Hide table of contents

Thanks to Jojo Lee, Bruce Tsai, Sam Buckman, and Poutasi Urale for helpful comments.

Hillary Greaves described "moral cluelessness" in her paper Cluelessness. Cluelessness about the long-term consequences of actions even when they have unambiguously positive consequences in the short term could stop a conscientious longtermist from doing the short-term good. Moral cluelessness of this kind might prevent us from, for instance, knowing whether donating mosquito bednets is net positive, in the very long-term, because we don't know its influence on existential risk and the long-term future. In this post, I'll run a back-of-the-envelope calculation to take you through the steps of calculating longtermist value by identifying some of the most important causal pathways from donating a bednet to existential risk.

How did we get to wondering whether donating bednets is net negative? It is true the short-term consequences are unambiguously positive. However, the long-term consequences are complicated to understand. In fact, Hillary Greaves has argued a longtermist committed to doing the most good regardless of whether it is now or in the future could conclude that donating mosquito bednets, which has been held to be one of the most effective ways to do good, actually has unknowable moral value. Many people think this seems absurd when they first hear it. But it does seem to be a situation where, as Greaves puts it,

For some pair of actions of interest A1, A2, 

 - (CC1) We have some reasons to think that the unforeseeable consequences of A1 would systematically tend to be substantially better than those of A2

 - (CC2) We have some reasons to think that the unforeseeable consequences of A2 would systematically tend to be substantially better than those of A1

 - (CC3) It is unclear how to weigh up these reasons against one another.

In that case, donating the bednets saves lives in the short term, but also has many unforeseeable positive effects. Probably there is substantial life-long economic benefit to bednet recipients who miss a case of malaria, and this, in turn, has benefits to their entire community. This may lead to more economic activity, better education, and as a result, someone from that community becomes an x-risk researcher and invents a technology that reduces risk, say, by reducing climate change risk.

On the other hand, there might be unforeseeable negative effects. Lower mortality rates lead to higher populations, and a higher population leads to more consumption on overstretched resources and higher carbon emissions. We don't know, but any additional person could be the straw that breaks the proverbial camel's back, and realizes the odds (estimated by Ord as 1 in 1000) that climate change turns out to be an existential catastrophe.

It's very counterintuitive to evaluate short-term interventions by the their long-term impact

In a podcast with Rob Wiblin, Greaves notes that even as she has advanced the moral cluelessness thesis, she still donates to short-termist causes:

Maybe I should say upfront I’m also one of those people [who are basing their charitable giving now on the measurable near-term effects of their actions]. I do in fact base my own decisions on GiveWell’s recommendations.

I want to be very clear: in writing this essay, I by no means intend to endorse the moral cluelessness argument. Even Greaves says she doesn’t fully follow through on it, in practice, and I’ve already written about my skepticism elsewhere. But having this concern about moral cluelessness, we might still want to work through the implications. Even if, as individuals, we don’t follow the moral cluelessness arguments to their conclusions, such ideas could still guide policy of major donors or institutions, and so it is important to understand how much it is actually true that the moral value of bednets is mostly unknowable. Is it really, in fact, unknowable, or just unknown? 

We can actually reduce uncertainty

Greaves says

charity evaluators clearly need (inter alia) estimates of the consequences of distributing bednets, per extra net distributed (and hence per dollar donated). Equally clearly, however, these charity evaluators, just like everyone else, cannot possibly include estimates of all the consequences of distributing bednets, from now until the end of time. In practice, their calculations are restricted to what are intuitively the ‘direct’ (‘foreseeable’?) consequences of bednet-distribution

But what if we tried to calculate the long-term consequences of distributing bednets? Let’s assume, for a moment, any effects on animal welfare are negligible, and the hinge of history thesis is really correct. It might imply that the overwhelmingly dominant way people living right now can improve the long-term future is to reduce X risk in the upcoming next few hundred years. To dissolve the moral cluelessness of the action by determining the long-term consequences, if the hinge of history thesis is correct, we don't have to quite go to the end of time; we only ("only"!) have to work out their overall impact on existential risk. It is true that will require a lot of guesswork and calculations far into the future. But it also might be a tractable intellectual problem, albeit one that takes some work.

Not everyone agrees that most dominant way we can improve the long-term future is to reduce x-risk in the next few hundred years. Plausibly what we really need is more emphasis on geopolitical stability, well-being enhancing values, and resilient, well-being enhancing governance institutions. If that were the case, I’d expect the case for altruistically donating bednets to help the less well-off is fairly straightforward. For that reason, I’m focusing this essay on the harder case about how donating bednets could influence existential risk.

We'll demonstrate the exercise, not make a judgment call

I am not, in this post, going to tell you the longtermist value of bednet distribution. What I will do is perform a back-of-the-envelope calculation to take you through the steps of calculating longtermist value by identifying some of the most important causal pathways. In this process, you'll see how the long-term value of causal pathways about which most people are, at first glance, completely clueless, can be estimated to some degree of precision.

Most likely, we won't get quite far enough in this post to actually make a clear determination about bednets. But I hope to demonstrate that we can begin to reduce uncertainty, and hence cluelessness, and if we continue this process, we will reduce uncertainty enough to be able to say with at least some non-zero degree of confidence whether the good is likely to outweigh the bad. Then, we can make a longtermist determination.

Plotting the pathways

The first step in this process is to try to understand the possibilities that bednet distribution could bring about. Greaves makes a start in her article on moral cluelessness, but we can continue the exercise.

It turns out that lower childhood mortality, the primary consequence of donating bednets, although partially offset by lower fertility (people have fewer children), leads to higher net reproduction rate, i.e., more children reach adulthood and have children of their own Doepke (2005). However, lower mortality rates typically goes hand in hand with economic development over time, which will eventually lead in turn to lower birthrates following the initial boom:

The process is iterative, and as we consider the consequences of economic development and population growth followed by decline, we can understand the consequences that follow those. For example:

  • Higher population growth leads to
    • a larger talent pool of inventors and entrepreneurial innovators
    • higher carbon emissions
  • Economic growth plausibly leads to
    • a more stable political system, a more efficient logistics system, and more access to natural resources
    • more efficient farming and a consequent increase in food supply
    • greater purchasing power in the hands of developing countries
      • leaving them better equipped to purchase food when they don't produce enough of their own
      • equalising power between them and richer countries, leading to stronger and more multilateral international institutions
    • Again, higher carbon emissions
  • The act of donating bednets by rich countries could lead to more favorable relations between rich countries and developing nations, leading to a more stable international system
  • More inventors and entrepreneurial innovators, alongside access to natural resources such as lithium, could exert downward pressure on carbon emissions
  • Any increase in carbon emissions could lead to more geopolitical instability
  • Net change in carbon emissions has an influence on the total existential risk due to climate change

Here's what all this looks like, by my watch:

So far, all we've done is highlighted Greaves point, which is that there are many, many factors involved in donating bednets, and who knows how it might impact existential risk, and hence the long term future?

Quantifying the risks and opportunities

However, many of the hypothetical effects we've plotted out in the previous section can be quantified, at least in expectation. Once we understand the expected relative size of each effect, we can better understand whether, in expectation, the positive effects on existential risks outweigh the negative effects.

In order to make the problem more tractable, I have begun with estimating the long-term effects of donating nets in one particular country, the Democratic Republic of the Congo. The DRC is an interesting choice for a few reasons. It has one of the lowest per capita GDPs in the world. At 90 million people, it is a fairly large country, larger than any US state or any European country, including the UK, Germany, or Türkiye. Finally, it is a major recipient of the AMF's bednet distribution program. Additionally, because Greaves' original argument focuses on climate change impact, I'll also consider only climate change existential risk, alongside nuclear war existential risk because it seems closely related. A full analysis would have to consider other environmental risks, possibly pandemic risks, and other risks that could be relevant to economic development and population growth.

You can view my original calculations here. In this section, I'll describe the calculations made in that spreadsheet and how we can begin to quantify the existential risk as a result of donating mosquito nets in the DRC.

We can start with GiveWell's assessment of AMF's program in the DRC. For a given year, they estimate 65 lives saved, alongside 3000 units of value in development benefits, which I think works out to about $27,000 in development benefits per life saved. These amounts are almost certainly negligible in their impact on the long-term future. But because we are interested in even tiny influences, because of the incredibly large value of the long-term future, we still want to know what they are. To make the calculations more vivid, let's imagine we could scale up our bednet distribution until it saved the lives of 2% of the population. Doepke (2005) tells us that while lower child mortality yields somewhat greater population increase overall, it is less than the amount of lives saved because people have fewer children as a result. Let’s split the difference of the 0-2% range and surmise the effect translates into a 1% increase in population size. Considering the development benefits as spread over a 20-year time span, and the DRC's current GDP per capita, the development benefits would be worth about a one-time but permanent 5.94% increase in the DRC's GDP per capita.

We've already highlighted that we want to understand the increase in climate existential risk due to the increase in emissions we'd expect from the consequent population and GDP increase. So what is the impact of these factors on economic growth? Amazingly, Acheampong (2018) informs us that although carbon emissions drive economic growth at a rate of a 2.8% increase in economic growth for every 1% increase in carbon emissions, the reverse is apparently not true: economic growth by other means, in Sub-Saharan Africa, does not generate carbon emissions on net. 

However, we might still want to assume population growth generates emissions, at a rate proportional to the DRC's emissions yield. Assuming the DRC's economy grows and its per capita emissions grow along with it (again, with the causation not being driven by economic growth, per se), by 2050, the DRC will have a per capita GDP of $1700. Countries currently at that level, Liberia, Madagascar, & Chad, currently emit around 0.13 tonnes of CO2 per capita. At that level, a 1% increase in the population of the DRC yields 0.12 million tonnes of CO2 in aggregate. This is the expected increase in CO2 emissions due to the hypothetical bednet intervention.

We can calculate the impact of additional emissions as a fraction of the estimated worldwide impact of all emissions. World anthropogenic carbon emissions is about 35 billion tonnes of CO2 per year. The 0.12 million tonnes of increase as a result of our bednet intervention equates to a whopping 0.00033% of global emissions. In fact, because we're not measuring cumulative emissions going back to the 20th century, an increase in the mid-21st century will be smaller than that. But let's use 0.00033% as a starting point.

What is the likely impact? Ord (2020) gives the risk of climate change and nuclear war at each of 1 in 1000. If the additional risk yielded the population increase of our intervention is proportional to this, it's something like 0.00000067%, or 1 in 150,000,000. 

If this seems like an astronomically small number, well, humanity's future is astronomically vast, so a consistent longtermist might still be concerned with a risk increase of this infinitesimally small magnitude. However, before we conclude our intervention is net-negative, we have to calculate all the other indirect ways that existential risk could be affected through our intervention.

Of all the factors previously listed, possibly the largest could be DRC's potential as a future major lithium and cobalt source. In fact, the DRC’s reserves are so large it is possible they could make the DRC the world's largest supplier of lithium. Australia, currently the world’s top supplier, produced 40,000 tonnes of lithium in 2020, so let's conservatively imagine, if DRC made it to become the world's largest supplier, that it reached this amount. That's enough lithium to build batteries for 4 million electric vehicles. A typical combustion engine vehicle consumes around 10,000 L of fuel in its lifetime, producing around 2.3 kg of CO2 per L fuel, equating to about 22 tonnes of CO2 over the life of the vehicle. If half of those 4 million electric vehicles counterfactually replaced a combustion engine, that would equate to 45 million tonnes Co2 emissions per year. Over 20 years of lithium mining, we might expect lithium for 80 million electric vehicle mined, equivalent to 900 million tonnes of CO2 emissions averted.

So, how would our intervention yield this huge lithium supply? The mining industry is particularly concerned about economic, social and governance risks in the DRC, so economic development that brings social stability could substantially increase the odds of lithium yield. Considering the relationship between hunger and GDP per capita across countries, a 5.94% increase in GDP per capita in the DRC might yield a 2.25% decrease in hunger. Assuming commensurate economic, social, and governance benefits, and assuming these translate to an increase in lithium producing capacity, we might expect 2.2% of 900 million tonnes of CO2 averted. Discount that by another 50% to express uncertainty about whether we’ll really need that much lithium (perhaps hydrogen cars will dominate instead?) and we get to around 10 million tonnes.

That 10 million tonnes CO2 averted must be set against the extra CO2 arising from population growth we already mentioned above, of 0.1 million tonnes.

Internal stability in some countries seems likely to lead to more geopolitical stability. In the DRC, internal conflict has led to around 1 million refugees displaced from the DRC and hosted in other countries. Primarily, these are other African countries, but around 6000 a year request asylum in the European Union. Congolese refugees made up 0.4% of Europe’s asylum seekers in 2016, when anti-migrant sentiment in the UK led to “Brexit”, the UK’s departure from the EU. It is difficult to quantify the benefit of more stability in the DRC for global geopolitical stability, but it seems like it should have some positive value. In this analysis, I’ll avoid assigning any positive value because it seems unclear to me how to quantify it, but further work could reveal a positive appropriate expected value.

There are other factors we've described. But risks arising from arable land declines, pressure on the global food system, and consequent geopolitical instability are already implicitly built into Toby Ord's estimate of existential risk due to climate change and nuclear war that we used earlier. A further increase in scientific innovation and entrepreneurship may also help, but the effect seems marginal considering the low base of $1700 per capita we are starting from.

So, all up, a net 10 million tonnes of CO2 emissions averted, against an insignificant amount increased due to population increase, using the same method of CO2 emissions to x-risk conversion as above, yields our hypothetical intervention saving 1% of the population a existential risk reduction of 1 in 3,400,000. The effect is small, but as I described before, longtermists who buy the hinge of history hypothesis might want to take any effect on x-risk quite seriously.

So why not just focus on the lithium mines?

Of course, if you're a strong longtermist and you bought this argument, you might go as far as the lithium mines and say, "well, new cause area: lithium mines. Can we forget the bednets and focus on that?" The argument here might not convince such a long-termist to donate to bednets. To a donor who would like to save lives in the present without worsening the long-term future, however, we may just have reduced moral cluelessness enough for them to feel comfortable donating bednets.

Is that decided then? Caveats and limitations

Not at all! I’ve spent 10-15 hours on this work, and I don’t think I’ve nearly got to the point where I could stand behind this analysis enough to drive real dollars with it.

There are a few things we'd want to do to really build confidence in the analysis. We need to understand some rough credible intervals around each point estimate, and apply a sensitivity analysis by testing how much our conclusion would change if estimates were at the high or low end of each interval. We’d want to try and get more precision around the parts of the estimate that make the most difference, which might include our expectation of lithium production in the DRC, how much improving the DRC’s economy and health could facilitate lithium production, and how much all that would matter for climate change. We’d also want to explore factors we might have missed. Chief among them might be other existential risks (Toby Ord puts the odds of an environmental existential risk other than climate change at 1/30, much larger than climate change itself), other ways that economic growth could affect global politics (which is a somewhat chaotic system), or any other completely unexplored ways that economic growth in the DRC could affect climate change.

What about donating bednets in countries without huge lithium deposits?

That's another analysis for another day. It is worth noting that many of those countries have substantially greater baseline GDP per capita. We might be more optimistic about improving innovation and technology through further improving their economic growth and stability. We might also think, particularly if inhabitants are well-equipped enough to migrate, legally or not, to other countries, that more internal stability in those countries could lead to more global stability and thus a safer long-term future.

Is this all really necessary?

I don’t have a great read on how seriously anyone takes moral cluelessness about short-term interventions like donating mosquito nets. Even Hillary Greaves said it hasn’t stopped her donating to short-term causes. But EA institutions concerned about the long-term future might avoid short-term interventions if they are worried about their impact on the long term. People outside the EA community might object to EA interventions on moral cluelessness grounds; as Greaves wrote

Many who would otherwise be drawn to Effective Altruism nonetheless refrain from donating any significant portion of their earnings, not because of any positive belief that refraining from donating will have better consequences, but from a sense that they would require more confidence that their donations really would be doing some significant amount of good – less cluelessness – before they are willing to take the bold-feeling step of donating a significant proportion of their income. And, among those who do donate, many donate significantly less than they would if they had no such cluelessness-based worries

In particular, outside the EA community, worries about overpopulation driving climate change are extremely common, and probably motivate people to avoid donating to causes that could result in a larger population, perhaps even if it saves lives. Highlighting how those causes actually would not, on net, worsen climate change, could be convincing outside the community as well as within it.


Longtermism seems like a strange lens from which to view short-term effective interventions. However, the moral cluelessness problem leaves us with a challenge: if we are committed to improving the longtermist future, can we really be morally motivated to perform presentist actions which could be much more likely to worsen the long-term future as improve it? I've elsewhere argued that even with a cursory glance, we might still be able to make a judgment call. Even if that's false, though, even if we need to carefully evaluate possible influences on existential risk in order to evaluate the goodness of a presentist action, we might have a solution. By examining the specific circumstances associated with a proposed presentist action, even if we rely on ballpark estimates and expected values, we can still identify a quantified likelihood which tells us whether the proposed presentist intervention is likely to improve the long-term future, in balance.









More posts like this

Sorted by Click to highlight new comments since:

To a donor who would like to save lives in the present without worsening the long-term future, however, we may just have reduced moral cluelessness enough for them to feel comfortable donating bednets.

I have to admit I find this slightly bizarre. Such a person would accept that we can improve/worsen the far future in expectation and that the future has moral value. At the same time, such a person wouldn't actually care about improving the far future, they would simply not want to worsen it. I struggle to understand the logic of such a view.

They might not be willing to commit 100% to EV maximization no matter how low the probability of making a difference, but entertain EV maximization as one of multiple views over which they have decision-theoretic (normative) uncertainty. Then they want to ensure their actions look good across views they find plausible. That being said, I think it's the entire portfolio that matters and you would want to be robustly positive over the combined portfolio, not on each individual act in it.

Also, they might think no far future-targeted option looks robustly positive in expectation.

Unsurprisingly I disagree with many of the estimates, but I very much like this approach. For any analysis of any action, one can divide the premises arbitrarily many times. You stop when you're comfortable that the granularity of the priors you're forming is high enough to outweigh the opportunity cost of further research, which is how any of us can literally take any action.

In the case of 'cluelessness', it honestly seems better framed as 'laziness' to me. There's no principled reason why we can't throw a bunch of resources at refining and parameterising cost-effectiveness analyses like these, but Givewell afaict don't do it because they like to deal in relatively granular priors and longtermist organisations don't do it afaict because post-'Beware Suprising and Suspicious Convergences' no-one takes the idea seriously that global poverty research could be a good use of longtermist resources. I think that's a shame, both because it doesn't seem either surprising or suspicious to me that high granularity interventions could be more effective long-term than low-granularity ones (eg 'more AI safety research') - IMO the planning fallacy gets much worse over longer periods - and because this...

Plausibly what we really need is more emphasis on geopolitical stability, well-being enhancing values, and resilient, well-being enhancing governance institutions. If that were the case, I’d expect the case for altruistically donating bednets to help the less well-off is fairly straightforward.

... seems to me like it should be a much larger part of the conversation. The only case I've seen for disregarding it amounts to hard cluelessness - we 'know' extinction reduces value by a vast amount (assuming we think the future is +EV) - whereas trajectory change is difficult to map out. But as above, that seems like lazy reasoning that we could radically improve if we put some resources into it.

Thanks for writing this! I've seen Hilary Greaves' video on longtermism and cluelessness in a couple university group versions of the Intro EA Program (as part of the week on critiques and debates), so it's probably been influencing some people's views. I think this post is a valuable demonstration that we don't need to be completely clueless about the long-term impact of presentist interventions.

I appreciate this attempt - I do think trying to understand the impact of reduced mortality on population sizes is pretty key (considering this paper and this paper together implies that population size could be quite crucial for a longtermist perspective). I'm not quite sure you've given this specific point enough attention though. You seem to acknowledge that whilst population should increase in the short-term that it could cause a population decline in several generations - but you don't really discuss how to weight these two points against each other, unless I missed it?

You're right that I didn't discuss it much. Perhaps I should have.

I have a head model that world per capita net GHG emissions will begin to decline at some point before 2050, and reach net zero some time between 2050 and 2100. The main relevance for population here was that higher population would increase emissions. But once the world reaches net zero per capita emissions, additional people might not produce more emissions.

I think it's quite plausible that population decline due to economic growth induced in 2022 won't show up for a couple of generations--potentially after we reach net zero. So I didn't include it in the model. If I had done, we'd get a result more in favour of donating bednets.

Ok, although it’s probably worth noting that climate change is generally not considered to be an existential risk so I’m not sure considerations of emissions/net zero are all that relevant here. I think population change is more relevant in terms of impacts on economic growth / tech stagnation which in turn should have an impact on existential risk.

Ord (2020) listed climate change as an x-risk. Though, on reflection, he may have said that 1/1000 was an absolute upper bound and he thought the actual risk was lower than that.

I have a hard time understanding stories not mediated through climate change or resource shortage (which seems closely linked to climate change, in that many resource limits boil down to carbon emissions) about how population growth in Africa could lead to higher existential risk--particularly in a context where global population seems like it will hit a peak and then decline sometime in the second half of the 21st century. Most of the pathways I can imagine would point to lower existential risk. If the starting point is that bednet distribution leads to lower existential risk, there isn't really a dilemma, and so that case seemed less interesting to analyse. So that's probably one reason I saw more value in starting my analysis with the climate change angle.

However, there are probably causal possibilities I've missed. I'd be interested to hear what you think they might be. I do think someone should try to examine those more closely in order to try and put reasonable probabilistic bounds around them.

I certainly don't think the analysis above is complete. As I said in the post, the intent was to demonstrate how we could "dissolve" or reduce some moral cluelessness to ordinary probabilistic uncertainty using careful reason and evidence to evaluate possible causal pathways. I think the analysis above is a start and a demonstration that we can reduce uncertainty through reasoned analysis of evidence. But we'd definitely need a more extended analysis to act. Then, we can take an expected value approach to work out the likely benefit of our actions.

Curated and popular this week
Relevant opportunities