Grid-search reduction for fair classification
Algorithms AIMech:GridSearchFairClassification DRAFT
Aliases: Agarwal grid-search fairness reduction
A fairness-aware classification reduction that evaluates cost-sensitive classifier best responses over a grid of Lagrange-multiplier-derived costs and selects a deterministic candidate with a desired empirical accuracy–fairness tradeoff.
Identity Notes & Concept Scope
The paper’s Section 3.4 grid search is a separate reduction producing deterministic candidates. It shares a cost-sensitive oracle with the randomized exponentiated-gradient reduction but does not inherit that procedure’s mixture or saddle-point stopping rule. It is not generic hyperparameter grid search or the bivariate causal-sensitivity grid procedure. All 442 pinned AIO labels/scoped synonyms and 104 preceding native identities were compared, together with hidden/ignored-inclusive searches of both checkouts; source citations and pending frontier mentions were distinguished from existing concept identities.
Claims & Grounded Evidence
1 Claim
Mechanism
#method-scope
Use the same linear conditional-moment constraint formulation and per-example cost-sensitive oracle as the exponentiated-gradient reduction, but enumerate candidate multiplier values or equivalent cost adjustments. For binary protected groups the paper reduces demographic-parity search to one adjustment dimension and equalized-odds search to two, then compares candidate classifiers to select the desired tradeoff.
⚠️ Curated Limitations & Negative Results
- A finite grid and heuristic cost-sensitive learner do not guarantee a globally optimal fair deterministic classifier.
- The paper explicitly allows lower accuracy or constraint violation when choosing a deterministic candidate; the randomized-mixture guarantees should not be transferred automatically.
- Grid search is intended for small numbers of effective constraint dimensions and scales poorly as that dimension grows.
- Fairness is assessed under the specified empirical constraints and sampling context; evaluation on unseen or shifted data is separate.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:BinaryClassification
Binary Classification
|
related_match | Constructs binary classification candidates using fairness-dependent costs. This relates the reduction to its task without equating algorithm and task. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
original-paper
|
Alekh Agarwal, Alina Beygelzimer, Miroslav Dudík, John Langford, Hanna Wallach
|
primary paper
full_text
|
ID: https://proceedings.mlr.press/v80/agarwal18a.html
sha256:bcdbc1454b2a…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-11T05:03:15Z
Updated: 2026-10-11T05:03:15Z
Source File:
data/curated/GridSearchFairClassification.yaml