Explanation-aware clustering

Research Method · All research methods
Evidence · 2 claims (2 for) · 1 study (1 associational), q2 · 0 of 1 report an effect size · 2 claims rest on one study

Description

This framework pairs a model explainer with unsupervised learning to reveal heterogeneity hidden by binary classification. LIME "perturbs the input data locally around a single instance and fits a simple interpretable model" to identify predictors contributing to class membership; clustering of these LIME profiles (via K-Means, cluster number selected by lowest Davies-Bouldin Index) then identifies subgroups within each class, addressing explainer models' "tendency to overstate class homogeneity due to such local approximations". The article uses it to uncover 2 responder and 4 non-responder subgroups with distinct longitudinal outcomes.

Accounts

• Explanation-aware clustering: combining model explainers with unsupervised clustering to uncover heterogeneous subgroups within predicted classes: This framework pairs a model explainer with unsupervised learning to reveal heterogeneity hidden by binary classification. LIME "perturbs the input data locally around a single instance and fits a simple interpretable model" to identify predictors contributing to class membership; clustering of these LIME profiles (via K-Means, cluster number selected by lowest Davies-Bouldin Index) then identifies subgroups within each class, addressing explainer models' "tendency to overstate class homogeneity due to such local approximations". The article uses it to uncover 2 responder and 4 non-responder subgroups with distinct longitudinal outcomes. (Tan et al. (2026))

Claims

• K-Means clustering of LIME feature-importance profiles identifies 2 responder and 4 non-responder subgroups within binary intervention-response classes [+M]
• Subgroup membership predicted distinct long-term literacy trajectories 68.2 months after intervention ended [+M]

Related Research Methods

-

Key Sources

• Tan, A.S.-C., & Ali, F. (2026). Uncovering subgroup heterogeneity in dyslexia intervention outcomes using explainable machine learning. npj Science of Learning. https://doi.org/10.1038/s41539-026-00429-3

Page metadata

Field	Value
type	research-method
id	explanation-aware-clustering
title	Explanation-aware clustering
description	This framework pairs a model explainer with unsupervised learning to reveal heterogeneity hidden by binary classification.
status	review
generated	by process:settle-candidates on 2026-10-26
trust	unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.