I've found myself referring to the concept of disentanglement research quite a bit lately.
As best as I can tell, it was first mentioned rather casually in this EA post: https://forum.effectivealtruism.org/posts/RCvetzfDnBNFX7pLH/personal-thoughts-on-careers-in-ai-policy-and-strategy#rxAi3ssD8DtSHJtMG
It might be useful to have a focused place to be able to link people to efficiently convey what disentanglement research is.
How would you define it?