From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Models
Pretraining data political biases propagate through language models into unfair hate speech and misinformation detectors, reinforcing polarization.
Published 2023138 citationsPaper ↗
Only vote on papers you've read. Sign in with GitHub to vote.
Valued for tracing pretraining political bias into downstream hate speech and misinformation fairness, the study is undermined by correlational methods that leave label bias unexamined and offer no mitigation or code.
Abstract
Language models (LMs) are pretrained on diverse data sources, including news, discussion forums, books, and online encyclopedias.A significant portion of this data includes opinions and perspectives which, on one hand, celebrate democracy and diversity of ideas, and on the other hand are inherently socially biased.Our work develops new methods to (1) measure political biases in LMs trained on such corpora, along social and economic axes, and (2) measure the fairness of downstream NLP models trained on top of politically biased LMs.We focus on hate speech and misinformation detection, aiming to empirically quantify the effects of political (social, economic) biases in pretraining data on the fairness of high-stakes social-oriented tasks.Our findings reveal that pretrained LMs do have political leanings that reinforce the polarization present in pretraining corpora, propagating social biases into hate speech predictions and misinformation detectors.We discuss the implications of our findings for NLP research and propose future directions to mitigate unfairness. 1 Warning: This paper contains examples of hate speech.