Preliminary Experiments to Examine the Stability of Bias-Aware Techniques