An Empirical Study of Metrics to Measure Representational Harms in Pre-Trained Language Models

Saghar Hosseini, Ahmed Hassan Awadallah, Hamid Palangi

Keywords: Natural Language Processing, Fairness, Safety

TL;DR: Measuring implicit hate in pretrained language models

Abstract: Large-scale Pre-Trained Language Models (PTLMs) capture knowledge from massive human-written data which contains latent societal biases and toxic contents. In this paper, we leverage the primary task of PTLMs, i.e. language modeling, and propose a new metric to quantify manifested implicit representational harms in PTLMs towards 13 marginalized demographics. Using this metric, we conducted an empirical analysis of 24 widely used PTLMs. Our analysis provides insights into the correlation between the proposed metric in this work and other related fairness metrics. We observe that our metric correlates with the majority of gender-specific fairness metrics in the literature. Through extensive experiments, we explore the connections between PTLMs architectures and representational harms across two dimensions: depth and width of the networks. We found that prioritizing depth over width, mitigates representational harms in some PTLMs.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Social Aspects of Machine Learning (eg, AI safety, fairness, privacy, interpretability, human-AI interaction, ethics)

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2301.09211/code)