I don’t think the issue is that we don’t have any people willing to be radicals and lose credibility. I think the issue is that radicals on a certain issue tend to also mar the reputations of their more level-headed counterparts. Weak men are superweapons, and groups like PETA and Greenpeace and Westboro Baptist Church seem to have attached lasting stigma to their causes because people’s pattern-matching minds associate their entire movement with the worst example.
Since, as you point out, researchers specifically grow resentful, it seems really important to make sure radicals don’t tip the balance backward just as the field of AI safety is starting to grow more respectable in the minds of policymakers and researchers.
I really appreciate you looking into this topic. I think you want to have much much bigger error bars on these, however. Interventions like this are known to have massive selection effects and difficulty with determining causality—giving point estimates is kind of sweeping under the rug the main thing that I’m interested in regarding whether these interventions work.
For example, ACE had a problem similar to this when it was beginning. For one of the charities, they relied on survey data to look for an effect and gave estimates of how effective interventions were based on this, but all of the interesting question was basically “whether we should believe at all the type of conclusion they drew from the surveys”. In the end, of course the answer was no.
I didn’t read the whole post but the reasoning in the summary and early sections seemed to be centered around point estimates and taking-data-at-face-value. The type of analysis that would convince me to change my actions here would be reliability analysis, seeking to show any place within this domain that has extremely clear support for a real effect. By default this basically doesn’t exist for social interventions ime, so the conclusions are unfortunately more affected by the vagaries of the input data rather than the underlying reality.