Equal opportunity in binary classification
Evaluation Criteria AIMech:EqualOpportunity DRAFT
Aliases: Equal opportunity (true-positive parity), Equality of opportunity (true-positive parity)
A binary-classification fairness criterion requiring equal probabilities of a positive prediction across specified protected groups among individuals whose target outcome is the designated positive outcome.
Identity Notes & Concept Scope
The identity is Hardt, Price and Srebro's Definition 2.2, with the positive target designated as the advantaged outcome. It is not the wider philosophical meaning of equal opportunity, the full equalized-odds condition, or a threshold-adjustment algorithm. Identity searches covered all 442 AIO labels and scoped synonyms, 94 preceding native records and ignored/hidden-inclusive text in both checkouts.
Claims & Grounded Evidence
1 Claim
Scope
#criterion-scope
For each compared protected group a, evaluate P(prediction=1 | A=a,Y=1) and require equality across groups. This constrains true-positive rates and imposes no corresponding equality requirement at Y=0. It relaxes the binary equalized-odds criterion by retaining only the designated positive-outcome condition.
⚠️ Curated Limitations & Negative Results
- The definition depends on which target outcome is designated positive/advantaged; changing that designation changes the constrained population.
- Equal opportunity alone does not equalize false-positive rates, selection rates, predictive values or individual treatment.
- The criterion requires meaningful target labels and sufficiently supported group/positive-outcome strata. Finite-sample uncertainty is not eliminated by writing an equality.
- Like other oblivious criteria, this condition does not determine the causal process or establish overall social fairness, empirical utility or legal compliance.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:BinaryClassification
Binary Classification
|
related_match | The criterion directly constrains true-positive rates of binary classifiers. BinaryClassification is the task on which the condition is evaluated; it is not a broader class of fairness criteria or an equivalent identity. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
original-paper
|
Moritz Hardt, Eric Price, Nathan Srebro
|
primary paper
full_text
|
ID: arXiv:1610.02413
sha256:e92b4f579180…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-11T04:48:50Z
Updated: 2026-10-11T04:48:50Z
Source File:
data/curated/EqualOpportunity.yaml