AIF360 popular metrics
The analysis below computes all fairness metrics provided by the AIF360 library's core given a model and dataset. These metrics are often used in both research literature and in practical applications.
As an older fairness framework, AIF360 lacks some features. For example, it does not support intersectional of sensitive attributes, and focuses on supporting ad-hoc instead of standardized exploration. The side-by-side comparison between multiple sensitive attributes with multiple values is a generalization engineered by MAI-BIAS.