SHapley additive explanations
Mathematical Methods AIMech:SHAP DRAFT
Aliases: SHAP
An additive feature-attribution framework assigning Shapley values to a prediction through a specified function for evaluating feature subsets.
Identity Notes & Concept Scope
SHAP denotes the attribution framework, not the software package or any one estimator such as Kernel SHAP. LIME instead fits a locally weighted surrogate objective that need not recover these Shapley values.
Claims & Grounded Evidence
1 Claim
Mechanism
#mechanism
In the paper's conditional-expectation formulation, feature attributions average marginal changes in expected model output across feature orderings and combine with a baseline prediction.
⚠️ Curated Limitations & Negative Results
- Feature-subset evaluation and dependence assumptions affect the interpretation; the paper distinguishes conditional expectations from independence approximations.
- Attribution values do not by themselves establish causal effects or that a model is safe to deploy.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:Model
Model
|
related_match | The framework attributes predictions of a predictive model through feature-subset evaluations. This directly supported analysis relationship does not equate the explanation method with a model or assert a model superclass. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
original-paper
|
Scott M. Lundberg, Su-In Lee
|
primary paper
full_text
|
sha256:32518459c313…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-11T03:12:53Z
Updated: 2026-10-11T04:02:04Z
Source File:
data/curated/SHAP.yaml