A very big problem with this study is that participants that don’t delegate are not asked to do anything regarding cheating, while the ones that do are presented with various choices that influence them. “choosing a setting on a dial that ranged from ‘maximize accuracy’ to ‘maximize profit’”. In this way, it doesn’t “control” for anything.

One of the authors is from the Center for Adaptive Rationality, obviously making a reference to the lesswrong CFAR.