I agree, and I’d add that what I see as one of the key ideas of effective altruism, that people should give substantially more than is typical, is harder to get off the ground in this framework. Singer’s pond example, for all its flaws, makes the case for giving a lot quite salient, in a way that I don’t think general considerations about maximizing the impact of your philanthropy in the long term are going to.
Alexander_Berger
Karma: 1,239
I think this argument is wrong for broadly the reasons that pappubahry lays out below. In particular, I think it’s a mistake to deploy arguments of the form, “the benefit from this altruistic activity that I’m considering are lower than the proportional benefits from donations I’m not currently making, therefore I should not do this activity.”
Ryan does it when he says:
How long would it take to create $2k of value? That’s generally 1-2 weeks of work. So if kidney donation makes you lose more than 1-2 weeks of life, and those weeks constitute funds that you would donate, or voluntary contributions that you would make, then it’s a net negative activity for an effective altruist.
Toby says:
One way I look at it is that I wouldn’t donate a kidney in order to get $2,000 (whether that was to be spent on myself or donated to effective charities), or equivalently, that I am prepared to pay $2,000 to keep my second kidney. This means that, for me at least, donating is dominated by extra donations.
The problem with these comparisons is that they’re totally made up. There’s a potential one-off activity (donating a kidney) which, Thomas argued above, has large benefits to recipients relative to costs to the giver. There’s also a question about how much you donate to charity. Based on the rationales you’re giving here, someone who is happy with the cost/benefit tradeoff of donating a kidney as a one off, but is convinced that it’s not as good cost/benefit as further donations, should actually increase their donations. However, my impression is that that has not been the reaction to these arguments; instead they justify current behavior/levels of altruism. (Toby, Ryan, correct me if I’m wrong here.) But donating a kidney would, according to most parties to the discussion, be net beneficial on its own terms. So the net impact of these arguments is to prevent people from taking positive sum altruistic actions, thereby reducing value.
There are kinds of costs that do mix between these two activities—genuinely foregone wages. And if your foregone wages were large and you decided that you would offset donations rather that consumption or savings with them, it would be perfectly appropriate to conduct this comparison. (Similarly, if the financial risk to future donations were higher, that would also make sense to offset.) But idly speculating about how much you’d have to be paid to do something, while taking the current level of donation as fixed, results in net negative impacts.
I think it’s a problem when the “effective” side of “effective altruism” is used as a argument against the “altruism” side. I should note that Jeff Kaufman and I had this framework argument on his post on this topic a while back on Less Wrong.
Someone emailed me this and asked for thoughts, so I thought I’d share some cleaned up reactions here. Full disclosure—I work at Open Phil on some related issues:
Thanks for the post—I think it’s helpful, and I agree that I would like to see the EA community engage more with Lant’s arguments.
If we’re focused primarily on near term human welfare (which seems to be the frame for the post), I think it’s really important to think (and do back of the envelope calculations) more explicitly in terms of utility rather than in terms of absolute dollars. In the post, you allude to the need to adjust for this (“It should be noted that the later stages of the growth accelerations affect progressively richer people, so produce less utility from additional consumption.”), but I think it’s actually first order. In general, I think true humanitarian welfare is distributed much more linearly than exponentially, and that Jones and Klenow’s welfare concept doesn’t map very well to how I think about utility. I don’t have any knock-down arguments here, but I think looking at life satisfaction survey data and lifespan data both suggest the relevant metric is much closer to log(GDP) than pure GDP: many people in rich countries are 100x richer than in poor countries, but I don’t think their lives are 100x (or even 10x) better, and they live <2x longer on average. I could propose some thought examples here (willingness-to-pay to save different lives, how many lives would you live in one place in exchange for just one in another), but I think the intuition is pretty straightforward. Some other thoughts on using GDP $ instead of something like average(log($)) or total(log($)) as your unit:
Using GDP $ ignores distribution, which is key to, e.g., the GiveDirectly case, which explicitly isn’t aiming at global GDP. More generally, growth accelerations typically increase inequality (at least for a while), and in the quick data I googled for India, the median income is less than half of the average; just adjusting for GDP/capita will miss some of the median income dynamics you recognize as important.
Using raw GDP makes you more likely to focus on rich countries: for instance, if I thought that relaxing zoning constraints would increase US GDP by 5% (example source), the perpetuity of that increase would be worth 6 times Lant’s calculation of the Indian growth episodes combined, even though it seems far less morally valuable to me. Log($) type considerations are, I think, a lot of what motivates a focus on developing countries in the first place, and would push towards more attention to poorer countries and domestic inequality within those countries, relative to a GDP-first framework.
Logging $ to get to utility terms generally removes the compounding dynamic in absolute $ that I think partially attracts people to growth arguments; I tend to think that’s good and correct, but am not sure, and would be interested in reading more on this if people have pointers.
One part of the case for a focus on GDP that I think might be right but that I’m uncertain about and would be interested in seeing quantified more is that growth itself causes others benefits (like health, education, etc) that should be counted separately from the more direct economic/subjective wellbeing benefits of growth. That seems like an obvious way that a log(GDP) utility function could be understating the value of growth. My intuition is that it would be surprising if more of the humanitarian impact (according to my values) of growth ran through second order impacts on health than through the direct impact on income/SWB, but I’m not sure about how the causal magnitudes would pencil out.
I think Carl Shulman’s old posts on log income and the right proxies for measuring long-run flow-through effects are interesting on this.
I totally grant that GDP is important and tightly correlated with lots of other good things, but I think using it as your comparison unit biases the calculation towards growth, since the randomistas are explicitly not aiming to increasing growth, while both groups are aiming to increase welfare all things considered.
Overall, I do find it likely that some form of policy interventions (maybe focused on growth, maybe not) will likely pencil out better than the current GW top charities, but I think measuring impacts in terms of raw GDP is likely to be more distracting than beneficial on that path.
If people are interested in taking up the mantle here and pushing these arguments further, I would be interested to see more of a focus on (a) detailed historical cases where outside efforts to improve policy to accelerate growth have worked—I think it’s unfortunate that we don’t have more concrete evidence on the amount of funding or role of the Indian think tank Lant cites; and (b) concrete proposed interventions/things to do. I agree that the “there’s nothing to do” argument is not a show-stopper, but I do think a real weakness of this overall direction of argument at the moment is answering that question. And in general, I expect an inverse relationship between “clarity/convincingness of policy impact/benefit” and “tractability/policy feasibility” (it seems to me that the clearest growth prescription is more immigration to rich countries, and ~equally clear that we shouldn’t expect major policy changes there), so I think getting the argument down into the weeds about what has worked in the past and where opportunities exist now might be more productive.
FWIW, I agree with the comment from @cole_haus speculating that part of the reason these arguments haven’t gotten traction in EA is that it seems most people who are willing to bite the “high uncertainty, high upside” bullet tend to go further, towards animals or the far future, rather than stop at “advocate for policies to promote growth.”
Again, just want to reiterate that I think this is an interesting and worthwhile question and that I’m sympathetic to the case that EAs should focus more on policy interventions writ large.
(Also, not sure I got formatting right, so let me know if links don’t work and I can try to edit—thanks!)
Re your last paragraph, I just wanted to drop @jefftk’s (IMO) amazing post here: https://www.jefftk.com/p/candy-for-nets
[Question] Post on maximizing EV by diversifying w/ declining marginal returns and uncertainty
Really liked this post, thanks.
Minor comment, wanted to flag that I think “Open Philanthropy has also reduced how much they donate to GiveWell-recommended charities since 2017.” was true through 2019, but not in 2020, and we’re expecting more growth for the GW recs (along with other areas) in the future.
[Question] Annual donation rituals?
Technical Updates to Our Global Health and Wellbeing Cause Prioritization Framework
Thanks, I thought this was interesting!
This question you called out in “Relevance” particularly struck me: “More concretely, it could help us estimate the potential market size of effective altruism. How many proto-EAs are there? Less than 0.1% of the population or more than 20%?”
How would you currently answer this question based on the research you report here?
If a five or higher on both scales is one way to operationalize proto-EA (you said 81% of self-ID’d EAs had that or higher), do you think the NYU estimates (6%?) or MTurk estimates (14%?) are more representative of the “relevant” population?
This isn’t an answer to the question, but two additional considerations I think you’re missing that point the opposite direction and I think would make AMF look even better than GiveWell counts it as, on the total view:
There’s some evidence that bednets lead to higher fertility and that channels sound somewhat intuitively plausible.
Roodman’s report is only counting the first generation. If preventing two under-5 deaths leads to ~one fewer birth, that’s still one more kid net making it to adulthood and being able to have kids of their own. Given fertility rates in the places where AMF works, I think that would more than offset the Roodman adjustment in just a few decades.
I haven’t seen any concerted public attempts to grapple with these factors or to think about how to bound the considerations in #2 in a way that would lead to a numerical answer to your question. I would be curious if people have other references to point to on this.
Agree that the paper leaves open the ultimate impact on completed fertility and on your #3. On #2 - I think it would be a mistake to try to adjust for this and neglect long run effects, as in your estimate in fn1.
I wrote a long twitter thread with some replies here FWIW: https://twitter.com/albrgr/status/1532726108130377729
Update on Our Planned Allocation to GiveWell’s Recommended Charities in 2022
GiveWell could answer more confidently but FWIW my take is:
-December 2022 is totally fine relative to today.
-I currently expect this increase in marginal cost-effectiveness to persist in future years, but with a lot of uncertainty/low confidence.
Thanks for the thorough engagement, Michael. We appreciate thoughtful critical engagement with our work and are always happy to see more of it. (And thanks for flagging this to us in advance so we could think about it—we appreciate that too!)
One place where I particularly appreciate the push is on better defining and articulating what we mean by “worldviews” and how we approach worldview diversification. By worldview we definitely do not mean “a set of philosophical assumptions”—as Holden writes in the blog post where he introduced the concept, we define worldviews as:
a set of highly debatable (and perhaps impossible to evaluate) beliefs that favor a certain kind of giving. One worldview might imply that evidence-backed charities serving the global poor are far more worthwhile than either of the types of giving discussed above; another might imply that farm animal welfare is; another might imply that global catastrophic risk reduction is. A given worldview represents a combination of views, sometimes very difficult to disentangle, such that uncertainty between worldviews is constituted by a mix of empirical uncertainty (uncertainty about facts), normative uncertainty (uncertainty about morality), and methodological uncertainty (e.g. uncertainty about how to handle uncertainty, as laid out in the third bullet point above).
We think it is a mistake to collapse worldviews in the sense that we use them to popular debates in philosophy, and we definitely don’t aim to be exhaustive across worldviews that have many philosophical adherents. We see proliferation of worldviews as costly for the standard intellectual reason that they inhibit optimization, as well as carrying substantial practical costs, so we think the bar for putting money behind an additional worldview is significantly higher than you seem to think. But we haven’t done a good job articulating and exploring what we do mean and how that interacts with the case for worldview diversification (which itself remains undertheorized). We appreciate the push on this and are planning to do more thinking and writing on it in the future.
In terms of disagreements, I think maybe the biggest one is a meta one about the value of philosophy per se. We are less worried about internal consistency than we think it is appropriate for philosophers to be, and accordingly less interested in costly exercises that would make us more consistent without carrying obviously large practical benefits. When we encounter critiques, our main questions are, “how would we spend our funding differently if this critique were correct? How costly are the deviations that we’re making according to this critique?” As an example of a case where we spent a lot of time thinking about the philosophy and ended up thinking it didn’t really have high utility stakes and so just deprioritized it for now, see the last footnote on this post (where we find that the utility stakes of a ~3x increase in valuations on lives in some countries would be surprisingly small, not because they would not change what we would fund but because the costs of mistakes are not that big on the view that has higher valuations). You mentioned being confused by what’s going on in that sheet, which is totally fair—feel free to email Peter for a more detailed explanation/walkthrough as the footnote indicates.
In this particular writeup, you haven’t focused as much on the upshot of what we should fund that we don’t (or what we do fund that we shouldn’t), but elsewhere in your writing I take your implication to be that we should do more on mental health. Based on my understanding of your critiques, I think that takeaway is wrong, and in fact taking on board your critiques here would lead us to do more of what most of OP Global Health and Wellbeing already does—save kids’ lives and work to fight the worst abuses of factory farming, potentially with a marginal reduction in our more limited work focused on increasing incomes. Three particular disagreements that I think drive this:
Set point. I think setting a neutral point on a life satisfaction scale of 5⁄10 is somewhere between unreasonable and unconscionable, and OP institutionally is comfortable with the implication that saving human lives is almost always good. Given that we think the correct neutral point is low, taking your other points on board would imply that we should place even more weight on life-saving interventions. We think that is plausible, but for now we’ll note that we’re already really far in this direction compared to other actors. That doesn’t mean we shouldn’t go further, but we do think it should prompt some humility on our part re: even more extreme divergence with consensus, which is one reason we’re going slowly.
Hedonism. We think that most plausible arguments for hedonism end up being arguments for the dominance of farm animal welfare. We seem to put a lot of weight on those arguments relative to you, and farm animal welfare is OP GHW’s biggest area of giving after GiveWell recommendations. If we updated toward more weight on hedonism, we think the correct implication would be even more work on FAW, rather than work on human mental health. A little more abstractly, we don’t think that different measures of subjective wellbeing (hedonic and evaluative) neatly track different theories of welfare. That doesn’t mean they’re useless—we can still learn a lot when noisy measures all point in the same direction—but we don’t think it makes sense to entrench a certain survey-based measure like life satisfaction scores as the ultimate goal.
Population ethics. While we’re ambivalent about how much to bet on the total view, we disagree with your claim that doing so would reduce our willingness to pay for saving lives given offsetting fertility effects. As I wrote here, Roodman’s report is only counting the first generation. If he is right that preventing two under-5 deaths leads to ~one fewer birth, that’s still one more kid net making it to adulthood and being able to have kids of their own. Given fertility rates in the places where we fund work to save lives, I think that would more than offset the Roodman adjustment in just a few decades, and potentially cumulatively lead to much higher weight on the value of saving kids’ lives today (though one would also have to be attentive to potential costs of bigger populations).
Related to the point about placing less weight on the value of philosophy per se, we’re reluctant to get pulled into long written back and forths about this kind of thing, so I’m not planning to say more on this thread by default, but happy to continue these discussions in the future. And thanks again for taking the time to engage here.
- Open Phil Should Allocate Most Neartermist Funding to Animal Welfare by (19 Nov 2023 17:00 UTC; 489 points)
- Winners of the EA Criticism and Red Teaming Contest by (1 Oct 2022 1:50 UTC; 226 points)
- Prioritising animal welfare over global health and development? by (13 May 2023 9:03 UTC; 107 points)
I also hadn’t seen these slides, thanks for posting! (And thanks to Michael for the post, I thought it was interesting/thought-provoking.)
Hi MHR,
I really appreciate substantive posts like this, thanks!
This response is just speaking for myself, doing rough math on the weekend that I haven’t run by anyone else. Someone (e.g., from @GiveWell) should correct me if I’m wrong, but I think you’re vastly understating the difficulty and cost of running an informative replication given the situation on deworming. (My math below seems intuitively too pessimistic, so I welcome corrections!)
If you look at slide 58 here you get the minimum detectable effect (MDE) size with 80% power can be approximated as 2.8*the standard error (which is itself effectively inversely proportional to the square of the sample size).
I didn’t check the original sources, but this GiveWell doc on their deworming replicability adjustment implies that the standard error for log(income/consumption) in the most recent replications is ~.066 (on a “main effect” of .109). The original RCT involved 75 schools, and according to figure A1 here the followup KLPS 4 involved surveying 4,135 participants in the original trial. GiveWell’s most recent cost-effectiveness analysis for Deworm the World makes 2 key adjustments to the main effect from the RCT:
A replicability adjustment of .13 (row 11)
A geography-specific adjustment for worm burden which averages about .12 (row 40) (this is because worm burdens are now much lower than they were at the time of MK)
Together, these adjustments imply that GiveWell projects the per-capita benefit to the people dewormed to be just .13*.12=1.56% of the .109 impact on log income in the late followups to the original Miguel and Kremer RCT. So if we wanted to detect the effect GiveWell expects to see in mass deworming, we’d have an MDE of ~.0017 on log income, which with 80% power and the formula above (MDE=2.8*standard error) implies we’d need the standard error to be .0017/2.8=~.00061 log points. So a well-powered study to get the effect GiveWell expects would need a standard error roughly 108 times smaller than the standard error (.066) GiveWell calculates on the actual followup RCTs.
But because standard errors are inversely proportional to the square root of sample size, if you used the same study design, getting a 108x smaller standard error would require a 108*108=11,664 times larger sample. I think that might imply a sample size of ~all the elementary schools in India (11,664*75=874K), which would presumably include many schools that do not in fact actually have significant worm burdens.
If the original MK study and one followup cost $1M (which I think is the right order of magnitude but may be too high or too low), this implies that a followup powered to find the effect GiveWell expects would cost many billions of dollars. And of course it would take well over a decade to get the long term followup results here. (That said, it wouldn’t surprise me if I’m getting the math wrong here—someone please flag if so!)
I’m sure there are better study designs than the one I’m implicitly modeling here that could generate more power, or places where worm burdens are still high enough to make this somewhat more economical, but I’m skeptical they can overcome the fundamental difficulty of detecting small effects in cluster RCTs.
I think a totally reasonable reaction to this is to be more skeptical of small cheap interventions, because they’re so hard to study and it’s so easy to end up driven by your priors.
Hi Nicole,
I think this is a cool choice and a good post—thanks for both! I agree with your bottom line that kidney donation can be a good choice for EAs and just wanted to flag a few additional resources and considerations:
I think these other EA forum posts about the costs and benefits of donation are worth checking out. In my mind the most important update relative to when I donated is that the best long-run studies now suggest a roughly 1 percentage point increase in later-life risk of kidney failure because of donating. I think that translates less than 1:1 to mortality for a variety of reasons (ability to get a transplant, maybe xenotransplantation or other things will be easy in 20-50 years) but I think that factor probably swamps the near-term (roughly 1⁄3,000) risk of death in surgery when thinking about the EV calculation.
I think I took ~3 weeks off work to recover from donation (it was also around the holidays for me), and I think for folks who work in altruistic jobs that may dominate the cost calculation. 52 hours seems like a very low estimate of the expected time cost to me all in though.
I think people sometimes assume that the original donor gets full counterfactual “credit” for all the steps in a chain. My read of this evidence is that even though average chain length is ~4, the marginal social value of an altruistic donor starting a chain is “only” ~.8-1.7 transplants (depending on blood type) because the relevant counterfactual can be other chains being longer.
I think things like this post are themselves a pretty important channel for impact. I think the impact of my personal donation was dominated by the small influence I had on getting Dylan Matthews to donate, which then had a big knock-on impact because his writing led a number of other people to donate.
Overall, I think these kinds of persuasion considerations can play a weirdly big role in how you evaluate kidney donation, and I don’t have a clear bottom line on which way they cut.
I don’t have a particularly good estimate on total time, but my impression is that most doctors recommend people plan to take a couple weeks off from office work, which would maybe 2-3x your 52 hr estimate?
Yes, kidney selling is officially banned in nearly every country. My preference, at least in the U.S. context, would be to have the government offer benefits to donors to ensure high quality and fair allocation: http://www.nytimes.com/2011/12/06/opinion/why-selling-kidneys-should-be-legal.html