Glossary term
Glossary term
Governance and Compliance
The idea that some notions of fairness are mutually incompatible and cannot be satisfied simultaneously. As a result, there is no single universal metric for quantifying fairness that can be applied to all ML problems.
While this may seem discouraging, incompatibility of fairness metrics doesn't imply that fairness efforts are fruitless. Instead, it suggests that fairness must be defined contextually for a given ML problem, with the goal of preventing harms specific to its use cases.
See "On the (im)possibility of fairness" for a more detailed discussion of the incompatibility of fairness metrics.
Created for this library
A bank's model risk team discusses the incompatibility of fairness metrics with stakeholders to set realistic expectations about which trade-offs apply.
A hiring-tech vendor explains the incompatibility of fairness metrics to enterprise customers so they choose a single primary metric for evaluation.
A health-tech startup acknowledges the incompatibility of fairness metrics in its model card and reports several to make the trade-off explicit.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License