Luisa Rodriguez is research analyst at 80,000 Hours. Previously, she researched civilisational collapse at the Forethought Foundation for Global Priorities Research, and nuclear war at Rethink Priorities and as a visiting researcher at the Future of Humanity Institute. Before that, she conducted cost-effectiveness evaluations of nonprofit and government programs at ImpactMatters, Innovations for Poverty Action, and GiveWell (as a summer intern).
Luisa_Rodriguez
Karma: 2,075
My experience with imposter syndrome — and how to (partly) overcome it
Long-Term Future Fund: July 2021 grant recommendations
The Forethought Foundation is hiring!
Oh dear — yes it should! Edited, thanks for flagging!
What is the likelihood that civilizational collapse would directly lead to human extinction (within decades)?
Updated estimates of the severity of a nuclear war
Yea, I find this really difficult to think about. I think if I’d never joined Rethink, I’d have ended up continuing to work in the global poverty space (>70%). If I left Rethink now, I’d probably look for (longtermist-oriented) research and research-adjacent jobs at EA orgs and EA-aligned think tanks.
Hey Aaron, good question!
I’m currently in touch with folks at the Nuclear Threat Initiative and a few other similar think tanks, but I don’t think my work has meaningfully influenced their views/activities. My hope is that this will change as I continue building my relationships with them.
To date, I think the audience that has engaged most with (and gotten the most value out of) the nuclear risks series is funders in the EA space. For example, I understand that multiple EA funders/grantmakers have drawn on (and augmented) some of my nuclear risks models as part of some cause prioritization work they’ve done.
Will the Treaty on the Prohibition of Nuclear Weapons affect nuclear deproliferation through legal channels?
Rethink Grants: an evaluation of Donational’s Corporate Ambassador Program
Note: I can’t discuss this, since it’s covered by an NDA, and I haven’t seen the report that OpenPhil received, but compared to what I see as a superforecaster on the questions it looks like the numbers you have from GJP are wrong.
Davidmainheim, thanks for raising this! The GJI data should be correct now — let me know if you notice any other inconsistencies.
Thanks for flagging. Edited!
How many people would be killed as a direct result of a US-Russia nuclear exchange?
Hi Topher,
I really appreciate you engaging so meaningfully with the arguments on countervalue and counterforce targeting. It’s a critical factor in understanding how much harm a nuclear war would cause, so it’s important to get right.
I actually think we may not disagree as much as you seem to think (this makes me think my posts weren’t clear enough on some key points). I want to clarify my position on targeting strategies in the hope that we might tease out exactly how much we disagree:
~
Clarification 1: You note that the US and Russia would likely target each others’ cities (e.g. Moscow) during counterforce targeting. I completely agree with this. I think a key miscommunication here stems from the fact that I’m using the following definitions of countervalue and counterforce targeting:
Countervalue targeting: targeting an enemy’s cities with the sole and explicit aim of killing civilians and disabling industry.
Counterforce targeting: targeting an enemy’s nuclear forces — for example, its missile silos and military bases, including those in and around cities.
So even though I’ve made the case that the US and Russia may be less likely to use countervalue targeting, I certainly grant that US and Russian cities near counterforce targets would likely be caught in the crossfire. I attempted to account for this to some extent in the two posts where I estimate the number of deaths caused directly and indirectly by a US-Russia nuclear war, but 1) this wasn’t clear in my posts, and 2) I didn’t do a good enough job of taking this into account in one of the two posts.
In my post on the number of deaths that would be caused directly by nuclear detonations, I quantified the deaths caused by counterforce targeting by the US and Russia by extrapolating from estimates generated by other people (academic researchers in the case of counterforce targeting against the US, and US government analysts in the case of counterforce targeting against Russia). I wasn’t clear about this in the post, but in both cases, the authors assumed that the counterforce scenarios they analyzed would involve the targeting of counterforce targets in and around major US and Russian cities. For example, the study on the expected fatalities caused by Russian counterforce targeting assumed that Russia would target Sacramento, San Francisco, Chicago, among others, as all of these places have key military bases other and facilities that relate to the US’s nuclear forces. Given this, my estimate of the deaths caused directly by nuclear detonations should do a reasonable job of accounting for the fact that some cities near counterforce targets would be bombed. Clearly, this was not well-explained in my write up. I’ll edit the post to make this clearer. If you have other suggestions for how to better account for this, I’d be happy to hear them.
By contrast, my post on the severity of a nuclear winter does not do a good enough job of accounting for the fact that counterforce targeting would involve bombing counterforce targets in or near cities. While I did take account of the fact that both the US and Russia would be quite likely to target each others’ capitals, I didn’t consider the smoke that would be generated by counterforce targeting in and around other cities that might be within the affected radius of a nuclear detonation during a counterforce strike by either country. This could end up making a difference to my conclusions, so I appreciate you flagging it as a gap in my work. I’ll spend some time in the next few weeks revising my model and post to reflect this (and I’ll update this thread once I’ve made those changes).
~
However, there’s still a big difference in total direct deaths and nuclear winter risk between some metropolitan areas being caught in a counterforce crossfire and major metropolitan areas being intentionally targeted to maximize civilian casualties (as would likely happen in a countervalue war). This brings me to my next clarification:
Clarification 2: You’ve argued that the US and Russia would definitely use both counterforce and countervalue targeting. My view is that the US and Russia might use countervalue targeting but I don’t think it’s certain that they would. In my post, I hypothesize that there’s a 25% chance (90% CI: 5%–59%) that the US would target Russian cities as part of countervalue targeting (as opposed to targeting Russian cities to destroy its nuclear weapons or incapacitate its leadership). Similarly, I think there’s a 40% chance (90% CI: 7%–81%) that Russia would target US cities as part of countervalue targeting.
I suspect that one of the reasons our views differ on this comes from the fact that you’ve cited discussions of targeting strategies from 2009 and earlier. For example, you noted that “US war planning have not changed significantly between 1976 and 2009.” But while that’s true, the US has updated its war planning as of 2010. In 2010, President Obama, updated the US’s nuclear weapons policy. According to the Nuclear Employment Strategy (2013), a document summarizing the policy changes produced by the Department of Defense, the new guidance from President Obama “requires the United States to maintain significant counterforce capabilities against potential adversaries… The new guidance does not rely on a ‘counter-value’ or ‘minimum deterrence’ strategy.” The report later says, “the United States will not intentionally target civilian populations or civilian objects.”
While I don’t believe this means the US definitely wouldn’t ever use countervalue targeting, I think this offers evidence that 1) nuclear planning strategies have been updated since the Cold War, and 2), nuclear war plans have likely shifted away from countervalue targeting, at least somewhat.
The evidence on whether Russia would target US cities just to maximize casualties and hurt industry is weaker (hence the higher probability and wider confidence interval). I agree that Russia’s development of a nuclear weapon designed to produce a bunch of fallout offers some evidence that Russia would consider using nuclear weapons to target US cities for non-counterforce purposes. And you’re absolutely right that I conflated the nuclear target list discussed in the Popular Mechanics article with targets that might already be part of Russia’s nuclear targeting plan. I’m going to spend a bit more time considering whether/how much to update my estimate of the probability that Russia would use countervalue targeting against the US in light of these points. My current feeling is that there are still compelling reasons to think Russia would be more likely not to implement countervalue targeting relative to implementing it — in particular, because it would threaten the destruction of its society, and possibly (independently) put Russia at a strategic disadvantage during the nuclear exchange.
Lastly, I’d encourage us to be probabilistic here. I’d be quite curious to know how high you’d put the probability that the US and Russia would engage in countervalue targeting (as I’ve defined it), and what kind of results you get given those views. I’d also be curious if you have any other feedback on where my Guesstimate model (for example, see https://www.getguesstimate.com/models/13506) may be wrong. Regardless of where I personally come down on the probability of countervalue targeting by the US and Russia, the Guesstimate models I built are publicly available, and I encourage anyone who disagrees with my views on US and Russian targeting strategies to play with the probability of countervalue targeting and see how tweaks change the results.
How likely is a nuclear exchange between the US and Russia?
How bad would nuclear winter caused by a US-Russia nuclear exchange be?
The one significant thing I was confused about was why the upper bound survivability for stationary, land-based ICBMs is only 25%.
Good catch! I wasn’t considering the fact that a countervalue attack might be ‘launched on warning,’ rather than after a first strike had already destroyed a portion of a country’s nuclear arsenal. I’ve updated the Guesstimate model in the third post to reflect that full-scale countervalue targeting could get a bit deadlier than I originally accounted for, and I’ll make sure my post on nuclear winter takes this into account as well. I don’t have the bandwidth to update all of the figures in the post immediately, but I should be able to do so soon.
Also, maybe you intend for your adjustment for US missile defence systems to be negating 15% of the lost warheads rather than adding 15% to the total arsenal?
Right, again! Thanks for flagging this. Here’s an updated version of my calculations. I now find that somewhere between ~990 and 1,500 US nuclear weapons would survive a counterforce first strike by Russia. I’ll update the post to reflect these changes soon.
Also, I see you’ve left some great feedback on posts 2 and 3. I’ll be replying to those comments shortly.
Hi Kit,
Thanks for your comments — I’m glad to hear you’re enjoying the series!
If you have written or do write about how future changes in arsenals may change your conclusions about what scenarios to pay the most attention to, I’d be interested in hearing about it.
I haven’t written about this yet, but I’ll consider working it in as I continue to explore the topic in the next few months. I’ll update this thread if I do.
To what extent do you think the survey you use for the probabilities of particular nuclear scenarios is a reliable source?
I’ll be sharing a post on the probability of a US-Russia nuclear war soon. It talks a little bit about the relative merits and weaknesses of some of these probability estimates.
2 votes
Overall karma indicates overall quality.
Total points: 0
Agreement karma indicates agreement, separate from overall quality.
Do you have a citation for the 100-1000 figure?
Comes from here https://mason.gmu.edu/~rhanson/collapse.pdf and the papers it cites: