Classification Fairness ======================= .. automodule:: responsibly.fairness The `demos section `_ contains two examples of measuring the fairness of a classifier and applying intervention to adjust it: 1. `COMPAS by ProPublica `_ 2. `FICO credit score `_ Metrics ------- .. automodule:: responsibly.fairness.metrics Independence ^^^^^^^^^^^^ .. autofunction:: responsibly.fairness.metrics.independence_binary .. autofunction:: responsibly.fairness.metrics.separation_binary Separation ^^^^^^^^^^ .. autofunction:: responsibly.fairness.metrics.separation_binary .. autofunction:: responsibly.fairness.metrics.separation_score ROC ~~~ The separation criterion has strong relation to the ROC, therefore these functions can generate ROC and ROC-AUC per sensitive attribute values: .. autofunction:: responsibly.fairness.metrics.roc_auc_score_by_attr .. autofunction:: responsibly.fairness.metrics.roc_curve_by_attr Plotting ~~~~~~~~ .. autofunction:: responsibly.fairness.metrics.plot_roc_by_attr .. autofunction:: responsibly.fairness.metrics.plot_roc_curves Sufficiency ^^^^^^^^^^^ .. autofunction:: responsibly.fairness.metrics.sufficiency_binary .. autofunction:: responsibly.fairness.metrics.sufficiency_score Report ^^^^^^ .. autofunction:: responsibly.fairness.metrics.report_binary A Dictionary of criteria ^^^^^^^^^^^^^^^^^^^^^^^^ https://fairmlbook.org/demographic.html#a-dictionary-of-criteria Algorithmic Interventions ------------------------- .. automodule:: responsibly.fairness.interventions Threshold - Post-processing ^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. automodule:: responsibly.fairness.interventions.threshold :members: