Hide table of contents

Update 2016-12-14: GiveWell’s 2016 cost-effectiveness analysis has updated the way it handles population ethics. It now explicitly takes the value of saving a 5-year old’s life as input and no longer assumes that it’s worth 36 life-years.

Update 2018-08-14: I recently revisited GiveWell’s 2018 cost-effectiveness analysis. Although the analysis spreadsheet no longer enforces the “GiveWell view” described in this essay, most GiveWell employees still implicitly adopt it. As a result, I believe GiveWell is still substantially mis-estimating the cost-effectiveness of the Against Malaria Foundation.

Cross-posted to my blog.

GiveWell claims that the Against Malaria Foundation (AMF) is about 10 times as cost-effective as GiveDirectly. This entails unusual claims about population ethics that I believe many people would reject, and according to other plausible views of population ethics, AMF looks less cost-effective than the other GiveWell top charities.

GiveWell’s Implicit Assumptions

A GiveWell-commissioned report suggests that population will hardly change as a result of AMF saving lives. GiveWell’s cost-effectiveness model for AMF assumes that saving one life creates about 35 quality-adjusted life years (QALYs), and uses this to assign a quantitative value to the benefits of saving a life. But if AMF causes populations to decline, that means it’s actually removing (human) QALYs from the world; so you can’t justify AMF’s purported cost-effectiveness by saying it creates more happy human life, because it doesn’t.

You could instead justify AMF’s life-saving effects by saying it’s inherently good to save a life, in which case GiveWell’s cost-effectiveness model shouldn’t interpret the value of lives saved in terms of QALYs created/destroyed, and should include a term for the inherent value of saving a life.

GiveWell claims that AMF is about 10 times more cost-effective than GiveDirectly, and GiveWell ranks AMF as its top charity partially on this basis (see “Summary of key considerations for top charities” in the linked article). This claim depends on the assumption that saving a life creates 35 QALYs.

To justify GiveWell’s cost-effectiveness analysis, you could say that it is good to cause existing people to live longer, but it is not bad to prevent people from existing. (Sean Conley of GiveWell says he and many other GiveWell staffers believe this.)

In particular, you’d have to assume that:

  1. It’s valuable to cause a currently-existing person to live longer, but not harmful to prevent people from existing.
  2. The value of causing a currently-existing person to live longer derives entirely from the additional life-years lived.

(Note that (2) implies that saving a life does not have value in itself, but that the value derives from causing a person to get to live additional life-years.)

I will refer to this view as the GiveWell view, although I do not believe all GiveWell employees would necessarily endorse these two assumptions. Rather, I mean that this view implicitly follows from the way GiveWell calculates AMF’s cost-effectiveness. GiveWell’s recommendations include many inputs other than explicit cost-effectiveness calculations, but many donors nonetheless heavily rely on the claim that AMF is about 10 times more cost-effective than GiveDirectly.

GiveWell’s cost-effectiveness calculations not only claim that preventing deaths from malaria is good, but that its goodness comes in proportion to the number of additional life-years lived by people who otherwise would have died from malaria.

A Paradox

This position leads to counter-intuitive and possibly contradictory conclusions. Let’s look at a thought experiment to see why.

You can choose between three possible acts. When you perform act1 A, a child (let’s call her Afiya) is born, gets malaria, and dies. Under act B, you cause Afiya not to be born. According to assumption (1), this act is not worse than A. Standard person-affecting view says that it is not wrong to cause someone to exist whose life is net positive, so A is not worse than B. Under act C, you cause Afiya to be born and prevent her from getting malaria. This beats act A according to (2), and is not better than act B according to (1). Thus, A = B, B ≥ C, and C > A. But this creates a contradiction: B > A and B = A.

This same concern applies to any person-affecting view. I believe the most plausible resolution of this paradox comes from Roberts (2003)2: It is not wrong to fail to bring someone into existence, but it is wrong to bring someone into existence who will be less happy than they could be. So you ought not allow Afiya to be born and contract malaria when you could have prevented her from existing or caused her not to contract malaria.In other words, B > A, C > A, and B ≥ C.

For a standard person-affecting view, this successfully avoids creating a contradiction (although I and many others find it unsatisfying). However, the GiveWell view has additional problems that a standard person-affecting view does not, because the GiveWell view includes the additional assumption that the value of causing a currently-existing person to live longer derives entirely from the additional life-years lived.

By assumption (2), since Afiya gains an additional 35 QALYs when she doesn’t get malaria, we can say that C = A + 35 (that is, act C is “35 QALYs” better than act A). Since B ≥ C, that means B ≥ A + 35. In other words, if we prevent Afiya from existing, that’s at least 35 QALYs better causing Afiya to be born and then letting her die of malaria.

This might not appear explicitly contradictory, it nonetheless comes across as very weird. How can act B be worth 35 life-years “better” than act A if it in fact contains fewer happy life-years?

In short, to avoid internal contradiction, the GiveWell view must claim causing Afiya not to be born is worth at least 35 QALYs “more” than causing her to be born and then die of malaria. (There exist a few other possibilities, such as that transitivity does not hold for orderings over acts, but I find these sufficiently implausible that I will not address them here.)

A couple of commenters have proposed that a child’s life may be net negative on balance when she dies of malaria at a young age. This still does not avoid the problem with the GiveWell view where the difference between the child’s existence and nonexistence exactly equals the value of the additional life-years she would have if she didn’t get malaria. And, more significantly, this does not resolve the additional problems with the GiveWell view described in the next section.

Bigger Problems

We can actually make a stronger claim than this: the GiveWell view appears to explicitly contradict Roberts’s resolution to the person-affecting view. Showing this gets a bit complicated, so bear with me.

In this scenario, there are two children, Afiya and Brian, who might potentially get malaria. You may take one of four possible acts.

  • Act A: Afiya gets malaria and dies. Brian is then born, gets malaria, and dies.
  • Act B: You prevent Afiya from getting malaria. Brian is still born and dies of malaria.
  • Act C: You prevent Afiya from getting malaria and Brian is never born.
  • Act D: Both Afiya and Brian are born, and you prevent them both from getting malaria.

3

According to the GiveWell view, A < B and B < D.

Under the standard person-affecting view, this creates the same paradox as our previous thought experiment. But even Roberts’s resolution does not work here for the GiveWell view. Roberts’s resolution entails that C = D and therefore B < C. But GiveWell’s cost-effectiveness calculations for AMF implicitly endorse B = C. When we donate to AMF and save an Afiya, standard cost-effectiveness estimates for AMF don’t give any regard to whether a Brian exists, even though under Roberts’s view we should prefer Brian not to exist. If donating to AMF looks like act B, then saving a life is worth 35 QALYs; but if it looks like act C, then saving a life is worth 70 QALYs. Similarly, acts C and D count for the same under Roberts’s resolution, even though GiveWell would count act D as twice as good as act C.

The GiveWell view assumes that the benefits of preventing Afiya from getting malaria come entirely from causing her to live longer. The GiveWell view is indifferent about whether saving Afiya counts as act B, C, or D. In other words, B = C = D. But at the same time, D > B because the GiveWell view wants us to save Brian. This is not just controversial--it's internally inconsistent.

Some philosophers still accept something like the GiveWell view in spite of its problems4. But even if we admit the plausibility of the GiveWell view, I do not believe GiveWell should claim that AMF is 10 times more cost-effective than GiveDirectly. This claim relies on fairly specific assumptions about population ethics that many potential donors will reject. If donors understood these assumptions, I expect that many of them would prioritize their donations differently. For example, last year when Stanford Effective Altruism was considering making donations to charity, we preferred the Schistosomiasis Control Initiative over AMF because we believed that GiveWell gave too much significance to the “GiveWell view” of population ethics and not enough to the total view. Had we not understood the significance of population ethics in GiveWell’s recommendation, we might have prioritized AMF even though our actual beliefs about population ethics would imply that the other top charities look better.

Attempted Resolutions

The GiveWell view has undesirable implications for population ethics. This gives rise to two concerns: (i) donors may end up following recommendations that contradict their views about population ethics; and (ii) GiveWell is probably in fact wrong about its cost-effectiveness calculations, which count for a large part of its final charity recommendations. Concern (i) suggests that GiveWell should make its assumptions about population ethics more explicit, and concern (ii) suggests that GiveWell should change its cost-effectiveness estimates and possibly its top charity recommendations.

I believe that even people who endorse the GiveWell view should agree with concern (i). On concern (ii), I see two strong candidates for how to resolve the problems with the GiveWell view:

  1. Reject assumption (1)–probably by adopting a total view, which claims that it is good rather than neutral to create new happy people.
  2. Reject assumption (2), perhaps by claiming that preventing deaths has value beyond just the additional life years created. This would bring us closer to something like preference utilitarianism with a person-affecting view of population ethics.

Let’s see if these produce similar counterintuitive conclusions to the GiveWell view, and how we would prioritize charities differently if we adopted them.

First Resolution

If we follow the first resolution, the value of AMF derives from its global effect on increasing happiness and reducing suffering. Because AMF hardly changes humans’ lifespans, it does not have a clear beneficial effect for humans (and it’s unclear whether creating more happy human life benefits or harms sentient life in general). So under this resolution, AMF looks much less effective. I personally prefer this view, as do many people at Stanford Effective Altruism (as mentioned above); and as do many philosophers for that matter5. Thus, I substantially discount the life-saving benefits of AMF when prioritizing GiveWell top charities, and I believe GiveWell ought to as well.

Second Resolution

The second resolution may have even worse implications than the GiveWell view. Suppose we reject assumption (2) by claiming that preventing death has extra value beyond its effect on someone’s lifespan. (Let’s call this the updated person-affecting view.) Returning to our initial thought experiment, the differential between act A (in which a child is born and dies of malaria) and act C (in which a child is born and prevented from getting malaria) appears bigger under the updated person-affecting view than under the GiveWell view. But we still cannot say that we prefer act C to act B, which means now instead of having B ≥ A + 35, we have B > A + 35.

We can avoid this problem by claiming that the life-years added by act C have sub-linear value with the length of life. This view appears more plausible, but it means we have no easy way to directly compare AMF against GiveDirectly (or against other charities that serve to improve people’s lives rather than prevent them from dying). We would need to create some new way to trade off preventing deaths against improving lives, so our cost-effectiveness calculations for AMF would look substantially different.

Conclusion

Cost-effectiveness estimates for GiveWell top charities, and therefore GiveWell’s recommendations6, depend heavily on controversial questions about population ethics. I believe GiveWell takes an incorrect stance here, and it ought to give more weight to the total view. But more importantly, donors should be aware of how questions of population ethics affect the expected value of different interventions.

To GiveWell’s credit, it has written about how its cost-effectiveness estimates require making judgment calls. However, if I am reading them correctly, the cost-effectiveness analyses done by different GiveWell employees all uniformly assume that saving one life with malaria nets has the same value as adding ~35 QALYs, which means GiveWell’s cost-effectiveness spreadsheet enforces a single judgment call on population ethics (i.e., what I have called the GiveWell view). GiveWell’s list of key assumptions on AMF does not include anything about population ethics.

A large part of the case for AMF rests on the fact that it appears highly cost-effective when you make certain assumptions about population ethics. Donors who do not make these assumptions, or who have substantial uncertainty about them, should consider which charities look best on their own view of population ethics.

Errata: I originally used DALYs instead of QALYs because GiveWell’s cost-effectiveness estimates use DALYs. I was unaware that DALYs have the opposite sign of QALYs. I changed this post to use QALYs instead of DALYs.

Notes

  1. I originally had this evaluating world states rather than acts, but Eitan Fischer pointed out that the world-orientation creates more substantial problems than the act-orientation.

  2. Roberts, M. A. (2003). Is the person-affecting intuition paradoxical?. Theory and Decision, 55(1), 1-44. http://link.springer.com/article/10.1023%2FB%3ATHEO.0000019052.80871.b3

  3. As you can clearly see, I am not a graphic designer.

  4. All ethical views have counterintuitive parts; it’s just a matter of choosing which of your intuitions you want to violate.

  5. The most common objection to the total view is the repugnant conclusion, although I do not believe that the repugnant conclusion is in fact repugnant.

  6. This is not to say that recommendations are entirely determined by cost-effectiveness estimates, but that such estimates are a major input.

Comments52
Sorted by Click to highlight new comments since: Today at 10:11 AM

(Edit: I no longer endorse suffering-focused ethics.)

Regardless of your stance on population ethics, I think in general it makes sense to take DALYs as a heuristic for how much good you can do with your money. Clearly all population ethical views consider improving existing lives in quality (decreasing YLDs, years lived with disability) a good thing. Preventing deaths expressed through reducing YLLs (Years of Life Lost) is probably overall good as well, although different views will assign more or less value to it. I agree with Michael Dickens that if the value of longer lives comes from adding life-years (reducing YLL) alone, this would indeed amount to something like total utilitarianism.

I think a steelman of GiveWell's view would be that in fact the YLL component of DALYs can be motivated by some other things, like preference dissatisfaction or decreasing the suffering of the parents of children as well. I believe that for reasons of cooperation between agents it always makes sense to consider the preferences of other beings at least to some degree. Fulfilling already existing preferences seems like something most people would agree to, whether they would also like to bring additional fulfilled preferences into existence or not. Therefore, death is intrinsically bad according to most reasonable views, since it violates the preferences of existing beings severely. In that sense, decreasing YLLs should be always good, even for non-classical utilitarians.

Unlike Michael, I personally would be less reluctant to accept a ranking of world states that can’t be boiled down to an easy mathematical function of the aggregated wellbeing, i.e. I’d be less turned off by more “complex” moral views. And I’d be less willing to bite bullets like the repugnant conclusion, or the “very repugnant conclusion,” where a world with fewer, but very happy individuals can be worse than a world containing any finite amount of extreme torture that is outweighed by an even greater amount of beings that live lives just barely worth living. Accepting this conclusion is a quite a controversial stance in my eyes. Given anti-realism, it is absolutely unclear to me why GiveWell would have to adhere to a total utilitarian view. They could very well accept all the inconsistencies Michael mentions and still just maximize EV according to their own (complex) values. I agree that they should probably specify their view more explicitly and it remains unclear what they are really optimizing for (see also http://blog.givewell.org/2008/08/22/dalys-and-disagreement/).

A candidate I am favouring that could possibly match a lot of people's intuitions would be something like negative idealized preference utilitarianism or more generally any form of suffering-focused ethics (e.g. trying to reduce extreme involuntary suffering without doing anything crazy or anything that would be considered really bad by other agents).

(cross-posted here: https://www.facebook.com/groups/effective.altruists/permalink/1071588459564177/)

Cross-posting my reply:

I believe this is the most plausible attempt at a resolution I've heard so far. Thanks, Johannes.

Like some other responses I've heard, if we accept your proposed view on population ethics, we'd still have to substantially update the common view on the value of AMF. Remember, I'm not saying that YLL's don't have value; I'm saying that it's controversial and probably incoherent to claim that the value of AMF's lives saved equal the (time-discounted) number of additional life-years lived.

If the importance of YLL's comes from the suffering of parents, as you suggest, YLL's will look really different than just one DALY per year of life lost. If we adopt more of a preference-utilitarian view, we end up producing contradictory conclusions in the same scenarios that I discussed in my original essay—you can't claim that AMF saves 35 DALYs without knowing AMF's population effects.

They could very well just accept all the inconsistencies Dickens mentions and still just maximize EV according to their own (complex) values.

If you're inconsistent, you cannot coherently maximize EV. You can only maximize EV if you can apply a unique real-valued EV function over states or actions, and such a function only exists in a consistent system.

(Edit: I no longer endorse negative utilitarianism or suffering-focused ethics.)

Thank you! Cross-posting my reply as well:

If we adopt more of a preference-utilitarian view, we end up producing contradictory conclusions in the same scenarios that I discussed in my original essay—you can't claim that AMF saves 35 DALYs without knowing AMF's population effects.

Shouldn't this be fixed by negative preference utilitarianism? There could be value in not violating the "preference-equivalent" of dying one year earlier, but no value in creating additional "life-year" preferences. A YLL would be equivalent to a violated life-preference, then. You could avert YLLs by not having children, of course, which seems plausible to me (if noone is born, whose preference is violated by dying from Malaria?). Being born and dying from Malaria would be worse than non-existence, so referring to your "Bigger Problem"-scenarios, A < B < C and C = D.

Regarding EV: I agree, there has to be one ranking mapping world-states onto real numbers (or R^n if you drop the continuity-axiom). So you're right in the sense that the supposed GiveWell-ranking of world-states that you assume doesn't work out. I still think that there might be a way to make a creative mapping in the real world so that the GiveWell focus on DALYs without regarding population size can be somehow translated into a utility function. Anyway, I would kind of agree that AMF turns out to be less effective than previously thought, both from an SFE and a classical view smile emoticon

One thing that seems noteworthy is the fact that the population effect actually brings people closer together than they were before: Ignoring population effects, AMF has high impact from a CU perspective but low impact from a suffering-focussed perspective; accounting for population effects, the difference almost vanishes. Another way of looking at it: In situations where the population remains constant, population ethics becomes irrelevant.

So accounting for population effects mainly gives us these two updates:

  1. Population-ethical views become less relevant for prioritisation between various GiveWell charities (and not more relevant, as some seemed to suggest (possibly with the exception of the negative preference view)).
  2. AMF might be less effective than deworming charities according to most population-ethical views (but still more effective than cash transfers due to developmental effects of malaria prevention).

If we consider wild-animal suffering, I think AMF looks better than charities that don't create as many human lives. This could once again make AMF more cost-effective according to many population-ethical views (unless you consider wild insects to have good lives on average).

They could very well accept all the inconsistencies Michael mentions and still just maximize EV according to their own (complex) values.

Excuse me, what does EV stand for?

EV stands for Expected Value. (I think I actually meant Expected Utility more precisely)

I thought you might like to know that, after a few days' thought and discussion, this reminder of my population ethics concerns has made me switch my donations from AMF to SCI; I originally favoured this over life-saving interventions for population ethics reasons back in 2010, six years ago. (My concerns are slightly broader ones about the difficulty of knowing how and why deaths are bad - without doubting that they are - but I don't have time to go into them any time soon.)

This is sad from my point of view, because increasing the human population plausibly prevents lots of bug suffering, while increasing quality of life does not (at least not as clearly; the sign of impact of quality-of-life improvements on bug suffering is more ambiguous).

This stuff is genuinely tricky, and I'm deeply aware of that! Does deworming increase population though? My guess is it might even decrease it. Remember that my concerns are slightly different than Michael's, and focused on the general difficulty of knowing how and why deaths are bad, making me want to prioritise easing suffering (and increasing welfare).

But more importantly, donors should be aware of how questions of population ethics affect the expected value of different interventions.

Thank you for emphasizing this--I think it's very important.

I've realized lately that my views on questions of population ethics are very underdeveloped, which is problematic because it leaves me very uncertain about the relative importance of different causes and the expected value of different interventions, which leads me to postpone donating more until I have better information (and also possibly leads me to not engage in direct-impact work that I possibly should be engaging in due to not knowing what that work that I should be engaging in is).

Note that because questions of population ethics can change the expected value of possible interventions from positive to negative (or vice versa) and by orders of magnitude rather than just a few percentage points, my lack of confident answers to questions of population ethics seems to be a good reason to postpone making any further donations until I have better information on my views on those questions.

If donors understood these assumptions, I expect that many of them would prioritize their donations differently.

I wonder... To what extent it is true that donor ignorance about their views on questions of population ethics (and related questions about their values) leads donors to confidently choose one charity or intervention over another when in fact if they understood their views on population ethics correctly then they would have chosen the other charity or intervention?

I used to think that I knew what I valued well enough to choose where to donate, but now I realize that I have to think more on certain questions of population ethics to at least figure out what approximate probability I would assign to each possible way of valuing things before I can know which cause and intervention I believe has the highest expected value and is worth donating to.

Here's a third resolution. Consider a utility function that is a weighted sum of:

  1. how close a region's population level is to the "ideal" population level for that region (i.e. not underpopulated or overpopulated)
  2. average utility of individuals in this region (not observer-moments in this region)

AMF is replacing lots of lives that are short (therefore low-utility) with fewer lives that are long (therefore higher utility), without affecting population level much. The effect of this could be summarized as "35 DALYs", as in "we increased the average lifespan by 35 DALYs / total population".

(warning: made-up numbers follow). Suppose we make someone live for 40 years instead of 5 years by curing malaria. This reduces the fertility rate; let's say one fewer 35-year life happens as a result. This has no effect on the average population level (part 1). We replaced a 5-year life plus a 35-year life with a 40-year life. If average lives in the region are 35 years long (and we're pretending that life utility = length of life), then most of the effect on part 2 of the utility function comes from preventing a worse-than-average life from happening.

Suppose instead that we extend someone's life from 40 years to 75 years (a gain of 35 DALYs). This reduces the fertility rate; let's pretend that this prevents a 35-year life from happening. So we're replacing a 40-year life plus a 35-year life with a 75-year life. From the perspective of part 2 of the utility function, this is exactly as good as curing a case of malaria. So it seems like you can measure life-extending measures in DALYs pretty naively and things work out (both 35-DALY improvements are equally good under the utility function).

GiveWell claims that AMF is about 10 times more cost-effective than GiveDirectly, and GiveWell ranks AMF as its top charity partially on this basis (see “Summary of key considerations for top charities” in the linked article). This claim depends on the assumption that saving a life creates 35 QALYs.

I'm sorry but I didn't have time to read this post in full and can't comment on the philosophical issues. I also cannot speak for GiveWell and the following is just my personal interpretation of their cost-effectiveness analysis.

I think the statement above is a based on a misunderstanding of Givewell's cost-effectiveness analyses.

GiveWell writes here: "We do include possible developmental impacts on children who sleep under an LLIN: we estimate that sleeping under an LLIN provides the same developmental impacts that a deworming pill provides. However, most of the benefit provided by LLINs is in the lives that LLINs save, not in their impact on development." http://www.givewell.org/international/top-charities/amf#Costperlifesaved

I think this is what the 10x better than GiveDirectly refers to. Deworming charities have a similar effect on development and are also ranked as about 10x as good as cash-transfers and rarely cause mortality, which leads me to believe that the child mortality reduction is not included in that part of the cost-effectiveness analysis, but rather modelled separately.

Have you looked at GiveWell's cost-effectiveness model? I don't believe that's correct. GiveWell estimates the cost-effectiveness of bednets as

((long-term benefits to income) / (cost per person-year of protection * relative development benefits)) + ((DALYs per life * conversion factor from DALYs to income) / (cost per life saved))

(making some simplifying assumptions but that's the gist of it.)

So it's a sum of the development benefits to income and the life-saving benefits. Estimates vary between employees, but on average the life-saving benefits account for about 2/3 of the total benefits. The claim that AMF is 10x more cost-effective than GiveDirectly comes from the total figure, and most of that comes from life-saving benefits. Therefore the cost-effectiveness estimate for AMF critically depends on how you estimate the value of saving lives.

And anyway, many people cite the benefits of AMF as "saving a life for $3000" or something along those lines, which falls prey to the problems I discuss here.

Sorry I was being unclear. The part of the equation that you call 'Life saving benefits' (e.g. ((DALYs per life * conversion factor from DALYs to income) / (cost per life saved)) is only of instrumental value - it crucially depend on the conversion factor of from DALYs to income (if you were to set it to zero so that there would be no increase in income due to the morbidity and mortality, the life saving benefits would be zero). So I believe there is no intrinsic valuing of QALYs/Life in the cost-effectiveness model that suggests that bednets are 10 times more effective than cash-transfers, as I thought your argument implies. Rather I believe cost-per-life saved is modelled separately.

It sounds like you're saying the conversion factor from DALYs to income is based on how much more income people have when their lives are saved, but that's not correct. The conversion factor from DALYs to income is written on GiveWell's spreadsheet as "1 DALY is equivalent to increasing ln(income) by one unit for how many years". In other words, how good is increasing income relative to extending someone's life? This number is determined by taking the median of value judgments by about a dozen different GiveWell staff members. So what I said originally is correct.

Thanks for writing this up; this is important to think about.

Standard person-affecting view says that it is not wrong to cause someone to exist whose life is net positive, so A is not worse than B

GiveWell's data on deaths prevented is entirely based on children under age 5. I don't think we have good information on children's subjective experience, and I'm not totally willing to assume that their lives are particularly enjoyable. Being a newborn in particular seems pretty unpleasant to me, and I could easily believe that the experience of living to age 1 month and then developing a fatal case of malaria is net negative. But given that happiness by age seems generally U-shaped, perhaps it's reasonable to assume that being a toddler is pretty fun.

That's an important point. It still doesn't resolve the problems with the GiveWell view though:

A couple of commenters have proposed that a child’s life may be net negative on balance when she dies of malaria at a young age. This still does not avoid the problem with the GiveWell view where the difference between the child’s existence and nonexistence exactly equals the value of the additional life-years she would have if she didn’t get malaria. And, more significantly, this does not resolve the additional problems with the GiveWell view described in the next section.

Interesting post, thanks! Updating somewhat after reading it.

One concern I had about this claim:

When you perform act 1 A, a child (let’s call her Afiya) is born, gets malaria, and dies. Under act B, you cause Afiya not to be born. According to assumption (1), this act is not worse than A. Standard person-affecting view says that it is not wrong to cause someone to exist whose life is net positive, so A is not worse than B.

To me it seems that a child being born and then immediately dying of malaria is net negative for the family of the child, in terms of suffering due to loss of a child, compared to the child not being conceived at all. So from this perspective, B<A. If you think this makes sense to you, how does this affect your calculations?

This was my first thought too. Since malaria is most deadly to young children and pregnant women, I'd want to consider the effects on Celeste, who watches her two-year-old die, and Dayo, who watches his pregnant wife die. As the parent of a toddler, I can say that investing a lot of resources into her and then watching her die doesn't seem like a net positive, even considering that she has enjoyed much of her two years so far.

While GiveWell's calculations around malaria are based on the lives saved (not other benefits like preventing illness or benefits to family members of not having their loved ones die), this consideration makes it seem very reasonable to me to consider that preventing deaths of people who are already known and loved is a good thing in itself.

I agree that this matters, but my argument just considers the effects on the primary individual, not effects on the parents, economic flow-through effects, etc. Similarly, cost-effectiveness calculations for AMF don't typically include a term for parental suffering.

Even so, the suffering of birth, death and perhaps of being an inarticulate infant can be high enough to make it net-negative for the primary individual. It can even be high enough to outweigh 35 average human life-years if it is severe enough and/or the average experience value of one life-year is low enough or negative.

Even if this isn't true for the majority, it can still be true for the average, e.g. if 1% of life-years contain unusual suffering 100 times as severe as 1 life-year is good.

Ok, I see how for the sake of your argument of specifically trying to optimize GiveWell rankings, this point may not seem central. Still, this may be part of what you call the "GiveWell view," consciously or unconsciously - I'd be interested in what GiveWell folks have to say about it.

On "A Paradox":

" According to assumption (1), this act is not worse than A. Standard person-affecting view says that it is not wrong to cause someone to exist whose life is net positive, so A is not worse than B. Under act C, you cause Afiya to be born and prevent her from getting malaria. This beats act A according to (2), and is not better than act B according to (1). Thus, A = B, B ≥ C, and C > A. But this creates a contradiction: B > A and B = A. "

This argument appears to assume completeness, but it's far from clear that those who believe that adding good lives does not make an outcome better should accept completeness. (Broome 2005, "Should We Value Population?", shows that they should not, provided they accept transitivity and the sort of choice-set independence implicitly assumed here).

I don't have a particularly good understanding of population ethics and I haven't read Broome (2005) yet, so I could be off base here. But it seems to me that when GiveWell recommends AMF as a top charity, this requires claiming that AMF is in principle comparable to other charities, which requires completeness (or, at least, completeness over the set of charities being compared).

I could also argue that rejecting completeness seems borderline nonsensical, but that's more complicated to argue, and I don't really have anything original to contribute on the subject.

Commenting here to raise a counter-argument to one of my claims that no one has brought up yet. In the section "Bigger Problems", I claim that the cost-effectiveness calculations implicitly endorse that acts B, C, and D are equivalent. I don't think this is necessarily true though. Just because a cost-effectiveness calculation doesn't include something doesn't mean that thing doesn't matter. The calculations also don't include flow-through effects on the economy, or on factory farmed animals, or on lots of other things. So you could say that the GiveWell view endorses B < D and B < C, and then just doesn't include this in the cost-effectiveness calculations.

I think this is right. People can have different opinions about whether it's good to prevent (or cause) a birth. (Like I said, the mainstream consensus tends to relate to these issues to reproductive freedom.) But GiveWell isn't weighing in on this, because of both empirical uncertainty and ethical ambivalence.

I haven't seen GiveWell emphasize that AMF is 10x as cost-effective as GiveDirectly. In fact, they frequently say that cost-effectiveness estimates shouldn't be taken literally. Nor have I seen them say that they're consequentialist. Many of their heuristics are roughly consequentialist, but that doesn't mean that they endorse strict utilitarianism. It's unclear what you mean by "controversial", but the GiveWell view is far more popular among the general public (and probably even among moral philosophers) than the total view of utilitarianism.

If you reject pure consequentialism, then there's no paradox. Non-consequentialist views tolerate path-dependence. It can be morally neutral to go from 99 happy people to 100 happy people, but morally bad to go from 100 happy people to 99 (via preventable death).

You're free to disagree with GiveWell's values, but calling them paradoxical or contradictory is unjustified. Rejecting an argument's premise isn't the same as refuting the argument.

A few clarifications:

  1. Lots of people rely on the claim that AMF is 10x as cost-effective as GiveDirectly, and lots reject the GiveWell view of population ethics (even if more people accept it). Saying that GiveWell should change its recommendations is a pretty strong claim and I don't know that I would go that far, but I do believe GiveWell should make it more clear how population ethics affects this calculation. This does actually affect some people's decisions—it changed where Stanford Effective Altruism prioritized donations, as I discuss in the article, and it changed Tom Ash's mind.

  2. What I'm calling the GiveWell view is the view that you need to adopt for GiveWell's cost-effectiveness calculations to hold water, not necessarily the view that people at GiveWell endorse. See this quote from the original article:

I will refer to this view as the GiveWell view, although I do not believe all GiveWell employees would necessarily endorse these two assumptions. Rather, I mean that this view implicitly follows from the way GiveWell calculates AMF’s cost-effectiveness. GiveWell’s recommendations include many inputs other than explicit cost-effectiveness calculations, but many donors nonetheless heavily rely on the claim that AMF is about 10 times more cost-effective than GiveDirectly.

  1. I'm not calling this view contradictory because I reject its premises, I'm calling it contradictory because it's contradictory. See the section "Bigger Problems." Do you believe my reasoning here is incorrect? If so, please explain why, or explain why path-dependence resolves the problem and doesn't conflict with the premises.

Also worth reiterating that I'm not criticizing GiveWell's values in general, I'm criticizing the way that GiveWell does cost-effectiveness calculations and the way people interpret them.

I'm not sure where you're getting the idea that GiveWell or GiveWell's donors would be indifferent between B, C, and D in "Bigger Problems". I think most people would agree that D is best. The difference between B and C would be more controversial, but I think the "mainstream" view is that preventing a birth is good if it's due to more reproductive freedom and bad if it's due to less. If a birth is prevented because a woman gets access to modern contraception, people generally think this is good. If it's prevented because a woman becomes infertile against her will, for example due to malnutrition, people generally think this is bad.

If AMF has a negative effect on population, which is debatable, I assume it would be because women feel like they can have fewer children because they won't "lose" as many. This seems to fall under more reproductive freedom, so mainstream opinion would hold that this is a good thing.

I don't think GiveWell's estimates of population effects were supposed to figure into the QALY estimates. They were probably intended to refute the common claim that aid leads to overpopulation (an argument often raised by EAs in the form of the meat-eater problem).

Both your original comment and this one are attributing claims to me that I never made, and largely ignoring my actual arguments. I don't know whether it's that you didn't read my post very carefully or that I didn't explain myself well. I believe you could benefit from reading my post again. I'll also try to explain the part where I think we're miscommunicating.

The claim that B = C = D isn't a claim that anyone made. It follows from the assumptions required for GiveWell's cost-effectiveness analysis to be correct, as I spell out in the post. If donors disagree that B = C = D, which I believe many of them do, then they should not agree with the standard claims about AMF's cost-effectiveness. It sounds like you think you're disagreeing with me when you say that most people wouldn't endorse B = C = D, but actually you're agreeing with me.

Does that make sense?

No, I don't understand. Why does AMF cost-effectiveness require indifference between B, C, and D?

According to the GiveWell view, A < B and B < D.

Under the standard person-affecting view, this creates the same paradox as our previous thought experiment. But even Roberts’s resolution does not work here for the GiveWell view. Roberts’s resolution entails that C = D and therefore B < C. But GiveWell’s cost-effectiveness calculations for AMF implicitly endorse B = C. When we donate to AMF and save an Afiya, standard cost-effectiveness estimates for AMF don’t give any regard to whether a Brian exists, even though under Roberts’s view we should prefer Brian not to exist. If donating to AMF looks like act B, then saving a life is worth 35 QALYs; but if it looks like act C, then saving a life is worth 70 QALYs. Similarly, acts C and D count for the same under Roberts’s resolution, even though GiveWell would count act D as twice as good as act C.

In other words, the cost-effectiveness analysis assumes that B, C, and D are all exactly 35 DALYs better than A, which means B = C = D. It also makes some other assumptions that contradict this assumption. Basically no matter what your view of population ethics, you have to look at population effects to determine how good it is to save a life, and GiveWell's cost-effectiveness analysis doesn't do that.

No, if you were to just count lives saved, D is twice as good as B.

Okay, I can see why this was unclear. In act B and D, you're doing the same thing—preventing Afiya from getting malaria. In act B, you cause Afiya not to get malaria, and Brian is born and dies of malaria. In act D, Brian does not get malaria. According to the GiveWell view, it doesn't matter which of these things happens when you $3000 donate to AMF—either way it's 35 DALYs "better" than act A. But the GiveWell view also claims that B < D, so it contradicts itself.

Which is it: Brian doesn't get malaria (for reasons that have nothing to do with AMF), or Brian is prevented from getting malaria by AMF? Those are pretty different. The cost-effectiveness analysis is indifferent to the first option, but is affected by the second option.

I also didn't understand this part. Switching from A to D adds 70 years to what would have been lives that would have existed under option A.

It's the former. In the second scenario, the only thing donating to AMF does directly is causes Afiya not to get malaria.

You should edit the post then. Right now it says "you prevent them both from getting malaria", which is the exact opposite of what you said just now. In this case I agree that the GiveWell view is indifferent between B, C, and D. Births that may result from or be prevented by AMF (which I would argue is pretty negligible) simply aren't considered in the cost-effectiveness analysis. There's nothing contradictory about this.

I broadly agree, the moral uncertainty over additional life years has also reduced my estimates in the intervention (not written up yet). You’ve added a few considerations I hadn’t taken into account yet, and those seem to weaken the case for LLIN distributions further.

I have a separate set of doubts about the comparison between YLL and YLD. For most highly malarial countries that I checked, the YLD per affected person per year through malaria are around 0.018 while the YLL per affected (dead) person per year are 1. So it seems that being dead is 50 times worse than suffering from malaria. Some of the elicitation methods of disability weights seem to determine a ranking of disabilities and then scale it to the 0–1 interval or not allow for values > 1 inherently. If the worst disabilities are “really” (according to subjective views we empathize with) far worse than death, some that are about as bad as death will be scaled down to a really small disability weight. There may be other elicitation methods I haven’t looked at though.

One note: DALYs are something negative, so one typically wants more QALYs but wants to avert DALYs.

Another note: The “GiveWell view” is more widely known as “the Asymmetry”—unless there are subtle differences I’m overlooking.

Somewhat tangential, but:

"For most highly malarial countries that I checked, the YLD per affected person per year through malaria are around 0.018 while the YLL per affected (dead) person per year are 1. So it seems that being dead is 50 times worse than suffering from malaria." No I think you misunderstood. The primary reason the YLD per affected person per year is so low is because a case of malaria generally last significantly less than 1 month, so an "affected person with malaria" will actually not have malaria for most of the year.

When I was researching the disability weight of dengue, yellow fever etc., for the Oxitec policy comment, the disability weights tended to be in the .7-.8 region per duration of suffering from those illnesses (ie, worse than clinical depression, which was surprising to me). Because the symptoms of most mosquito-borne illnesses seem to be quite similar, I suspect that malaria will also have a similar disability weight. So dying is probably no more than 2x as bad as suffering from malaria.

Another point to note is that most people who get malaria don't die. Extrapolating from WHO statistics*, there is roughly one death from malaria every 500 cases! Nonetheless, i believe the median GW employee estimates that the morbidity-impact of malaria prevented through LLINs is about 50% that of the mortality-impact. So unless I'm misunderstanding something (maybe AMF distributes to high-mortality regions or demographics), 1 death is roughly as bad as 1000 cases of non-fatal malaria averted, mostly because of large differences in duration.

http://www.who.int/features/factfiles/malaria/en/

Oh, I see. Thanks! Here is a study that seems to use two of the following disability weights for malaria from this one:

  • Infectious disease: acute episode, mild 0.005 (0.002–0.011)
  • Infectious disease: acute episode, moderate 0.053 (0.033–0.081)
  • Infectious disease: acute episode, severe 0.210 (0.139–0.298)
  • Infectious disease: post-acute consequences (fatigue, emotional lability, insomnia) 0.254 (0.170–0.355)

So that is probably the range where the weighed average should end up too.

These were derived using a combination of two elicitation methods that eventually scale the results to the [0, 1] interval: “The surveys used paired comparison questions, in which respondents considered two hypothetical individuals with different, randomly selected health states and indicated which person they regarded as healthier. The web survey added questions about population health equivalence, which compared the overall health benefits of different life-saving or disease-prevention programmes. We analysed paired comparison responses with probit regression analysis on all 220 unique states in the study. We used results from the population health equivalence responses to anchor the results from the paired comparisons on the disability weight scale from 0 (implying no loss of health) to 1 (implying a health loss equivalent to death).”

I’ve corrected my Guesstimate accordingly. A duration of 2–8 weeks puts the disability weights roughly in that area, but I haven’t found any data on the duration distribution of malaria episodes. (The much larger number of nonlethal episodes is something I already took into account.)

“So dying is probably no more than 2x as bad as suffering from malaria”: But per year, right? Death lasts much longer than a year for most people who die of malaria according to the system. Some 78 years using the Japanese life expectancy at birth and just a decade or two less using sub-Saharan African life expectancies. That’s in the context of the first paragraph, though, which you maybe didn’t even mean to comment on.

While I haven't heard the term "asymmetry" before, it looks like this is a more general claim about population ethics. The "GiveWell view" implies asymmetry, but it also implies that extending a life is good in proportion to the (quality-adjusted?) length of life added, and that it's irrelevant what other people you could affect (i.e., in my second thought experiment, B = C). As I discuss in OP, this has additional problems that merely assuming asymmetry does not. Most significantly, to fix certain problems with asymmetry, you probably have to assume B < C, but the GiveWell view assumes B = C.

Thanks for making the distinction clear!

You can choose between three possible acts. When you perform act A, a child (let’s call her Afiya) is born, gets malaria, and dies. Under act B, you cause Afiya not to be born. According to assumption (1), this act is not worse than A. Standard person-affecting view says that it is not wrong to cause someone to exist whose life is net positive, so A is not worse than B. Under act C, you cause Afiya to be born and prevent her from getting malaria. This beats act A according to (2), and is not better than act B according to (1). Thus, A = B, B ≥ C, and C > A. But this creates a contradiction: B > A and B = A.

If the fact that "act [B] is not worse than A" leads to the equation A = B, then why does the fact "[act C] is not better than act B" lead to the equation B ≥ C? It would make more sense if you simply said B = C, as the current equation seems to raise the possibility that B is better than C without offering any justification for that view.

You could claim that causing someone to come into existence and have a happy life but then die prematurely is a bad act. Some people do claim this, but most people don't and I thought it was sufficiently implausible that it was worth rejecting. If you do make this assumption, it raises new concerns.

You wrote that act B (nonexistence) is greater than or equal to act C (full happy life). I understand that they're equal under a standard person affecting view, but I'm asking if there's any view under which act B is greater. If there's no such view, it may make more sense to say B = C instead of B ≥ C, as the latter equation implies that such a view does exist.

Sure you could say that, but it doesn't matter because A = B, B = C, and C > A is still a contradiction.

Okay thanks. Just wanted to clarify.

[Edited]

For example, last year when Stanford Effective Altruism was considering making donations to charity, we preferred the Schistosomiasis Control Initiative over AMF because we believed that GiveWell gave too much significance to the “GiveWell view” of population ethics and not enough to the total view.

I'm confused about how the differences between SCI and AMF connect to population ethics. Neither charity seems like it would have obvious effects on the birth rate. Both schistomiasis and malaria do harm through a mix of killing people and lowering their subsequent quality of life, but I guess it's a different mix and the demographics of the people affected is different? It would help a lot to lay out specifically what those differences are.

Almost all the benefits of SCI come from improving quality of life, which is a good thing under any view of population ethics. The case for AMF being better than SCI requires adopting what I call the GiveWell view of population ethics (unless you believe that the non-life-saving benefits of AMF make it worthwhile, which seems fairly plausible to me).

I think that adopting your first resolution, in addition to the assumption by commenters that being a child with malaria is a net negative experience, can rescue some of the value of AMF. Say in situation 1, a family has a child, Afiya, who eventually gets malaria and dies, and thus has a net negative experience. Because of this, the family decides to have a second child, Brian, who does not get malaria and lives a full and healthy life. In situation 2, where AMF is taken to have a contribution, a family has just one child, Afiya, who is prevented from getting malaria and lives a full and healthy life. The family does not decide to have a second child. Only taking into account the utility of the people directly affected by malaria, and not the family, it seems to me that situation 1 is worse than situation 2 by an amount equivalent to Afiya's net negative experience of getting malaria; the reverse of this could be said to be AMF's contribution. So while this is not the same as 35 QALY's, it still seems like a net positive.

EDIT: Note of clarification: The above is in particular a response to the statement, "Because AMF hardly changes humans’ lifespans, it does not have a clear beneficial effect for humans," which was stated as a problem for Givewell with adopting the first resolution.