Fighting Spurious Correlations in Text Classification via a Causal Learning Perspective
Yuqing Zhou, Ziwei Zhu
TL;DR
Spurious correlations degrade text classifier robustness under distribution shift. CCR integrates counterfactual-based causal feature selection with an unbiased IPW loss to emphasize causal features, supported by covariance-based disentanglement and a two-stage training regime. The approach introduces PN, PS, and PNS concepts and derives a practical loss L_IPW augmented by a PNS-based constraint, achieving state-of-the-art performance among non-group-label methods and competitive results with group-label baselines on four datasets. Empirically, CCR yields strong worst-group accuracy improvements and interpretable reductions in reliance on spurious cues, offering a scalable path to robust text classification without costly group annotations.
Abstract
In text classification tasks, models often rely on spurious correlations for predictions, incorrectly associating irrelevant features with the target labels. This issue limits the robustness and generalization of models, especially when faced with out-of-distribution data where such spurious correlations no longer hold. To address this challenge, we propose the Causally Calibrated Robust Classifier (CCR), which aims to reduce models' reliance on spurious correlations and improve model robustness. Our approach integrates a causal feature selection method based on counterfactual reasoning, along with an unbiased inverse propensity weighting (IPW) loss function. By focusing on selecting causal features, we ensure that the model relies less on spurious features during prediction. We theoretically justify our approach and empirically show that CCR achieves state-of-the-art performance among methods without group labels, and in some cases, it can compete with the models that utilize group labels.