Calibration Testing Across (Sub)populations

From The Foundation for Best Practices in Machine Learning
Technical Best Practices > Fairness & Non-Discrimination > Calibration Testing Across (Sub)populations

Calibration Testing Across (Sub)populations

Control

If applicable, test Model(s) for calibration. Evaluate whether (Sub)populations members with the same predicted Outcome have an equal probability of actually being in the positive class.


Aim

To (a) ensure that Subpopulations each have the same likelihood of deserving the Positive Outcome for a given Model prediction; and (b) highlight associated risks that might occur in the Product Lifecycle.


Additional Information