I wish the systemic change discussion was less focused on cost-effectiveness and more focused on uncertainty regarding the results of our actions. For example, in 2013 Scott Alexander wrote this post on how military strikes are an extremely cheap way to help foreigners (“at least potentially”). I’m glad he included the disclaimer, because although Scott’s article works off the premise that “life is ~10% better in Libya after Gaddafi was overthrown”, Libya isn’t looking too hot right now—Obama says Libya is the biggest regret of his presidency. Scott also failed to mention that American intervention in Libya may have reduced North Korea’s willingness to negotiate regarding its nuclear weapons program.
To me, uncertainty means it’s valuable to research systemic changes well in advance of trying to make them. If systemic changes aren’t cost-effective now, but might be cost-effective in the future, we should consider starting to theorize, debate, and run increasingly large experiments now anyway. (Disclaimer: Having productive disagreements about systemic changes is in itself a largely unsolved institution design problem, I’d argue! Maybe we should start by trying to solve that.)
Note that Scott himself has said he agrees the Libya intervention turned out poorly and is now no longer comfortable endorsing things similar interventions, see http://web.archive.org/web/20150731041537/https://slatestarscratchpad.tumblr.com/post/125060547081/how-have-your-political-positions-shifted-through