Fairness
Fairness: Evaluating and managing a model’s impacts and disparities on individuals and groups in a defined context. Demographic metrics, group errors and process analysis measure different and sometimes incompatible aspects. Choosing among them requires social and legal criteria, not technical optimization alone.
In italiano: FairnessLegge V · Overfitting, bias e collasso dei modelliDemographic metrics, group errors and process analysis measure different and sometimes incompatible aspects. Choosing among them requires social and legal criteria, not technical optimization alone.
Examples
- Group false-negative rates
- Equitable service access
- Decision auditing