Would it be progress if people start clustering effect sizes/ROI on the “right” side of a “too good to be true” threshold? How to balance quest for novelty vs. research that just confirms our priors from a well known distribution of effect sizes/ROI? Glad I’m not an editor.

But I agree that “no free lunch” and “be suspicious of psychology/information type interventions” are good (though hardly definitive) guiding principles.

“Major changes in important things” and “tiny interventions” are doing some heavy lifting there. The search for low cost, high ROI interventions will continue in Dev and Applied Micro regardless of the issues with replicability and/or research fraud in particular studies.