On the Independence of Bias and Fairness in Language Models

Laura Cabello, Anders Søgaard

Keywords: bias, fairness, language models, NLP

TL;DR: Association bias and empirical fairness in language models can be completely independent

Abstract: The societal impact of pretrained language models has prompted researchers to probe them for strong associations between protected attributes and value-loaded terms, from slur to prestigious job titles. Such work is said to probe models for bias or fairness--or such probes 'into representational biases' are said to be 'motivated by fairness'--suggesting an intimate connection between bias and fairness. We provide conceptual clarity by distinguishing between association biases and empirical fairness and show the two can be independent.

Submission Number: 16