Equalized odds¶

A classifier fairness criterion requiring protected groups to have equal true-positive and false-positive rates conditional on the actual outcome.

Version: v3 · 2026-09-19 · History
Domain-specific #: 1775
Domain group: Applied Sciences & Engineering
Origin domain: Computer Science & Software Engineering
Subdomains: Machine Learning, Group Fairness for Supervised Prediction → Computer Science & Software Engineering
Aliases: Equalized odds criterion

Core Idea¶

Equalized odds is a classifier fairness criterion requiring protected groups to have equal true-positive and false-positive rates conditional on the actual outcome.

A predictor satisfies equalized odds with respect to protected attribute A and outcome Y when its prediction is conditionally independent of A given Y. In binary classification, this requires equal true-positive rates and equal false-positive rates across protected groups, or equivalently equal prediction distributions within each actual-outcome stratum.

Its operative boundary is not supplied by the name alone. Preserve this identity: A classifier fairness criterion requiring protected groups to have equal true-positive and false-positive rates conditional on the actual outcome.

How would you explain it like I'm…

Same Mistakes for Every Team

Imagine a robot that guesses which kids are sick. To be fair this way, it should catch sick kids just as well in every group. And it should make the mistake of calling a healthy kid sick just as often in every group too.

Equal Hits, Equal Misses

Equalized odds is a fairness rule for computer programs that sort people into yes or no. First split people by what is actually true—say, who really will repay a loan and who won't. Among the people who really will, the program should say 'yes' at the same rate for every group. Among the people who really won't, it should wrongly say 'yes' at the same rate for every group too. Just giving every group the same number of 'yes' answers is not enough.

Equal Error Rates Given Truth

Equalized odds is a fairness criterion for classifiers. It requires protected groups to have the same true-positive rate (the share of actual positives correctly flagged) and the same false-positive rate (the share of actual negatives wrongly flagged). The key is that the comparison is made conditional on the real outcome. That distinguishes it from demographic parity, which only asks that each group gets positive predictions at the same overall rate—satisfying that alone does not satisfy equalized odds. Both rates must match; matching only one isn't enough.

Equalized odds is a classifier fairness criterion defined with respect to a protected attribute A and a true outcome Y. A predictor satisfies it when its prediction is conditionally independent of A given Y. In binary classification this means equal true-positive rates and equal false-positive rates across protected groups—equivalently, the distribution of predictions is the same across groups within each actual-outcome stratum. Because it conditions on ground truth, it differs from demographic parity, which requires equal positive-prediction rates unconditionally; demographic parity alone does not satisfy equalized odds. Equalizing only one of the two error-related rates is likewise insufficient. The criterion thus constrains how the classifier's errors and successes are distributed among groups, given what actually happened.

Scope of Application¶

The abstraction recurs literally within binary and multiclass supervised decisions where protected-group error rates are audited or constrained. The following habitats preserve the same recognition machinery; they are not invitations to extend the name metaphorically.

Lending. approval errors are compared conditional on repayment outcome.
Medical prediction. sensitivity and false-positive rates are compared across patient groups.
Risk assessment. error-rate disparities are measured by outcome strata.
Post-processing. randomized thresholds move a score onto an equalized-odds frontier.
Model selection. accuracy and group-conditional error constraints are optimized jointly.

Clarity¶

Declare A, Y, the prediction threshold, positive class, population, and whether equality is exact or tolerance-bounded. Report uncertainty for every group–outcome cell. Equalized odds can be technically satisfied against labels that encode unequal treatment, so label validity and decision consequences remain separate normative questions.

A practical identification audit begins with the typed roles rather than the title: establish the protected attribute, verify the ground-truth outcome, then test the remaining conditions and exclusions.

Manages Complexity¶

The criterion decomposes disparity into false-negative and false-positive treatment conditional on actual status. ROC geometry makes feasible post-processing explicit and reveals the performance cost of equalizing rates.

The compression remains accountable because each simplification has a named failure condition. Disagreement can be localized to a missing role, an invalid assumption, an ambiguous measurement, or a neighboring abstraction instead of being hidden inside an unanalyzed label.

Abstract Reasoning¶

R1. Define protected groups, outcome labels, prediction, and decision population. R2. Estimate true- and false-positive rates within every adequately sampled group. R3. Test both rate equalities with uncertainty or a declared tolerance. R4. If post-processing, derive group-specific randomized thresholds from the validation ROC curves. R5. Audit label quality, calibration, intersectional groups, and downstream consequences beyond the criterion.

Knowledge Transfer¶

The criterion transfers literally across supervised predictions with protected group, outcome, and both conditional error rates. Fairness and constraint are parents; equal acceptance rates or equal accuracy alone do not satisfy equalized odds.

The transfer boundary is explicit: DOMAIN-SPECIFIC PASS / PRIME FAIL: The criterion is applied across classifiers, protected attributes, and binary outcomes by comparing group-conditioned error rates. Literal recognition retains the specialist vocabulary and validity conditions of machine-learning fairness; outside that setting only broader parent operations transfer.

Relationships to Other Abstractions¶

Current abstraction Equalized odds Domain-specific

Parents (2) — outgoing typed relations

Equalized odds is a kind of Constraint Prime

Constraint (prime:constraint).
Equalized odds is a kind of Fairness Prime

Fairness (prime:fairness).

Hierarchy paths (2) — routes to 2 parentless roots

Equalized odds → is a kind of Constraint

Show alternative path (1)

Neighborhood in Abstraction Space¶

Equalized odds sits in a sparse region of the domain-specific corpus (71^st percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Statistical Bias & Inference Pitfalls (7 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25