Fairness parity#
Equalising a chosen metric across protected groups.
Important
✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.
What it is#
Fairness parity is the family of group-fairness criteria that demand some metric be equal across demographic groups — that the model treat protected groups comparably. Which metric you equalize defines the flavor of parity.
The main flavors#
Demographic (statistical) parity equalizes the positive-outcome rate across groups; equal opportunity equalizes the true-positive rate (recall) among those who qualify; equalized odds equalizes both TPR and FPR; predictive parity equalizes precision. Each is measured as a difference or a ratio between groups.
Why it’s hard#
The different parities conflict — impossibility results show you generally can’t satisfy all at once (the COMPAS debate turned on predictive parity holding while equalized odds failed). Enforcing any parity also usually costs accuracy, so teams choose the criterion that fits the harm they most need to prevent. The four-fifths (80%) rule is a common legal threshold.
Theme: Fairness & Calibration · All terminology
Hint
Mind map — connected ideas
Selection Rate · Recall · Fairness Guardrails · Fair Lending laws · High-Stakes Domains · Precision (a.k.a. Positive Predictive Value, PPV)
Hint
More in Fairness & Calibration
Demographic Parity (Statistical Parity) · Equal Opportunity (Fairness) · Equalized Odds (Fairness) · Fairness Guardrails · Four-Fifths (80%) Rule · Predictive Parity (Calibration) · Selection Rate
See also
Source article Adapted (context, re-expressed) in our own words from: Fairness parity (insightful-data-lab.com).