Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Causal Inference in Natural Language Processing: Estimation, Prediction, Interpretation and Beyond

Amir Feder, Katherine A. Keith, Emaad Manzoor, Reid Pryzant, Dhanya Sridhar, Zach Wood-Doughty, Jacob Eisenstein, Justin Grimmer, Roi Reichart, Margaret E. Roberts, Brandon M. Stewart, Victor Veitch, Diyi Yang

TL;DR

This paper surveys the intersection of causality and NLP, arguing that causal reasoning can improve robustness, fairness, and interpretability in language technologies. It organizes the field into two strands: estimating causal effects from text where text can be a confounder, outcome, or treatment, and using causal formalisms to build robust and explainable NLP models. It reviews key estimands, identification assumptions, and graphical models, surveys existing methods for text-based causal estimation, and outlines challenges and opportunities in benchmarks, representation learning, and controllable generation. The goal is to make causal reasoning an explicit ingredient in NLP research, enabling more reliable, transparent, and principled language understanding.

Abstract

A fundamental goal of scientific research is to learn about causal relationships. However, despite its critical role in the life and social sciences, causality has not had the same importance in Natural Language Processing (NLP), which has traditionally placed more emphasis on predictive tasks. This distinction is beginning to fade, with an emerging area of interdisciplinary research at the convergence of causal inference and language processing. Still, research on causality in NLP remains scattered across domains without unified definitions, benchmark datasets and clear articulations of the challenges and opportunities in the application of causal inference to the textual domain, with its unique properties. In this survey, we consolidate research across academic areas and situate it in the broader NLP landscape. We introduce the statistical challenge of estimating causal effects with text, encompassing settings where text is used as an outcome, treatment, or to address confounding. In addition, we explore potential uses of causal inference to improve the robustness, fairness, and interpretability of NLP models. We thus provide a unified overview of causal inference for the NLP community.

Causal Inference in Natural Language Processing: Estimation, Prediction, Interpretation and Beyond

TL;DR

This paper surveys the intersection of causality and NLP, arguing that causal reasoning can improve robustness, fairness, and interpretability in language technologies. It organizes the field into two strands: estimating causal effects from text where text can be a confounder, outcome, or treatment, and using causal formalisms to build robust and explainable NLP models. It reviews key estimands, identification assumptions, and graphical models, surveys existing methods for text-based causal estimation, and outlines challenges and opportunities in benchmarks, representation learning, and controllable generation. The goal is to make causal reasoning an explicit ingredient in NLP research, enabling more reliable, transparent, and principled language understanding.

Abstract

A fundamental goal of scientific research is to learn about causal relationships. However, despite its critical role in the life and social sciences, causality has not had the same importance in Natural Language Processing (NLP), which has traditionally placed more emphasis on predictive tasks. This distinction is beginning to fade, with an emerging area of interdisciplinary research at the convergence of causal inference and language processing. Still, research on causality in NLP remains scattered across domains without unified definitions, benchmark datasets and clear articulations of the challenges and opportunities in the application of causal inference to the textual domain, with its unique properties. In this survey, we consolidate research across academic areas and situate it in the broader NLP landscape. We introduce the statistical challenge of estimating causal effects with text, encompassing settings where text is used as an outcome, treatment, or to address confounding. In addition, we explore potential uses of causal inference to improve the robustness, fairness, and interpretability of NLP models. We thus provide a unified overview of causal inference for the NLP community.

Paper Structure

This paper contains 24 sections, 7 equations, 1 figure.

Table of Contents

  1. Introduction
  2. Background
  3. Causal Estimands
  4. Identification Assumptions for Causal Inference
  5. Causal Graphical Models
  6. Estimating Causal Effects with Text
  7. Causal Effects with Textual Confounders
  8. Previous approaches.
  9. Challenges for causal assumptions with text.
  10. Causal Effects on Textual Outcomes
  11. Challenges for causal assumptions with text.
  12. Causal Effects with Textual Treatments
  13. Previous approaches.
  14. Challenges for causal assumptions with text.
  15. Future Work
  16. ...and 9 more sections

Figures (1)

  • Figure 1: Causal graphs for the motivating examples. (Left) In \ref{['eg:1']}, the post icon ($T$) is correlated with attributes of the post ($X$), and both variables affect the number of likes a post receives ($Y$). (Right) In \ref{['eg:2']}, the label ($Y$, i.e., diagnosis) and hospital site ($Z$) are correlated, and both affect the clinical narrative ($X$). Predictions $f(X)$ from the trained classifier depend on $X$.

Theorems & Definitions (2)

  • Example 1
  • Example 2