Noisy labels as a Bias model# Fairness constraints for recovering from biased labels# Blum & Stengl Considers 3 cases: more errors in the disadvantaged group than the advantaged group fewer positive examples of the disadvantaged group both
