Causal Estimation for Text Data with (Apparent) Overlap Violations
This addresses a critical issue in causal inference for text data, enabling more reliable effect estimation in applications like analyzing email politeness effects, though it is incremental in adapting existing non-parametric estimation techniques.
The paper tackles the problem of estimating causal effects from text data when treatment attributes are perfectly determined by the text, which violates the overlap assumption required for causal inference. It proposes a method using supervised representation learning to adjust for confounders while satisfying overlap, resulting in a low-bias estimator with valid uncertainty quantification and showing strong improvements in bias and uncertainty relative to baselines.
Consider the problem of estimating the causal effect of some attribute of a text document; for example: what effect does writing a polite vs. rude email have on response time? To estimate a causal effect from observational data, we need to adjust for confounding aspects of the text that affect both the treatment and outcome -- e.g., the topic or writing level of the text. These confounding aspects are unknown a priori, so it seems natural to adjust for the entirety of the text (e.g., using a transformer). However, causal identification and estimation procedures rely on the assumption of overlap: for all levels of the adjustment variables, there is randomness leftover so that every unit could have (not) received treatment. Since the treatment here is itself an attribute of the text, it is perfectly determined, and overlap is apparently violated. The purpose of this paper is to show how to handle causal identification and obtain robust causal estimation in the presence of apparent overlap violations. In brief, the idea is to use supervised representation learning to produce a data representation that preserves confounding information while eliminating information that is only predictive of the treatment. This representation then suffices for adjustment and can satisfy overlap. Adapting results on non-parametric estimation, we find that this procedure is robust to conditional outcome misestimation, yielding a low-bias estimator with valid uncertainty quantification under weak conditions. Empirical results show strong improvements in bias and uncertainty quantification relative to the natural baseline.