Summary: My uncertainty about this post lies in how much your proposed interventions can contribute to the overall size of the problem, which interventions these are, and how neglected they are.
Comparing the US prison population & NK: While I agree that the overall size of the problem is greater in NK, it does not help us to understand how much of the problem we can address. The US estimate are limited to prison reforms, but the NK estimate includes the losses from the general population’s suffering, which is much broader and complicated.
Estimates of total GDP lost in NK: Comparing the $ lost from SK to NK is useful insofar as you’re trying to estimate the effect of the communist regime on the prospects of NK. However, I don’t think this is maybe the right number when thinking about the near-term gains made from “suceeding”. Things that would affect your estimate:
What time-scale of gains are you looking at? For example, on the timescale of 50 years perhaps it is reasonable that eventually NK will catch up to SK but in the interim the gains would be less.
What solutions are you envisioning and what is the end goal?
It seems that a fair number of your proposals are to reduce the suffering but not try to make bigger changes to the overall regime. In that case, it seems that the overall GDP gains are going to be much smaller, and each solution may have a natural limiting factor.
e.g. Safe escape routes will have a limiting factor at which point authorities may crack down/make it harder to escape, and mass exodus of millions of people doesn’t seem plausible without a breakdown of the entire state.
Some of your proposals could lead to regime change—e.g. the dissolution of the NK government or opening up of NK (e.g. lowering the price of exports), then the more appropriate reference class would be to look at the various Eastern bloc/Soviet countries immediately after their independence. Many countries faced several years of economic recession followed by a slow recovery. Additionally, there is no guarantee, even if things go better than the ex-Soviet countries, what timescale these changes would happen across.
Does this sound right?
Neglectedness: Do you have an estimate range for the amount of funding currently in this space and the number of actors (e.g. organisations?).
Tractability: It might have been useful to get a little better sense of the track record of these interventions to date to help compare them better. E.g. how many NK’s try and how many succeed to take a escape route?
Intervention—Capacity building: I think your point about improving coordination in capacity building is really interesting and I enjoyed reading the Attribution of Moloch post! You mention: “Metainterventions aiming at capacity building are said to be highly robust, and once a more coordinated space emerges, it will be easier for organizations to conduct research on further intervention in the open.”—I’m curious what the source is for this?
I’ve added a note to the post that I accidentally changed the publishing date of it yesterday. I wanted to edit it but made it a draft instead, and when I reversed that, it must’ve bumped the date. All of this information is 3–4 years out of date.
Today I would approach things differently in three main ways:
I would put a strong emphasis on trying to determine the robustness of all candidate interventions, by which I mean how likely I think they are to backfire in terrible ways, how many such ways I can think of, whether they can be preempted, or monitored and responded to, or whether they are more likely to happen suddenly and irreversibly or unknowably.
I would take more seriously the possibility that extreme suffering is worse than median-levels of suffering by factors of 10, 100, or 1000 than by a factor of around 2. That would make the suffering in concentration camps and that of factory-farmed chickens vastly more important than it appears according to the calculations here.
I would ask how this problem interacts with other problems, especially those that I deem even more urgent and important. For example, would a more human rights–respecting North Korea (or a united Korea under a South Korea–like government) improve existential security? Might it decrease the probability of some s-risks? Or would activism in this area, especially EA-adjacent activism, detract from the talent pools, funding, or attention that other problems, such as AI policy, have/receive?
Comparing the US prison population & NK: Yes, that comparison only serves to get an idea of the scale but leaves open how many comparatively difficult subproblems we’re dealing with in each case, what interventions there are to address them severally or all together, and hence how tractable the most tractable ones of these interventions are and at what rate they’re becoming less tractable up to the point where the problem is largely solved.
Estimates of total GDP lost in NK: I think I agree. Only two comments: (1) I think all the interventions I came up with except for the one concerning safer escape routes are either “meta” or aim at destabilizing the regime as a whole. Whatever connects and empowers the North Korean population or weakens the government makes it more likely that the government can be overthrown or at least becomes irrelevant. And (2), maybe the recovery of the former German Democratic Republic could be seen as a best-case scenario because there there was also a more developed part of the same country that it could reunite with and receive support from. But I think the cultural differences between North and South Korea are more extreme at this point than the differences between GDR and FRD.
Neglectedness: Indeed! But I don’t. I think people at Saram or the ICNK may have estimates though. I would imagine that government grants are transparent and that the charities in the space also publish their financials. Maybe the financials of the Ministry of Unification (other than grants) or South Korean intelligence are harder to know.
Tractability: Yes, I think that should be possible to estimate at least for someone who speaks Korean because a lot of North Koreans eventually flee to South Korea where they receive support and so are known to the authorities. The charities that support them to get through and out of China are often US-based, so maybe there’s even English-language information on this.
Intervention—Capacity building: Thank you! The first part is a mix of the idea that capacity building is a convergent instrumental goal in a lot of contexts and a general impression I got from discussions in EA. The second part may not be true. What I think was true at the time was that the space had these Attribution Moloch type of dynamics going on that made organizations more secretive and suspicious of each other. I don’t know if that’s still the case. But even without that problem, they may still not be able to do their research completely in the open because then the North Korean government can react to it more easily. For example, if you send aid via balloons into North Korea, the North Korean government can launch identical balloons of their own, ostensibly coming from the south, but with poisoned products or misinformation.
Thanks again for all your comments! I probably won’t write an updated version of the article, but I hope someone else will want to dig deeper!
Summary: My uncertainty about this post lies in how much your proposed interventions can contribute to the overall size of the problem, which interventions these are, and how neglected they are.
Comparing the US prison population & NK: While I agree that the overall size of the problem is greater in NK, it does not help us to understand how much of the problem we can address. The US estimate are limited to prison reforms, but the NK estimate includes the losses from the general population’s suffering, which is much broader and complicated.
Estimates of total GDP lost in NK: Comparing the $ lost from SK to NK is useful insofar as you’re trying to estimate the effect of the communist regime on the prospects of NK. However, I don’t think this is maybe the right number when thinking about the near-term gains made from “suceeding”. Things that would affect your estimate:
What time-scale of gains are you looking at? For example, on the timescale of 50 years perhaps it is reasonable that eventually NK will catch up to SK but in the interim the gains would be less.
What solutions are you envisioning and what is the end goal?
It seems that a fair number of your proposals are to reduce the suffering but not try to make bigger changes to the overall regime. In that case, it seems that the overall GDP gains are going to be much smaller, and each solution may have a natural limiting factor.
e.g. Safe escape routes will have a limiting factor at which point authorities may crack down/make it harder to escape, and mass exodus of millions of people doesn’t seem plausible without a breakdown of the entire state.
Some of your proposals could lead to regime change—e.g. the dissolution of the NK government or opening up of NK (e.g. lowering the price of exports), then the more appropriate reference class would be to look at the various Eastern bloc/Soviet countries immediately after their independence. Many countries faced several years of economic recession followed by a slow recovery. Additionally, there is no guarantee, even if things go better than the ex-Soviet countries, what timescale these changes would happen across.
Does this sound right?
Neglectedness: Do you have an estimate range for the amount of funding currently in this space and the number of actors (e.g. organisations?).
Tractability: It might have been useful to get a little better sense of the track record of these interventions to date to help compare them better. E.g. how many NK’s try and how many succeed to take a escape route?
Intervention—Capacity building: I think your point about improving coordination in capacity building is really interesting and I enjoyed reading the Attribution of Moloch post! You mention: “Metainterventions aiming at capacity building are said to be highly robust, and once a more coordinated space emerges, it will be easier for organizations to conduct research on further intervention in the open.”—I’m curious what the source is for this?
Please do not leave us wondering. ;)
Sorry! Must have missed that hanging sentence. There is no guarantee of the timescale across when these changes would happen.
Edited :)
Thank you for your comprehensive comment! :-D
I’ve added a note to the post that I accidentally changed the publishing date of it yesterday. I wanted to edit it but made it a draft instead, and when I reversed that, it must’ve bumped the date. All of this information is 3–4 years out of date.
Today I would approach things differently in three main ways:
I would put a strong emphasis on trying to determine the robustness of all candidate interventions, by which I mean how likely I think they are to backfire in terrible ways, how many such ways I can think of, whether they can be preempted, or monitored and responded to, or whether they are more likely to happen suddenly and irreversibly or unknowably.
I would take more seriously the possibility that extreme suffering is worse than median-levels of suffering by factors of 10, 100, or 1000 than by a factor of around 2. That would make the suffering in concentration camps and that of factory-farmed chickens vastly more important than it appears according to the calculations here.
I would ask how this problem interacts with other problems, especially those that I deem even more urgent and important. For example, would a more human rights–respecting North Korea (or a united Korea under a South Korea–like government) improve existential security? Might it decrease the probability of some s-risks? Or would activism in this area, especially EA-adjacent activism, detract from the talent pools, funding, or attention that other problems, such as AI policy, have/receive?
Comparing the US prison population & NK: Yes, that comparison only serves to get an idea of the scale but leaves open how many comparatively difficult subproblems we’re dealing with in each case, what interventions there are to address them severally or all together, and hence how tractable the most tractable ones of these interventions are and at what rate they’re becoming less tractable up to the point where the problem is largely solved.
Estimates of total GDP lost in NK: I think I agree. Only two comments: (1) I think all the interventions I came up with except for the one concerning safer escape routes are either “meta” or aim at destabilizing the regime as a whole. Whatever connects and empowers the North Korean population or weakens the government makes it more likely that the government can be overthrown or at least becomes irrelevant. And (2), maybe the recovery of the former German Democratic Republic could be seen as a best-case scenario because there there was also a more developed part of the same country that it could reunite with and receive support from. But I think the cultural differences between North and South Korea are more extreme at this point than the differences between GDR and FRD.
Neglectedness: Indeed! But I don’t. I think people at Saram or the ICNK may have estimates though. I would imagine that government grants are transparent and that the charities in the space also publish their financials. Maybe the financials of the Ministry of Unification (other than grants) or South Korean intelligence are harder to know.
Tractability: Yes, I think that should be possible to estimate at least for someone who speaks Korean because a lot of North Koreans eventually flee to South Korea where they receive support and so are known to the authorities. The charities that support them to get through and out of China are often US-based, so maybe there’s even English-language information on this.
Intervention—Capacity building: Thank you! The first part is a mix of the idea that capacity building is a convergent instrumental goal in a lot of contexts and a general impression I got from discussions in EA. The second part may not be true. What I think was true at the time was that the space had these Attribution Moloch type of dynamics going on that made organizations more secretive and suspicious of each other. I don’t know if that’s still the case. But even without that problem, they may still not be able to do their research completely in the open because then the North Korean government can react to it more easily. For example, if you send aid via balloons into North Korea, the North Korean government can launch identical balloons of their own, ostensibly coming from the south, but with poisoned products or misinformation.
Thanks again for all your comments! I probably won’t write an updated version of the article, but I hope someone else will want to dig deeper!