Argumentative Debates for Transparent Bias Detection [Technical Report]