Counterfactual Fairness

A fairness criterion stating that a decision is fair if it would remain the same in a counterfactual world where an individual's protected attribute were different.

Ask yourself: if everything about a person stayed the same except their race or gender, would the AI still make the same decision? If yes, it's counterfactually fair.

This concept is valuable for AI governance because it provides a rigorous way to test whether protected attributes are influencing decisions. Including counterfactual analysis in your AI audit processes strengthens your fairness assurance.