"Algorithmic bias" research is mostly trash. The more rigorous of it just pretends statisticians haven't been working on sample adjustment and balancing for causal inference for literally generations, and the less rigorous simply insists that we should ignore that things are not equally and identically distributed across population subgroups.