A fuzzy-rough uncertainty measure to discover bias encoded explicitly or implicitly in features of structured pattern classification datasets

•We propose a fuzzy-rough set measure to detected implicit and explicit bias.•Our measure does not depend on any prediction model but a distance function.•Explicit bias quantifies the granular changes after removing a protected feature.•Implicit bias quantifies the correlation between protected and unprotected features.•We conclude that the German credit dataset is more biased toward gender than age.