It's already known that debiasing efforts can backfire in specific ways - overcorrection, for instance, or the ironic process theory of thought suppression. But I believe there's value in pulling these together under one lens: the idea that debiasing, under certain conditions, can give rise to second-order bias. I call this integrated perspective "Debiasing Debiasing." * In AI/ML, similar concerns and expressions do exist around further tuning or redesigning debiasing methods. What I mean by "Debiasing Debiasing" here is something broader: a perspective focused on the fact that debiasing attempts - whether by humans, AI, or other agents - can themselves introduce new distortions, and that those second-order biases need to be examined and corrected in turn.
Hajime Tsui (Sun,) studied this question.