Abstract
We formalize a notion of differential robustness. An algorithm is differentially robust to an event if the event has disparate impact on the performance of an algorithm for different groups in the population. We illustrate it with a case study of the real world deployment of a predictive algorithm in a European public employment service.
| Original language | English |
|---|---|
| Journal | CEUR Workshop Proceedings |
| Volume | 3442 |
| Publication status | Published - 2023 |
| Event | 2nd European Workshop on Algorithmic Fairness, EWAF 2023 - Winterthur, Switzerland Duration: 7 Jun 2023 → 9 Jun 2023 |
Keywords
- algorithmic fairness
- causal mechanism
- concept drift
- differential robustness
- disparate impact
- distribution shift