Name: AI Ethics Lab
Author: MySimulator

Threshold controls

Separate threshold per group Group A threshold: 0.50 Group B threshold: 0.50

Fairness metrics

Approval rate — A—

Approval rate — B—

Demographic parity gap—

True positive rate — A—

True positive rate — B—

Equal opportunity gap—

Overall accuracy—

Group B's population is generated with a lower positive base rate — a common real-world pattern caused by historical data bias, not by anything inherent to the group.

Score distributions by group