAs have called into question the usefulness of cost-effectiveness estimates, a charge that was largely lead by GiveWell.
I'll speak up on this one. I am a booster of more such estimates, detailed enough to make assumptions and reasoning explicit. Quantifying one's assumptions lets other challenge the pieces individually and make progress, where with a wishy-washy "list of considerations pro and con" there is a lot of wiggle room about their strengths. Sometimes doing this forces one to think through an argument more deeply only to discover big holes, or that the key pieces also come up in the context of other problems.
In prediction tournaments training people to use formal probabilities has been helpful for their accuracy.
Also I second the bit about comparative advantage: CEA recently hired Owen Cotton-Barratt to do cause prioritization/flow-through effects related work. GiveWell Labs is heavily focused on it. Nick Beckstead and others at the FHI also do some work on the topic.
It seems like the EA mainstream either agrees with many of your critiques already (and therefore you're just trying to convince EAs to adopt the mainstream)
I think that on some of these questions there is also real variation in opinion that should not simply be summarized as a clear "mainstream" position.
But I think there's a large difference between "here's a first-pass attempt at a cost-effectiveness estimate purely so we can compare numbers" and "this is how much it costs to save a life".
You still have to answer questions like:
Those choices imply judgments about expected value. Being evasive and vague doesn't eliminate the need to make such choices, and tacitly quantify the relative value of options.
Being vague can conceal one's ignorance and avoid sticking one's neck out far enough to be cut off, and it can help guard against being misquoted and PR damage, but you should still ultimately be more-or-less assigning cardinal scores in light of the many choices that tacitly rely on them.
It's still important to be clear on how noisy different inputs to one's judgments are, to give confidence intervals and track records to put one's analysis in context rather than just an expected value, but I would say the basic point stands, that we need to make cardinal comparisons and being vague doesn't help.
I would argue (perhaps self-servingly) that academia is another example of such a path
Academia is, in my mind, the textbook example of people doing something because it's familiar, not because they've searched for it and it's the right choice. Most of the academics I know will freely state that it only makes sense to go into academia for fame, not for money- and so it's not clear to me what you think the EA benefit is. (Convincing students to become EA? Funding student organizations seems like a better way to do that.)
The history of effective altruism is littered with over-confident claims, many of which have later turned out to be false. In 2009, Peter Singer claimed that you could save a life for $200 (and many others repeated his claim). While the number was already questionable at the time, by 2011 we discovered that the number was completely off. Now new numbers were thrown around: from numbers still in the hundreds of dollars (GWWC's estimate for SCI, which was later shown to be flawed) up to $1600 (GiveWell's estimate for AMF, which GiveWell itself expected to go up, and which indeed did go up).
Another good example is GiveWell's 2009 estimate that "Because [our] estimate makes so many conservative assumptions, we feel it is overall reasonable to expect [Village Reach's] future activities to result in lives saved for under $1000 each."
"8 lives saved per dollar donated to the Machine Intelligence Research Institute. — Anna Salamon"
I agree with Luke's comment; compared to my views in 2009, the issue now seems more complicated to me; my estimate of impact form donation re: AI risk is lower (though still high); and I would not say that a particular calculation is robust.
"Why haven't more EAs signed up for a course on global security, or tried to understand how DARPA funds projects, or learned about third-world health? I've heard claims that this would be too time-consuming relative to the value it provides, but this seems like a poor excuse. If we want to be taken seriously as a movement (or even just want to reach consistently accurate conclusions about the world)."
This one worries me quite a bit. The vast majority of EA's (including myself) have not spent very much time learning about what the large players in third world poverty are (e.g. WHO, UN). In fact you can be an "expert" in EA content and know virtually nothing about the rest of the non-profit/charity sector.
It seems to me that the effective altruist movement over-focuses on “tried and true” options, both in giving opportunities and in career paths. Perhaps the biggest example of this is the prevalence of “earning to give”.
I would have guessed that the biggest example is the focus on poverty reduction / global health initiatives that GiveWell and GWWC have traditionally focused nearly all their attention on. E.g. even though Holden has since the beginning suspected that the highest-EV altruistic causes are outside global health, this point isn't mentioned on GiveWell's historical "top charities" pages (2012, 2011, 2010, 2009, 2008), which emphasize the important focus on "tried and true" charitable interventions.
One in six Yale graduates go into finance and consulting, seemingly due to the simplicity of applying and the easy supply of extrinsic motivation. My intuition is that this ratio is higher than an optimal society would have, even if such people commonly gave generously.
Because those one-in-six don't all give generously, we can't conclude whether it's right at the margins for graduates to go into earning to give, even if we grant the assumption about the ratio in an optimal society.
I agree that it's worth looking at a wider spread of career possibilities, but this isn't the argument to use to get there.
Thanks for the interesting critique. I agree with you that EAs often make over-confident claims without solid evidence, although I don't think it's a huge issue that people sometimes understate how much it costs to save a life, as even the most pessimistic realistic estimates of this cost don't undermine the case for donating significant sums to cost-effective charities.
Am I right in understanding that you think that too many EAs are pursuing earning to give careers in finance and technology, whereas you think they'd have greater impact if they worked in s...
careers in finance and software (the two most common avenues for this) are incredibly straight-forward and secure.
What are you talking about? Investment Banking, at least, has a huge attrition rate. Careers in IB are short and brutal.
This still feels like a "we need fifty Stalins" critique.
For me the biggest problems with the effective altruism movement are:
1: Most people aren't utilitarians.
2: Maximizing QALY's isn't even the correct course of action under utilitarianism - its short sighted and silly. Which is worse under utilitarianism: Louis Pasteur dying in his childhood or 100,000 children in a third world country dying? I would argue that the death of Louis Pasteur is a far greater tragedy since his contributions to human knowledge have saved a lot more than 100,000 liv...
Effective Altruism (and critiques of it) need to think at the margin. If I give $X to an organization doing good chances are this won't displace someone else from giving the organization money. In contrast, if I get a job at such an organization I have probably displaced someone else from taking that job. This kind of marginal analysis greatly strengthens the value of the “earning to give” path of effective altruism.
Naive efficient-market analysis suggests that if finance and computer programming are predictable and lucrative careers, there should be some less stable career option which is even more lucrative on average. For someone who's genuinely earning to give, and planning to keep only a pittance for their own survival regardless, that variability shouldn't matter.
(This comment is on career stuff, which is tangential to your main points)
I recently had to pick a computer science job, and spent a long time agonizing over what would have the highest impact (among other criteria). I'm not convinced startups or academia have a higher expected value than large companies. I would like to be convinced otherwise.
(Software) Startups:
1) Most startups fail. It's easy to underestimate this because you only hear the success stories.
2) Many startups are not solving "important" problems. They are solving relatively minor ...
What is particularly worrysome to me is that the positive effects of interventions such as improvements in the education are much harder to qualitatively calculate.
Say, an individual can make the choice to be a judge in the US, or to be a banker and donate a lot of money to the charities. The straightforward calculation does not take into account the importance of good people among the justices; without such people US would probably have been in no position to send aid (and would need monetary aid itself).
This post is now five years old. It seems to me that EA has shifted from where is was five years ago (not that I was around back then), and it seems that people in EA largely share your concerns. It is generally recognized that doing good effectively is a very complex challenge, much of which is hard to quantify, but that we should try our best to figure out how to do it (witness the increasing popularity of global priorities research).
I don't like it when people "burn the candle from both ends". You complain about over-reliance on quantitat...
The history of effective altruism is littered with over-confident claims, many of which have later turned out to be false. In 2009, Peter Singer claimed that you could save a life for $200 (and many others repeated his claim).
I think this sentence misrepresents Peter Singer's position. Here's a relevant excerpt from The Life You Can Save (pp. 85-87, 103). As you can see, Singer actually criticizes many organizations for providing excessively optimistic estimates, and doesn't himself endorse the $200 per-life-saved figure.
...For saving lives on a large s
It is interesting, what people inside EA find troubling, compared to people outside. (I do not identify myself as EA).
For me, the most repellent things are mentioned here: http://lesswrong.com/lw/j8n/a_critique_of_effective_altruism/#poor-psychological-understanding
In other words, self sacrifice is expected from me to the extent, that my life would suck. No, thanks.
Specifically, the issues about children:
Prominent EA Julia Wise and her husband have decided to have kids. IMO, a good way to think about EA is that everyone makes their own trade-off between their own quality of life and the quality of life of others. You can also think of helping people in terms of scoring points.
But part of the strength of the (EA) movement is the message that people can achieve a lot without great personal sacrifice. I wonder where that message got lost.
Hm, are You asking why I did not notice that message, or why did it, objectively, get lost ? I will answer the first part, why I never noticed such a message.
Short after learning, that EA exist, my CFAR friend, who is dating an EA, told me about a disagreement she had with her boyfriend. He did not not want her to go her best friends wedding, because travel expenses and time spent could be used better. (Although he later admitted it was a half joke from his side). She also told me, he periodically scolds her, her temperament is too prone to hapiness, which makes her less understand suffering, which makes less incentive for her to work on preventing it. That was not a joke.
I had some shocks at the EA facebook group. OK, Ben Kuhn complains that EA fb group is stupid these days. I was told that supporting less than optimal charity is immoral. I translate it into examples, that supporting any Slovakian charity is immoral, because money are better spent on AMF. Supporting this baby is probably even more immoral than my f
"Why haven't more EAs signed up for a course on global security, or tried to understand how DARPA funds projects, or learned about third-world health?"
A very interesting point, and you've inspired me to take such a course. Does anyone have any recommendations for a good (and preferable reputable, given our credential addicted world) course relating to global security and health?
Another reasonable concern has to do with informational flow-through lines. When novel investigation demonstrates that previous claims or perspectives were in error, do we have good ways to change the group consensus?
I agree with Luke's comment; compared to my views in 2009, the issue now seems more complicated to me; my estimate of impact form donation re: AI risk is lower (though still high); and I would not say that a particular calculation is robust.
my estimate of impact form donation re: AI risk is lower (though still high)
Out of curiosity, what's your current estimate? I recognize it'll be rough, but even e.g. "more likely than not between $1 and $50 per life saved" would be interesting.
Recently Ben Kuhn wrote a critique of effective altruism. I'm glad to see such self-examination taking place, but I'm also concerned that the essay did not attack some of the most serious issues I see in the effective altruist movement, so I've decided to write my own critique. Due to time constraints, this critique is short and incomplete. I've tried to bring up arguments that would make people feel uncomfortable and defensive; hopefully I've succeeded.
Briefly, here are some of the major issues I have with the effective altruism movement as it currently stands:
Over-focus on “tried and true” and “default” options, which may both reduce actual impact and decrease exploration of new potentially high-value opportunities.
Over-confident claims coupled with insufficient background research.
Over-reliance on a small set of tools for assessing opportunities, which lead many to underestimate the value of things such as “flow-through” effects.
The common theme here is a subtle underlying message that simple, shallow analyses can allow one to make high-impact career and giving choices, and divest one of the need to dig further. I doubt that anyone explicitly believes this, but I do believe that this theme comes out implicitly both in arguments people make and in actions people take.
Lest this essay give a mistaken impression to the casual reader, I should note that there are many examplary effective altruists who I feel are mostly immune to the issues above; for instance, the GiveWell blog does a very good job of warning against the first and third points above, and I would recommend anyone who isn't already to subscribe to it (and there are other examples that I'm failing to mention). But for the purposes of this essay, I will ignore this fact except for the current caveat.
It seems to me that the effective altruist movement over-focuses on “tried and true” options, both in giving opportunities and in career paths. Perhaps the biggest example of this is the prevalence of “earning to give”. While this is certainly an admirable option, it should be considered as a baseline to improve upon, not a definitive answer.
The biggest issue with the “earning to give” path is that careers in finance and software (the two most common avenues for this) are incredibly straight-forward and secure. The two things that finance and software have in common is that there is a well-defined application process similar to the one for undergraduate admissions, and given reasonable job performance one will continue to be given promotions and raises (this probably entails working hard, but the end result is still rarely in doubt). One also gets a constant source of extrinsic positive reinforcement from the money they earn. Why do I call these things an “issue”? Because I think that these attributes encourage people to pursue these paths without looking for less obvious, less certain, but ultimately better paths. One in six Yale graduates go into finance and consulting, seemingly due to the simplicity of applying and the easy supply of extrinsic motivation. My intuition is that this ratio is higher than an optimal society would have, even if such people commonly gave generously (and it is certainly much higher than the number of people who enter college planning to pursue such paths).
Contrast this with, for instance, working at a start-up. Most start-ups are low-impact, but it is undeniable that at least some have been extraordinarily high-impact, so this seems like an area that effective altruists should be considering strongly. Why aren't there more of us at 23&me, or Coursera, or Quora, or Stripe? I think it is because these opportunities are less obvious and take more work to find, once you start working it often isn't clear whether what you're doing will have a positive impact or not, and your future job security is massively uncertain. There are few sources of extrinsic motivation in such a career: perhaps moreso at one of the companies mentioned above, which are reasonably established and have customers, but what about the 4-person start-up teams working in a warehouse somewhere? Some of them will go on to do great things but right now their lives must be full of anxiousness and uncertainty.
I don't mean to fetishize start-ups. They are just one well-known example of a potentially high-value career path that, to me, seems underexplored within the EA movement. I would argue (perhaps self-servingly) that academia is another example of such a path, with similar psychological obstacles: every 5 years or so you have the opportunity to get kicked out (e.g. applying for faculty jobs, and being up for tenure), you need to relocate regularly, few people will read your work and even fewer will praise it, and it won't be clear whether it had a positive impact until many years down the road. And beyond the “obvious” alternatives of start-ups and academia, what of the paths that haven't been created yet? GiveWell was revolutionary when it came about. Who will be the next GiveWell? And by this I don't mean the next charity evaluator, but the next set of people who fundamentally alter how we view altruism.
The history of effective altruism is littered with over-confident claims, many of which have later turned out to be false. In 2009, Peter Singer claimed that you could save a life for $200 (and many others repeated his claim). While the number was already questionable at the time, by 2011 we discovered that the number was completely off. Now new numbers were thrown around: from numbers still in the hundreds of dollars (GWWC's estimate for SCI, which was later shown to be flawed) up to $1600 (GiveWell's estimate for AMF, which GiveWell itself expected to go up, and which indeed did go up). These numbers were often cited without caveats, as well as other claims such as that the effectiveness of charities can vary by a factor of 1,000. How many people citing these numbers understood the process that generated them, or the high degree of uncertainty surrounding them, or the inaccuracy of past estimates? How many would have pointed out that saying that charities vary by a factor of 1,000 in effectiveness is by itself not very helpful, and is more a statement about how bad the bottom end is than how good the top end is?
More problematic than the careless bandying of numbers is the tendency toward not doing strong background research. A common pattern I see is: an effective altruist makes a bold claim, then when pressed on it offers a heuristic justification together with the claim that “estimation is the best we have”. This sort of argument acts as a conversation-stopper (and can also be quite annoying, which may be part of what drives some people away from effective altruism). In many of these cases, there are relatively easy opportunities to do background reading to further educate oneself about the claim being made. It can appear to an outside observer as though people are opting for the fun, easy activity (speculation) rather than the harder and more worthwhile activity (research). Again, I'm not claiming that this is people's explicit thought process, but it does seem to be what ends up happening.
Why haven't more EAs signed up for a course on global security, or tried to understand how DARPA funds projects, or learned about third-world health? I've heard claims that this would be too time-consuming relative to the value it provides, but this seems like a poor excuse if we want to be taken seriously as a movement (or even just want to reach consistently accurate conclusions about the world).
Effective altruists tend to have a lot of interest in quantitative estimates. We want to know what the best thing to do is, and we want a numerical value. This causes us to rely on scientific studies, economic reports, and Fermi estimates. It can cause us to underweight things like the competence of a particular organization, the strength of the people involved, and other “intangibles” (which are often not actually intangible but simply difficult to assign a number to). It also can cause us to over-focus on money as a unit of altruism, while often-times “it isn't about the money”: it's about doing the groundwork that no one is doing, or finding the opportunity that no one has found yet.
Quantitative estimates often also tend to ignore flow-through effects: effects which are an indirect, rather than direct, result of an action (such as decreased disease in the third world contributing in the long run to increased global security). These effects are difficult to quantify but human and cultural intuition can do a reasonable job of taking them into account. As such, I often worry that effective altruists may actually be less effective than “normal” altruists. (One can point to all sorts of examples of farcical charities to claim that regular altruism sucks, but this misses the point that there are also amazing organizations out there, such as the Simons Foundation or HHMI, which are doing enormous amounts of good despite not subscribing to the EA philosophy.)
What's particularly worrisome is that even if we were less effective than normal altruists, we would probably still end up looking better by our own standards, which explicitly fail to account for the ways in which normal altruists might outperform us (see above). This is a problem with any paradigm, but the fact that the effective altruist community is small and insular and relies heavily on its paradigm makes us far more susceptible to it.