Introduces 'deconfounding scores' to enable reliable causal effect estimation even when treatment and control groups have very little overlap.
April 2, 2026
Original Paper
Deconfounding Scores and Representation Learning for Causal Effect Estimation with Weak Overlap
arXiv · 2604.00811
The Takeaway
In high-dimensional real-world data, the 'positivity' assumption (overlap) often fails, making causal inference brittle or impossible. This framework provides a way to learn representations that optimize for overlap while preserving the ability to identify causal effects, directly addressing the curse of dimensionality in policy evaluation and A/B testing.
From the abstract
Overlap, also known as positivity, is a key condition for causal treatment effect estimation. Many popular estimators suffer from high variance and become brittle when features differ strongly across treatment groups. This is especially challenging in high dimensions: the curse of dimensionality can make overlap implausible. To address this, we propose a class of feature representations called deconfounding scores, which preserve both identification and the target of estimation; the classical pr