Should effective altruists donate to fighting Ebola?
Argument against: usually very famous things that make the news are terrible effective altruist causes and you should stick to well-studied things like malaria.
Argument for: Ebola is very underfunded compared to sexier disasters. And it is a disease in the Third World, a category which has brought us most of the best-known effective altruism interventions.
Thoughts: The CDC estimates a best-case scenario of 20,000 cases by January and a worst-case scenario of about 1.5 million cases by January. They do not estimate risks past January. There are also black swan risks in which Ebola spreads to the entire Third World (eg India) and kills tens of millions of people there. However, on the margin individual donations are unlikely to shift the virus from one of these scenarios to another, so it’s probably more worth considering how much good the marginal donation does.
Doctors Without Borders is a very well known, GiveWell-approved charity. They are running clinics in the country, but it’s hard to tell how much more clinic they can run per dollar. On the other hand, they are also giving out home infection prevention kits by the tens of thousands. Other charities price these at about ten dollars per kit, although I’ve seen estimates that differ by an order of magnitude. I don’t think anybody knows how effective the kits are going to be, although everyone agrees they are a vastly inferior option to sufficient space in hospitals, which at the moment does not exist.
If we estimate likelihood of 100,000 Liberians (geometric mean of estimates) eventually infected = 2% of the population, then $1000 buys 100 kits buys 2 kits for people likely to be infected..
$1000 for malaria bed nets supposedly gives something like 20 to 100 DALYs, depending on whose estimate you trust.
Ebola death rate is about 50%. Suppose the average infected person has 30 DALYs left to lose. So each case of Ebola costs 15 DALYs directly. But it probably ends up costing more like 30, because I think on average each case infects one other person (I don’t think this is meant to be iterate, or else the estimate quickly goes to infinity). So if every Ebola kit was 100% effective, we would expect distributing the kits to save 60 DALYs.
That means in order for kits to be as good as the bottom range of estimates for bed nets, they would have to be at least 33% effective in preventing Ebola among people who get them, which they probably aren’t.
On the other hand, every number in this estimate is a total wild guess, and I don’t trust that I’m within two orders of magnitude of anything approaching reality. Kits likely cost more when including distribution (I expect charities to underreport costs to make people feel good about giving them), there’s no guarantee that there’s room for more kits, and my rate of how many subsequent cases are caused by each case is from a half-remembered news article. Does anyone have better ideas for how to figure this out?
The absolute numbers are far far below panic-levels but the underreporting and ridiculously-exponential curve is pretty disturbing. It’s showing little sign of saturation in the currently-affected populations (an apparent levelling off of infection rates in Liberia was accompanied by reports of difficulty gathering data), it could spread to other populations, and wherever it goes it brings not just ebola but economic disruption, famine, and disruption of health systems that deal with other, more common chronic diseases like malaria and childbirth complications. As of now the measured doubling time is circa 3.5 weeks (a bit longer than that which the worst-case CDC models used but not by much) with each case infecting about two new ones.
The scary possibility is it getting established in additional poor urban populations. It already might be just starting to set off famines where it already is. All exponentials eventually run into a wall and saturate, but it’s unclear exactly which walls will do the job here, behavioral or medical or geographc, and exactly where they are and at what order of magnitude they lay. The possibility of thicker spread through larger populations dominates any discussion of the potential effects of the situation.
My feeling is we will know with more certainty the approximate order of magnitude of the issue by Christmas. In the mean time I somehow managed to save a bit recently… money sent. I hope that was paranoid of me.
I don’t think this is meant to be iterate, or else the estimate quickly goes to infinity
It is meant to iterate, which is how we’ve gotten from one case to thousands already. Biology is all about exponential growth. If each person infected now infects an average of less than one new person, the outbreak dies out. If each person infects more than one, it grows until it has infected everyone within reach, which probably means a fixed proportion of the populations of the affected countries. Pushing that number below 1 is a really big deal.
$1000 for malaria bed nets supposedly gives something like 20 to 100 DALYs
Givewell’s DALY numbers are based entirely on preventing deaths, and do not account for the long term effects on malaria survivors. In addition to the DALYs of chronic disease, there is also lost productivity, which could have long term consequences for the society. If the conditions are (or will be) malthusian, it is better for people to die of acute starvation than that they be subject to chronic malnutrition; it is better that they die of acute diseases like ebola, that they suffer from chronic diseases like malaria.
If the conditions are (or will be) malthusian, it is better for people to die of acute starvation than that they be subject to chronic malnutrition; it is better that they die of acute diseases like ebola, that they suffer from chronic diseases like malaria.
No. Better medical care encourages women to get less children and therefore reduces population.
Well, if they’re roughly comparable in terms of moment-to-moment intensity of suffering, then obviously we (utilitarian consequentialists of any sane kind) would rather that a smaller number of people experience it for a brief period (and then die) than that a larger group of people experience it for a long time (and then die). It’s not even a Repugnant Conclusion issue, since it’s hard to argue that chronically malnourished lives have positive value on the margin.
That’s an easy argument but I’m not sure it’s a correct one. The answers to the following questions may be different:
Is it a good thing on the whole that A exists?
Now that A already exists, does A want to commit suicide?
for at least the following reasons:
Other people will be negatively affected (it may be worse to get to know someone and have them commit suicide, than never to encounter them at all).
People have an inbuilt preference for staying alive, for obvious biological reasons, so someone may go on wanting to live well beyond the point at which each day of life is substantially negative for them. (Just as someone may go on wanting to take heroin well beyond the point at which they’d be happier if they stopped.)
Even without taking into account cognitive biases and externalities, that argument only applies to people who don’t believe in afterlife: if you thought that by killing yourself you would go to hell and by enduring shit thirty more years you would go to heaven, you’d think twice before picking the former.
if they’re roughly comparable in terms of moment-to-moment intensity of suffering
So, um, you think that the life of the chronically malnourished consists of nothing but suffering comparable to that of someone dying from acute starvation??
It’s not even a Repugnant Conclusion issue, since it’s hard to argue that chronically malnourished lives have positive value on the margin.
It’s not a repugnant conclusion to decide that going around killing the chronically malnourished is a good thing—are you quite sure?
So, um, you think that the life of the chronically malnourished consists of nothing but suffering comparable to that of someone dying from acute starvation??
Not my area of expertise, but for the argument to work they only need to be within maybe an order of magnitude or so. Does that seem plausible?
It’s not a repugnant conclusion to decide that going around killing the chronically malnourished is a good thing—are you quite sure?
It certainly sounds repugnant, but I don’t think it’s equivalent to the Repugnant Conclusion, which is what I was referring to. In any case, you are correct that that would be the logical conclusion to draw from what I said, which is concerning me. Population ethics has always confused me—does anyone have any idea how this one might be resolved?
Well, if they have positive value on the margin, then that means it would be an increase in global utility if a bunch of children were born who were doomed to lives of permanent malnutrition. In fact, it would imply that a morally defensible solution to a Malthusian overpopulation problem would be to let the population explode in third world countries, and then just not let any resources be diverted to them. Heck, a sufficiently underserved African country, for example, could turn into a practically unstopppable utility factory—and quite cheaply!
In all seriousness, I think one of the issues here is that we’re conflating “has positive marginal utility in our utility function” with “considers their own life to be worth living (i.e., doesn’t kill themselves)”, when there doesn’t seem to be any particular reason those should always line up (especially if we’re not preference utilitarians).
Edit: This is a better summary of my line of thinking about this.
when there doesn’t seem to be any particular reason those should always line up
In which case it’s a good, moral, virtuous thing to go kill all those miserable wretches who are so misguided as to consider their own life to be worth living.
But you know what, maybe some good can be extracted out of them. I’ve got a Modest Proposal you might consider....
You may have noticed that I spent the entire first paragraph of my comment making that exact point. Again, I think that gjm summarized my line of thinking about this much better upthread, including laying out the more subtle points that I didn’t make in the parent. I think it should be clear by this point that we’re stuck in a false dilemma, since the two positions we’re considering both lead to highly unpalatable conclusions.
1 - Your malaria/ebola comparison must be off somewhere. Here’s how I see it:
Malaria - $5 to distribute a net, some small chance of preventing a person from getting malaria, some very small chance of fatality from malaria. ~$4k per life saved.
Ebola - $10 to distribute a kit, some small chance of preventing a person from getting ebola, 50% fatality.
So the effectiveness of Ebola kits does not need to be anywhere near as effective at preventing infection in order to be equally effective. Ebola kits need only prevent infection ~0.2% of the time to be effective. Note—my calculation probably has some errors and omisions.
2 - CDC is modelling the no-intervention case as exponential, doubling every few weeks. Reading their paper might inform an intermediate-case growth trajectory
3 - It’s worth considering how many millions are being poured into ebola reduction by national governments and thinking that if these governments are intervening effectively, then our inroads into ebola are likely to be proportionately smaller, according to the smaller size of our monetary contribution.
There are also black swan risks in which Ebola spreads to the entire Third World (eg India) and kills tens of millions of people there.
Not meaning to start fear mongering, but since we’re talking about highly improbable events, I wonder how probable a mutation is that makes it spread much more effectively via aerosol.
A Finnish official covered this question on the news and her answer was that such a mutation has never been observed, and Ebola is already transmitted effectively enough so that there’s no selection pressure for more infectivity.
A Finnish official covered this question on the news and her answer was that such a mutation has never been observed, and Ebola is already transmitted effectively enough so that there’s no selection pressure for more infectivity.
The first part of her answer is true, the second part is nonsense.
I’m not worrying about aerosol. That’s low probability.
I think it’s high probability that ebola will become endemic, at least in Africa. I don’t think it’s crazy to believe that America doesn’t have the institutional conscientiousness to stop it here.
The mildly good news is that I expect a vaccine to be developed.
And just as a side issue, I’m none too pleased to have a disease that’s more likely to hit helpful people.
There was enough sloppiness at that hospital in Texas that I’m worried ebola will spread through incompetence. The knowledge and resources might be theoretically available, but they aren’t being used adequately.
I suppose this is plausible since even the western medicine is ridden with certain very problematic hospital bugs that spread and are created at least in part through incompetence.
I’ve seen some terrible hygiene by even some highly regarded professionals. Disinfection is often seen as a nuisance unless you’re doing a procedure.
I can personally attest to that you really have to remind yourself that it’s important since you have to do it so often.
The fourth point seems ok but the study I linked makes me slightly doubtful about the other three. I didn’t do any searching for other such studies, I simply found that one in the NIH report. Here’s another relevant article.
I’m late to the thread—just got pointed here from Slate Star Codex. I gave to Doctors Without Borders recently. Your analysis is pretty similar to mine except on two points. Douglas_Knight pretty well covered the first, about iteration. Supposing an infection prevention kit successfully averts an infection, we should not only count the expected half-a-life saved right away, but all the expected subsequent generations of infection that the averted case would have led to, until the end of the current epidemic.
Second, instead of assuming that a prevention kit goes to a random person in the entire population, I think it’s more realistic to assume that it goes to the family member of a person who has either the flu, food poisoning, etc, or ebola.
I don’t think its worth dealing with Ebola because its not going to get that far. I think Ebola is overhyped. I think worrying about influenza this year and the upper respiratory infection from South America is far more pressing.
Should effective altruists donate to fighting Ebola?
Argument against: usually very famous things that make the news are terrible effective altruist causes and you should stick to well-studied things like malaria.
Argument for: Ebola is very underfunded compared to sexier disasters. And it is a disease in the Third World, a category which has brought us most of the best-known effective altruism interventions.
Thoughts: The CDC estimates a best-case scenario of 20,000 cases by January and a worst-case scenario of about 1.5 million cases by January. They do not estimate risks past January. There are also black swan risks in which Ebola spreads to the entire Third World (eg India) and kills tens of millions of people there. However, on the margin individual donations are unlikely to shift the virus from one of these scenarios to another, so it’s probably more worth considering how much good the marginal donation does.
Doctors Without Borders is a very well known, GiveWell-approved charity. They are running clinics in the country, but it’s hard to tell how much more clinic they can run per dollar. On the other hand, they are also giving out home infection prevention kits by the tens of thousands. Other charities price these at about ten dollars per kit, although I’ve seen estimates that differ by an order of magnitude. I don’t think anybody knows how effective the kits are going to be, although everyone agrees they are a vastly inferior option to sufficient space in hospitals, which at the moment does not exist.
If we estimate likelihood of 100,000 Liberians (geometric mean of estimates) eventually infected = 2% of the population, then $1000 buys 100 kits buys 2 kits for people likely to be infected..
$1000 for malaria bed nets supposedly gives something like 20 to 100 DALYs, depending on whose estimate you trust.
Ebola death rate is about 50%. Suppose the average infected person has 30 DALYs left to lose. So each case of Ebola costs 15 DALYs directly. But it probably ends up costing more like 30, because I think on average each case infects one other person (I don’t think this is meant to be iterate, or else the estimate quickly goes to infinity). So if every Ebola kit was 100% effective, we would expect distributing the kits to save 60 DALYs.
That means in order for kits to be as good as the bottom range of estimates for bed nets, they would have to be at least 33% effective in preventing Ebola among people who get them, which they probably aren’t.
On the other hand, every number in this estimate is a total wild guess, and I don’t trust that I’m within two orders of magnitude of anything approaching reality. Kits likely cost more when including distribution (I expect charities to underreport costs to make people feel good about giving them), there’s no guarantee that there’s room for more kits, and my rate of how many subsequent cases are caused by each case is from a half-remembered news article. Does anyone have better ideas for how to figure this out?
The absolute numbers are far far below panic-levels but the underreporting and ridiculously-exponential curve is pretty disturbing. It’s showing little sign of saturation in the currently-affected populations (an apparent levelling off of infection rates in Liberia was accompanied by reports of difficulty gathering data), it could spread to other populations, and wherever it goes it brings not just ebola but economic disruption, famine, and disruption of health systems that deal with other, more common chronic diseases like malaria and childbirth complications. As of now the measured doubling time is circa 3.5 weeks (a bit longer than that which the worst-case CDC models used but not by much) with each case infecting about two new ones.
The scary possibility is it getting established in additional poor urban populations. It already might be just starting to set off famines where it already is. All exponentials eventually run into a wall and saturate, but it’s unclear exactly which walls will do the job here, behavioral or medical or geographc, and exactly where they are and at what order of magnitude they lay. The possibility of thicker spread through larger populations dominates any discussion of the potential effects of the situation.
My feeling is we will know with more certainty the approximate order of magnitude of the issue by Christmas. In the mean time I somehow managed to save a bit recently… money sent. I hope that was paranoid of me.
It is meant to iterate, which is how we’ve gotten from one case to thousands already. Biology is all about exponential growth. If each person infected now infects an average of less than one new person, the outbreak dies out. If each person infects more than one, it grows until it has infected everyone within reach, which probably means a fixed proportion of the populations of the affected countries. Pushing that number below 1 is a really big deal.
Givewell’s DALY numbers are based entirely on preventing deaths, and do not account for the long term effects on malaria survivors. In addition to the DALYs of chronic disease, there is also lost productivity, which could have long term consequences for the society. If the conditions are (or will be) malthusian, it is better for people to die of acute starvation than that they be subject to chronic malnutrition; it is better that they die of acute diseases like ebola, that they suffer from chronic diseases like malaria.
No. Better medical care encourages women to get less children and therefore reduces population.
Oh, right. Damn real life, always giving us easy outs from the interesting ethical dilemmas.
Better for whom and by which metric?
Well, if they’re roughly comparable in terms of moment-to-moment intensity of suffering, then obviously we (utilitarian consequentialists of any sane kind) would rather that a smaller number of people experience it for a brief period (and then die) than that a larger group of people experience it for a long time (and then die). It’s not even a Repugnant Conclusion issue, since it’s hard to argue that chronically malnourished lives have positive value on the margin.
It’s rather easy to argue—they don’t kill themselves, so they presumably think their lives are worth living.
That’s an easy argument but I’m not sure it’s a correct one. The answers to the following questions may be different:
Is it a good thing on the whole that A exists?
Now that A already exists, does A want to commit suicide?
for at least the following reasons:
Other people will be negatively affected (it may be worse to get to know someone and have them commit suicide, than never to encounter them at all).
People have an inbuilt preference for staying alive, for obvious biological reasons, so someone may go on wanting to live well beyond the point at which each day of life is substantially negative for them. (Just as someone may go on wanting to take heroin well beyond the point at which they’d be happier if they stopped.)
Even without taking into account cognitive biases and externalities, that argument only applies to people who don’t believe in afterlife: if you thought that by killing yourself you would go to hell and by enduring shit thirty more years you would go to heaven, you’d think twice before picking the former.
So, um, you think that the life of the chronically malnourished consists of nothing but suffering comparable to that of someone dying from acute starvation??
It’s not a repugnant conclusion to decide that going around killing the chronically malnourished is a good thing—are you quite sure?
Not my area of expertise, but for the argument to work they only need to be within maybe an order of magnitude or so. Does that seem plausible?
It certainly sounds repugnant, but I don’t think it’s equivalent to the Repugnant Conclusion, which is what I was referring to. In any case, you are correct that that would be the logical conclusion to draw from what I said, which is concerning me. Population ethics has always confused me—does anyone have any idea how this one might be resolved?
I think the root issue is that you consider chronically malnourished lives to be not worth living. Is that so?
Well, if they have positive value on the margin, then that means it would be an increase in global utility if a bunch of children were born who were doomed to lives of permanent malnutrition. In fact, it would imply that a morally defensible solution to a Malthusian overpopulation problem would be to let the population explode in third world countries, and then just not let any resources be diverted to them. Heck, a sufficiently underserved African country, for example, could turn into a practically unstopppable utility factory—and quite cheaply!
In all seriousness, I think one of the issues here is that we’re conflating “has positive marginal utility in our utility function” with “considers their own life to be worth living (i.e., doesn’t kill themselves)”, when there doesn’t seem to be any particular reason those should always line up (especially if we’re not preference utilitarians).
Edit: This is a better summary of my line of thinking about this.
In which case it’s a good, moral, virtuous thing to go kill all those miserable wretches who are so misguided as to consider their own life to be worth living.
But you know what, maybe some good can be extracted out of them. I’ve got a Modest Proposal you might consider....
You may have noticed that I spent the entire first paragraph of my comment making that exact point. Again, I think that gjm summarized my line of thinking about this much better upthread, including laying out the more subtle points that I didn’t make in the parent. I think it should be clear by this point that we’re stuck in a false dilemma, since the two positions we’re considering both lead to highly unpalatable conclusions.
I have a couple of remarks
1 - Your malaria/ebola comparison must be off somewhere. Here’s how I see it: Malaria - $5 to distribute a net, some small chance of preventing a person from getting malaria, some very small chance of fatality from malaria. ~$4k per life saved. Ebola - $10 to distribute a kit, some small chance of preventing a person from getting ebola, 50% fatality.
So the effectiveness of Ebola kits does not need to be anywhere near as effective at preventing infection in order to be equally effective. Ebola kits need only prevent infection ~0.2% of the time to be effective. Note—my calculation probably has some errors and omisions.
For reference, GiveWell’s estimates, CDC projections
2 - CDC is modelling the no-intervention case as exponential, doubling every few weeks. Reading their paper might inform an intermediate-case growth trajectory
3 - It’s worth considering how many millions are being poured into ebola reduction by national governments and thinking that if these governments are intervening effectively, then our inroads into ebola are likely to be proportionately smaller, according to the smaller size of our monetary contribution.
4 - A recent relevant link, HT Carl
ETA: CellBioGuy gave an excellent response here.
Not meaning to start fear mongering, but since we’re talking about highly improbable events, I wonder how probable a mutation is that makes it spread much more effectively via aerosol.
A Finnish official covered this question on the news and her answer was that such a mutation has never been observed, and Ebola is already transmitted effectively enough so that there’s no selection pressure for more infectivity.
The first part of her answer is true, the second part is nonsense.
The answer was probably incomplete.
Slightly scary, isn’t it, with all the doctorates and stuff.
I’m not worrying about aerosol. That’s low probability.
I think it’s high probability that ebola will become endemic, at least in Africa. I don’t think it’s crazy to believe that America doesn’t have the institutional conscientiousness to stop it here.
The mildly good news is that I expect a vaccine to be developed.
And just as a side issue, I’m none too pleased to have a disease that’s more likely to hit helpful people.
Can you expand on that?
I wonder how this should impact the decision of being helpful under a consequentialist moral system, if at all.
There was enough sloppiness at that hospital in Texas that I’m worried ebola will spread through incompetence. The knowledge and resources might be theoretically available, but they aren’t being used adequately.
I suppose this is plausible since even the western medicine is ridden with certain very problematic hospital bugs that spread and are created at least in part through incompetence.
I’ve seen some terrible hygiene by even some highly regarded professionals. Disinfection is often seen as a nuisance unless you’re doing a procedure.
I can personally attest to that you really have to remind yourself that it’s important since you have to do it so often.
I think it’s more likely for a disease that spreads through aerosol to mutate to become as deadly as Ebola. Like with SARS.
Um, diseases are generally under selection pressure to become less deadly, not more.
That just means that the mutant strain won’t be as virulent as it otherwise would. It won’t keep the mutation from happening.
Yes, and SARS was quite easily contained, a less virulent strain would be even easier.
You’re probably right. I think if we’re interested in the risk of superbugs we should be interested in the aggregate, not individual pathogens.
Seconded.
CellBioGuy gave an excellent response here.
The fourth point seems ok but the study I linked makes me slightly doubtful about the other three. I didn’t do any searching for other such studies, I simply found that one in the NIH report. Here’s another relevant article.
What kind of confidence interval lies between best case and worst case?
A better question is what kind of a distribution do they think they’re looking at—in particular, whether it’s bimodal.
I’m late to the thread—just got pointed here from Slate Star Codex. I gave to Doctors Without Borders recently. Your analysis is pretty similar to mine except on two points. Douglas_Knight pretty well covered the first, about iteration. Supposing an infection prevention kit successfully averts an infection, we should not only count the expected half-a-life saved right away, but all the expected subsequent generations of infection that the averted case would have led to, until the end of the current epidemic.
Second, instead of assuming that a prevention kit goes to a random person in the entire population, I think it’s more realistic to assume that it goes to the family member of a person who has either the flu, food poisoning, etc, or ebola.
I don’t think its worth dealing with Ebola because its not going to get that far. I think Ebola is overhyped. I think worrying about influenza this year and the upper respiratory infection from South America is far more pressing.