Implementing Fairness Metrics

A fairness metric is only as trustworthy as its implementation: robust code counts once per group, returns undefined rather than zero for empty denominators, reports group sizes, handles missing and sparse groups deliberately and is validated against reference libraries.