I’ve just read the shorter version of your report, and gave some specific comments there. Here are some high-level thoughts:
First of all, great work! I think I had moderately high expectations for this report when you first sent me the proposal to give feedback, and I think this has mostly exceeded them, so congrats!
You might want to make the shorter version of the report a separate EA Forum post. The shorter version is ~4500 words and the current post is >20000 words. My guess is that many people will feel intimidated about commenting on this post before reading the whole post (and perhaps after, if they’re worried they missed an important part), particularly on mobile.
At the high level, how has your opinions on political inclusion/exclusion changed as a result of doing this research? In particular, do you think as a result of looking at the case studies, the primary effect was that your existing opinions have gotten more nuanced, or have you dramatically changed your mind from doing this work?
Any high-level takeaways/hot tips on doing EA-focused historical research, or hot takes on the quality of current EA epistemics?
What do other case studies suggest about the hypotheses in this report, and what other hypotheses do they suggest?
Re the above, do you have thoughts on how much predictive (postdictive?) power your framework has on other randomly generated historical case studies?
Relatedly, do you think it’s likely that you will change your mind a lot if you read five more analogous case studies in a similar level of detail? What probability will you assign to reversing one of the core conclusions were you to do so?
Any [. . .] hot takes on the quality of current EA epistemics?
It seems that many EAs have adopted Singer’s expanding circle narrative for thinking about important questions, without much scrutiny, and despite how Singer’s historical narrative is arguably highly incomplete (Singer’s relevant book also wasn’t trying to make a thorough historical argument). This suggests that many EAs aren’t giving enough scrutiny to other arguments from high-profile EAs, and pay too much attention to academic work that happens to come from EAs (even when it’s about questions like “why have many societies become more inclusive?”—questions that aren’t just of interest to EA-sympathetic researchers).
do you have thoughts on how much predictive (postdictive?) power your framework has on other randomly generated case studies?
Relatedly, do you think it’s likely that you will change your mind a lot if you read five more analogous case studies in a similar level of detail? What probability will you assign to reversing one of the core conclusions were you to do so?
Thoughts:
This framework seems (retroactively) predictively powerful for abolition and democratization in many countries. From a distance, it seems roughly predictive of other cases (e.g. factory farming, genocide), although there’s some cases that it seems to get wrong (e.g. it’s not clear to me what the economic incentives for decolonization were). It also seems less predictively useful when incentives seem balanced enough that predictions are ambiguous.
I’d be surprised but not shocked if I changed my mind about any given core conclusion. Maybe 30%? (Overall probability of reversing one core conclusion would depend on how narrowly we’re thinking of “core conclusion.”)
The main way that it seems like I could be wrong would be something like “under the right circumstances, social values are more influential than strong economic incentives.”
I’d be shocked if it turned out that social values are of dominant importance, and economic motives don’t matter much, for bringing about political inclusion/exclusion. That would require explaining away lots of historical evidence. 8%?
My median expectation is that I’d roughly keep the core conclusions and framework, add additional factors that contribute to one outcome or the other (additional ways in which political actors can be economically incentivized to support inclusion/exclusion), and change lots of finer details.
Hi Linch, thanks so much! I’ll reply to your first several bullet points here.
Good point about making the shorter version a separate post. I might do that.
At the high level, how has your opinions on political inclusion/exclusion changed as a result of doing this research? […] Any high-level takeaways [. . .] ?
I don’t think I had very clear/precise opinions on political inclusion/exclusion before this research. But here’s some high-level takeaways/ways in which I changed my mind:
Theory of change/heuristics about what kinds of things drive political progress:
I’m no longer fairly optimistic about value changes on their own when there are other big incentives at play, and I’m now fairly optimistic about value changes on their own when there aren’t other big incentives at play.
I’m now optimistic about looking for clever political strategies, e.g. a policy you can advocate that divides the opposition, or a policy that would spread internationally through a positive feedback loop. (Before, I hadn’t considered this option much.)
Methodology:
My original plan had been to try to predict future moral circle expansion (MCE) by graphing historical trends in MCE, and naively extrapolating them. I’m glad I ended up looking for causal explanations instead, since these helped me figure out when it would be useful, and when it would be misleading, to extrapolate past trends in MCE.
Before looking at these case studies, I spent a lot (~40%?) of my research time reading up on various more theoretical fields that seemed relevant (e.g. psych, IR). They ended up being a lot less helpful than I had expected . If I were to do a similar research project, I’d first look into case studies, and then decide which other sub-fields (if any) would be useful (since then, I’d have a better sense of what info and ideas would be helpful).
I found mentorship (which took the form of weekly memos for and chats with an academic, as well as initially creating a list of readings for each week) really helpful for time management, research design, and exposure to a different perspective.
Over the course of this research, I drifted somewhat from my original research goals, maybe due to a mix of forgetting them, locally optimizing, and letting myself be too influenced by my mentor/mistaking my research proposal for my goals. This seems to have worked out fine, but in the future I’d write out my goals, and regularly (each week?) adjust what I’m doing to better meet them.
My research reinforced my thinking that, for learning about general trends and why things happened, reading from political scientists and economists is often more useful than reading from historians.
I was surprised by the predictive power (especially in Economic Origins of Dictatorship and Democracy) of assuming that organized interests mostly act rationally, with the goal of advancing their own economic interests. This change of mind made me take on board the assumption as a core assumption of my model.
Looking at how similar things happened in many different countries seems to have been helpful for having a better-informed idea of what trends are general trends.
I’ve just read the shorter version of your report, and gave some specific comments there. Here are some high-level thoughts:
First of all, great work! I think I had moderately high expectations for this report when you first sent me the proposal to give feedback, and I think this has mostly exceeded them, so congrats!
You might want to make the shorter version of the report a separate EA Forum post. The shorter version is ~4500 words and the current post is >20000 words. My guess is that many people will feel intimidated about commenting on this post before reading the whole post (and perhaps after, if they’re worried they missed an important part), particularly on mobile.
At the high level, how has your opinions on political inclusion/exclusion changed as a result of doing this research? In particular, do you think as a result of looking at the case studies, the primary effect was that your existing opinions have gotten more nuanced, or have you dramatically changed your mind from doing this work?
Any high-level takeaways/hot tips on doing EA-focused historical research, or hot takes on the quality of current EA epistemics?
Re the above, do you have thoughts on how much predictive (postdictive?) power your framework has on other randomly generated historical case studies?
Relatedly, do you think it’s likely that you will change your mind a lot if you read five more analogous case studies in a similar level of detail? What probability will you assign to reversing one of the core conclusions were you to do so?
(Following up on my other reply)
It seems that many EAs have adopted Singer’s expanding circle narrative for thinking about important questions, without much scrutiny, and despite how Singer’s historical narrative is arguably highly incomplete (Singer’s relevant book also wasn’t trying to make a thorough historical argument). This suggests that many EAs aren’t giving enough scrutiny to other arguments from high-profile EAs, and pay too much attention to academic work that happens to come from EAs (even when it’s about questions like “why have many societies become more inclusive?”—questions that aren’t just of interest to EA-sympathetic researchers).
Thoughts:
This framework seems (retroactively) predictively powerful for abolition and democratization in many countries. From a distance, it seems roughly predictive of other cases (e.g. factory farming, genocide), although there’s some cases that it seems to get wrong (e.g. it’s not clear to me what the economic incentives for decolonization were). It also seems less predictively useful when incentives seem balanced enough that predictions are ambiguous.
I’d be surprised but not shocked if I changed my mind about any given core conclusion. Maybe 30%? (Overall probability of reversing one core conclusion would depend on how narrowly we’re thinking of “core conclusion.”)
The main way that it seems like I could be wrong would be something like “under the right circumstances, social values are more influential than strong economic incentives.”
I’d be shocked if it turned out that social values are of dominant importance, and economic motives don’t matter much, for bringing about political inclusion/exclusion. That would require explaining away lots of historical evidence. 8%?
My median expectation is that I’d roughly keep the core conclusions and framework, add additional factors that contribute to one outcome or the other (additional ways in which political actors can be economically incentivized to support inclusion/exclusion), and change lots of finer details.
Oh man your comments are so insightful! Strongly upvoted.
Will carefully consider if I have follow-up questions later.
Hi Linch, thanks so much! I’ll reply to your first several bullet points here.
Good point about making the shorter version a separate post. I might do that.
I don’t think I had very clear/precise opinions on political inclusion/exclusion before this research. But here’s some high-level takeaways/ways in which I changed my mind:
Theory of change/heuristics about what kinds of things drive political progress:
I’m no longer fairly optimistic about value changes on their own when there are other big incentives at play, and I’m now fairly optimistic about value changes on their own when there aren’t other big incentives at play.
I’m now optimistic about looking for clever political strategies, e.g. a policy you can advocate that divides the opposition, or a policy that would spread internationally through a positive feedback loop. (Before, I hadn’t considered this option much.)
Methodology:
My original plan had been to try to predict future moral circle expansion (MCE) by graphing historical trends in MCE, and naively extrapolating them. I’m glad I ended up looking for causal explanations instead, since these helped me figure out when it would be useful, and when it would be misleading, to extrapolate past trends in MCE.
Before looking at these case studies, I spent a lot (~40%?) of my research time reading up on various more theoretical fields that seemed relevant (e.g. psych, IR). They ended up being a lot less helpful than I had expected . If I were to do a similar research project, I’d first look into case studies, and then decide which other sub-fields (if any) would be useful (since then, I’d have a better sense of what info and ideas would be helpful).
I found mentorship (which took the form of weekly memos for and chats with an academic, as well as initially creating a list of readings for each week) really helpful for time management, research design, and exposure to a different perspective.
Over the course of this research, I drifted somewhat from my original research goals, maybe due to a mix of forgetting them, locally optimizing, and letting myself be too influenced by my mentor/mistaking my research proposal for my goals. This seems to have worked out fine, but in the future I’d write out my goals, and regularly (each week?) adjust what I’m doing to better meet them.
My research reinforced my thinking that, for learning about general trends and why things happened, reading from political scientists and economists is often more useful than reading from historians.
I was surprised by the predictive power (especially in Economic Origins of Dictatorship and Democracy) of assuming that organized interests mostly act rationally, with the goal of advancing their own economic interests. This change of mind made me take on board the assumption as a core assumption of my model.
Looking at how similar things happened in many different countries seems to have been helpful for having a better-informed idea of what trends are general trends.