Equalized odds
Equalized odds: A fairness criterion requiring comparable error rates across groups conditional on the actual class. It considers true- and false-positive rates by group. It may conflict with calibration and other criteria when prevalence differs. This perspective connects the definition with practical behaviour in artificial intelligence systems.
In italiano: Equalized OddsLegge V · Overfitting, bias e collasso dei modelliIt considers true- and false-positive rates by group. It may conflict with calibration and other criteria when prevalence differs.