# Explaining The Efficacy of Counterfactually Augmented Data

arXiv: https://arxiv.org/abs/2010.02114 · HTML: https://mlsanity.com/paper/2010.02114/

## Key facts

- Cited by 65 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Oct 2020.

## Built on by (most cited first, 30 of 65)

- [A Survey of Data Augmentation Approaches for NLP](https://mlsanity.com/paper/2105.03075/) (2021)
- [Last Layer Re-Training is Sufficient for Robustness to Spurious Correlations](https://mlsanity.com/paper/2204.02937/) (2022)
- [Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks](https://mlsanity.com/paper/2307.02477/) (2023)
- [Causal Inference in Natural Language Processing: Estimation, Prediction, Interpretation and Beyond](https://mlsanity.com/paper/2109.00725/) (2021)
- [On Feature Learning in the Presence of Spurious Correlations](https://mlsanity.com/paper/2210.11369/) (2022)
- [Competency Problems: On Finding and Removing Artifacts in Language Data](https://mlsanity.com/paper/2104.08646/) (2021)
- [Diffusion Causal Models for Counterfactual Estimation](https://mlsanity.com/paper/2202.10166/) (2022)
- [Counterfactual Invariance to Spurious Correlations: Why and How to Pass Stress Tests](https://mlsanity.com/paper/2106.00545/) (2021)
- [GLUE-X: Evaluating Natural Language Understanding Models from an Out-of-distribution Generalization Perspective](https://mlsanity.com/paper/2211.08073/) (2022)
- [Faithful Explanations of Black-box NLP Models Using LLM-generated Counterfactuals](https://mlsanity.com/paper/2310.00603/) (2023)
- [An Investigation of the (In)effectiveness of Counterfactually Augmented Data](https://mlsanity.com/paper/2107.00753/) (2021)
- [Probing Classifiers are Unreliable for Concept Removal and Detection](https://mlsanity.com/paper/2207.04153/) (2022)
- [DoCoGen: Domain Counterfactual Generation for Low Resource Domain Adaptation](https://mlsanity.com/paper/2202.12350/) (2022)
- [Equitable modelling of brain imaging by counterfactual augmentation with morphologically constrained 3D deep generative models](https://mlsanity.com/paper/2111.14923/) (2021)
- [On the Efficacy of Adversarial Data Collection for Question Answering: Results from a Large-Scale Randomized Study](https://mlsanity.com/paper/2106.00872/) (2021)
- [Exploring the Efficacy of Automatically Generated Counterfactuals for Sentiment Analysis](https://mlsanity.com/paper/2106.15231/) (2021)
- [A Rationale-Centric Framework for Human-in-the-loop Machine Learning](https://mlsanity.com/paper/2203.12918/) (2022)
- [On Behalf of the Stakeholders: Trends in NLP Model Interpretability in the Era of LLMs](https://mlsanity.com/paper/2407.19200/) (2024)
- [NeuroCounterfactuals: Beyond Minimal-Edit Counterfactuals for Richer Data Augmentation](https://mlsanity.com/paper/2210.12365/) (2022)
- [Informativeness and Invariance: Two Perspectives on Spurious Correlations in Natural Language](https://mlsanity.com/paper/2204.04487/) (2022)
- [Challenges in Applying Explainability Methods to Improve the Fairness of NLP Models](https://mlsanity.com/paper/2206.03945/) (2022)
- [Evaluating LLMs' Mathematical and Coding Competency through Ontology-guided Interventions](https://mlsanity.com/paper/2401.09395/) (2024)
- [Necessity and Sufficiency for Explaining Text Classifiers: A Case Study in Hate Speech Detection](https://mlsanity.com/paper/2205.03302/) (2022)
- [Nuisances via Negativa: Adjusting for Spurious Correlations via Data Augmentation](https://mlsanity.com/paper/2210.01302/) (2022)
- [A Review of the Role of Causality in Developing Trustworthy AI Systems](https://mlsanity.com/paper/2302.06975/) (2023)
- [Data Augmentations for Improved (Large) Language Model Generalization](https://mlsanity.com/paper/2310.12803/) (2023)
- [Towards Fine-grained Causal Reasoning and QA](https://mlsanity.com/paper/2204.07408/) (2022)
- [Flattery, Fluff, and Fog: Diagnosing and Mitigating Idiosyncratic Biases in Preference Models](https://mlsanity.com/paper/2506.05339/) (2025)
- [Making a (Counterfactual) Difference One Rationale at a Time](https://mlsanity.com/paper/2201.05177/) (2022)
- [GDA: Generative Data Augmentation Techniques for Relation Extraction Tasks](https://mlsanity.com/paper/2305.16663/) (2023)

## Abstract (arXiv)

In attempts to produce ML models less reliant on spurious patterns in NLP datasets, researchers have recently proposed curating counterfactually augmented data (CAD) via a human-in-the-loop process in which given some documents and their (initial) labels, humans must revise the text to make a counterfactual label applicable. Importantly, edits that are not necessary to flip the applicable label are prohibited. Models trained on the augmented data appear, empirically, to rely less on semantically irrelevant words and to generalize better out of domain. While this work draws loosely on causal thinking, the underlying causal model (even at an abstract level) and the principles underlying the observed out-of-domain improvements remain unclear. In this paper, we introduce a toy analog based on linear Gaussian models, observing interesting relationships between causal models, measurement noise, out-of-domain generalization, and reliance on spurious signals. Our analysis provides some insights that help to explain the efficacy of CAD. Moreover, we develop the hypothesis that while adding noise to causal features should degrade both in-domain and out-of-domain performance, adding noise to non-causal features should lead to relative improvements in out-of-domain performance. This idea inspires a speculative test for determining whether a feature attribution technique has identified the causal spans. If adding noise (e.g., by random word flips) to the highlighted spans degrades both in-domain and out-of-domain performance on a battery of challenge datasets, but adding noise to the complement gives improvements out-of-domain, it suggests we have identified causal spans. We present a large-scale empirical study comparing spans edited to create CAD to those selected by attention and saliency maps. Across numerous domains and models, we find that the hypothesized phenomenon is pronounced for CAD.