Fairness Metrics That Conflict
Discover why you mathematically cannot satisfy all fairness criteria at once. Explore five key metrics, see how they trade off against each other, and learn which to prioritise for your use case.
Fairness Metric Explorer
Click any metric card to explore what it measures, how to think about it visually, and when it matters most in HR contexts. Formulas are presented intuitively, not mathematically.
Trade-off Visualiser
Use the slider to adjust the model decision threshold and watch how different fairness metrics respond. Notice how improving one metric often worsens another — this is the fundamental tension in algorithmic fairness.
Which Metrics Matter Most?
Select your HR use case to see which fairness metrics should take priority and why. Context determines which trade-offs are acceptable.
Key insight: The impossibility theorem (proven independently by Chouldechova and Kleinberg et al.) demonstrates that when base rates differ between groups, you mathematically cannot satisfy demographic parity, equalised odds, and predictive parity simultaneously. This is not a technical limitation to solve — it is a fundamental mathematical reality. The question is not “which metric is correct?” but “which trade-off is most acceptable for this specific context?”