How do I use fairness metrics to compare model performance across demographic groups?

Example Concept: Fairness metrics are used to evaluate how a model's predictions differ across demographic groups. For instance, demographic parity checks if the positive prediction rate is equal across groups, while equal opportunity ensures that true positive rates are similar. Disparate impact measures the ratio of outcomes between groups, aiming for a threshold (e.g., 80%) to indicate fairness. These metrics are crucial for identifying bias and guiding model adjustments.