Causal Direction of Data Collection Matters: Implications of Causal and Anticausal Learning for NLP

TL;DR

This paper explores how the causal direction in data collection impacts NLP tasks, revealing that understanding causal mechanisms can explain variations in semi-supervised learning and domain adaptation performance.

Contribution

It introduces the first analysis of the independent causal mechanisms principle in NLP, linking causal direction to empirical NLP outcomes and offering new insights for modeling.

Findings

Results align with causal theory predictions

Meta-analysis supports causal explanations for SSL and DA differences

Provides guidelines for future NLP modeling based on causal insights

Abstract

The principle of independent causal mechanisms (ICM) states that generative processes of real world data consist of independent modules which do not influence or inform each other. While this idea has led to fruitful developments in the field of causal inference, it is not widely-known in the NLP community. In this work, we argue that the causal direction of the data collection process bears nontrivial implications that can explain a number of published NLP findings, such as differences in semi-supervised learning (SSL) and domain adaptation (DA) performance across different settings. We categorize common NLP tasks according to their causal direction and empirically assay the validity of the ICM principle for text data using minimum description length. We conduct an extensive meta-analysis of over 100 published SSL and 30 DA studies, and find that the results are consistent with our expectations based on causal insights. This work presents the first attempt to analyze the ICM principle in NLP, and provides constructive suggestions for future modeling choices. Code available at https://github.com/zhijing-jin/icm4nlp