Example Concept: Pre-processing techniques involve modifying the training data to reduce bias before model training, such as re-sampling, re-weighting, or transforming features. In-processing methods adjust the learning algorithm itself to minimize bias, often by incorporating fairness constraints or regularization terms. Post-processing techniques modify the model's predictions to achieve fairness, such as through threshold adjustments or re-ranking. Fairness metrics like demographic parity, equal opportunity, and disparate impact are used to evaluate the model's fairness across different groups.