fairscoring.metrics

A collection fairness metrics to measure the bias of scores.

The package consists of multiple classes that cover different fairness concepts and distance metrics. A concrete fairness metric is an instance of these classes with a concrete parametrization.

There are recommended default instances are already included:

Metric	Bias	Class
bias_metric_cal	Calibration Bias	CalibrationMetric
bias_metric_eo	Equal Opportunity Bias	WassersteinMetric
bias_metric_pe	Predictive Equality Bias	WassersteinMetric

Changed in version 0.2.0: The bias_ prefixed constants are now prefixed by bias_metric_

Submodules

• fairscoring.metrics.base
• fairscoring.metrics.calibration
• fairscoring.metrics.integral
• fairscoring.metrics.roc