SocioProbe: What, When, and Where Language Models Learn about Sociodemographics

TL;DR

This paper investigates whether pre-trained language models encode sociodemographic information like gender and age, revealing that such knowledge is present but varies across layers and requires extensive pre-training data.

Contribution

It introduces probing methods to analyze sociodemographic knowledge in PLMs, including multilingual and training data effects, filling a gap in understanding higher-level language knowledge.

Findings

PLMs encode sociodemographic information.

Knowledge is distributed across layers in some models.

More pre-training data enhances sociodemographic knowledge.

Abstract

Pre-trained language models (PLMs) have outperformed other NLP models on a wide range of tasks. Opting for a more thorough understanding of their capabilities and inner workings, researchers have established the extend to which they capture lower-level knowledge like grammaticality, and mid-level semantic knowledge like factual understanding. However, there is still little understanding of their knowledge of higher-level aspects of language. In particular, despite the importance of sociodemographic aspects in shaping our language, the questions of whether, where, and how PLMs encode these aspects, e.g., gender or age, is still unexplored. We address this research gap by probing the sociodemographic knowledge of different single-GPU PLMs on multiple English data sets via traditional classifier probing and information-theoretic minimum description length probing. Our results show that PLMs do encode these sociodemographics, and that this knowledge is sometimes spread across the layers of some of the tested PLMs. We further conduct a multilingual analysis and investigate the effect of supplementary training to further explore to what extent, where, and with what amount of pre-training data the knowledge is encoded. Our overall results indicate that sociodemographic knowledge is still a major challenge for NLP. PLMs require large amounts of pre-training data to acquire the knowledge and models that excel in general language understanding do not seem to own more knowledge about these aspects.