CaSPECT: Discovering Causally Homogeneous Subgroups via Directed Spectral Clustering
2026-07-03 • Machine Learning
Machine Learning
AI summaryⓘ
The authors introduce CaSPECT, a method that groups individuals based on shared cause-and-effect relationships rather than just their characteristics. They first learn a causal network from data, then use this to find subgroups with similar causal pathways influencing treatment effects. Their approach uses statistical tools to ensure the directions of causal effects are reliable and measures treatment impact robustly. They show through tests on simulations and real datasets that CaSPECT can identify groups with meaningful treatment effects even when confounding is strong and without needing a predefined model for treatment assignment.
causal inferencespectral clusteringdirected acyclic graphPC algorithmOrientation Validation Scoreaverage treatment effectdouble machine learningbackdoor criterionpropensity scorecausal homogeneity
Authors
Arghya Pratihar, Shinjon Chakraborty, Swagatam Das
Abstract
We propose \textbf{CaSPECT}, a causal spectral clustering framework for discovering causally homogeneous subgroups from observational data. Rather than clustering in covariate space, CaSPECT defines similarity through the topology of a learned directed acyclic graph (DAG); a bootstrap-stabilised PC algorithm recovers the causal skeleton; a novel \emph{Orientation Validation Score} (OVS) combines PC bootstrap evidence with DirectLiNGAM to orient edges robustly; directed edges are weighted by backdoor-identified average treatment effects estimated via OLS or double machine learning. Chung's directed Laplacian provides a spectral embedding in which individuals close together share the same causal propagation pathways. We establish almost-sure consistency of the full pipeline and validate the method through a controlled simulation study and on LaLonde CPS1, IHDP, and 401(k) datasets, where CaSPECT recovers a positive and statistically significant treatment effect within the causally comparable subpopulation and corrects for severe confounding without requiring a pre-specified propensity score model.