Demographic Parity
noun
say it: “dem-uh-GRAF-ik PAIR-ih-tee” /ˌdɛməˈɡræfɪk ˈpɛrəti/
A fairness rule that a model should give a positive decision to the same share of people in every group.
What is demographic parity?
Demographic parity is a fairness test that asks for the same selection rate in every group: the same share of each
district invited, whoever qualifies. Fairlearn measures the gap with demographic_parity_difference(). It ignores
whether the groups' real qualification rates differ, which equalized odds takes into account.
from fairlearn.metrics import demographic_parity_difference
qualified = [1, 0, 1, 0, 1, 0, 1, 0]
invited = [1, 1, 1, 0, 1, 0, 0, 0]
district = ["N", "N", "N", "N", "S", "S", "S", "S"]
print(demographic_parity_difference(qualified, invited, sensitive_features=district)) # 0.5
In a sentence
- “The screen missed demographic parity by 0.1: North's rate was 0.3, South's 0.2.”
- “Demographic parity would mean inviting the same share of applicants from every district.”
Level: Freelancer · More AI Basics terms at this level
Updated