Researchers find that debiasing doesn't eliminate racism from hate speech detection models

In a study, researchers find evidence that debiasing techniques don't make toxicity detectors trained on biased data any less discriminatory.