Techniques applied after model training to reduce bias, such as adjusting decision thresholds per group or recalibrating predictions.