November 2026
368 pages
13h 30m
English
|
Priority |
Metric |
Concise explanation (plain) |
Why it matters |
Healthcare example (glaucoma screening) |
|---|---|---|---|---|
|
1 |
Calibration |
Predicted risk scores match actual disease likelihood equally across groups. |
Bedrock of trust: Without calibration, predictions are clinically meaningless, making all other fairness checks secondary |
A risk score of 0.7 must mean ˜70% chance of glaucoma for every ethnic group |
|
2 |
False-Negative Rate (FNR) parity/equal opportunity |
Ensures diseased patients have the same chance of being correctly diagnosed across groups ... |
Read now
Unlock full access