It’s more of a idk-what-to-do-with-this-effect study

No info on inter-coder reliability either. These are basic robustness checks for such RCTs.

Speaking of AI agents mimicking human behaviour, I wrote this post on how models can reproduce human-like patterns on the surface, yet still miss the behavioural noise buraksonmez.substack.com/p/silicon-sa...

Methods training in sociology varies wildly unlike in econ. Tbh this isn’t even about experimental design, it’s intro-level potential outcomes framework stuff…

Often hard to justify whether this is the main reason, but reviewers frame this as a “major limitation” in the paper. Imo editors in these outlets care more about unanimous decisions given the submission volumes

I wish it was just one design or occasion, and these are the discipline’s top outlets.