Towards Understanding and Mitigating Social Biases in Language Models

Paul Pu Liang, Chiyu Wu, Louis-Philippe Morency, Ruslan Salakhutdinov
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:6565-6576, 2021.

Abstract

As machine learning methods are deployed in real-world settings such as healthcare, legal systems, and social science, it is crucial to recognize how they shape social biases and stereotypes in these sensitive decision-making processes. Among such real-world deployments are large-scale pretrained language models (LMs) that can be potentially dangerous in manifesting undesirable representational biases - harmful biases resulting from stereotyping that propagate negative generalizations involving gender, race, religion, and other social constructs. As a step towards improving the fairness of LMs, we carefully define several sources of representational biases before proposing new benchmarks and metrics to measure them. With these tools, we propose steps towards mitigating social biases during text generation. Our empirical results and human evaluation demonstrate effectiveness in mitigating bias while retaining crucial contextual information for high-fidelity text generation, thereby pushing forward the performance-fairness Pareto frontier.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-liang21a, title = {Towards Understanding and Mitigating Social Biases in Language Models}, author = {Liang, Paul Pu and Wu, Chiyu and Morency, Louis-Philippe and Salakhutdinov, Ruslan}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {6565--6576}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/liang21a/liang21a.pdf}, url = {https://proceedings.mlr.press/v139/liang21a.html}, abstract = {As machine learning methods are deployed in real-world settings such as healthcare, legal systems, and social science, it is crucial to recognize how they shape social biases and stereotypes in these sensitive decision-making processes. Among such real-world deployments are large-scale pretrained language models (LMs) that can be potentially dangerous in manifesting undesirable representational biases - harmful biases resulting from stereotyping that propagate negative generalizations involving gender, race, religion, and other social constructs. As a step towards improving the fairness of LMs, we carefully define several sources of representational biases before proposing new benchmarks and metrics to measure them. With these tools, we propose steps towards mitigating social biases during text generation. Our empirical results and human evaluation demonstrate effectiveness in mitigating bias while retaining crucial contextual information for high-fidelity text generation, thereby pushing forward the performance-fairness Pareto frontier.} }
Endnote
%0 Conference Paper %T Towards Understanding and Mitigating Social Biases in Language Models %A Paul Pu Liang %A Chiyu Wu %A Louis-Philippe Morency %A Ruslan Salakhutdinov %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-liang21a %I PMLR %P 6565--6576 %U https://proceedings.mlr.press/v139/liang21a.html %V 139 %X As machine learning methods are deployed in real-world settings such as healthcare, legal systems, and social science, it is crucial to recognize how they shape social biases and stereotypes in these sensitive decision-making processes. Among such real-world deployments are large-scale pretrained language models (LMs) that can be potentially dangerous in manifesting undesirable representational biases - harmful biases resulting from stereotyping that propagate negative generalizations involving gender, race, religion, and other social constructs. As a step towards improving the fairness of LMs, we carefully define several sources of representational biases before proposing new benchmarks and metrics to measure them. With these tools, we propose steps towards mitigating social biases during text generation. Our empirical results and human evaluation demonstrate effectiveness in mitigating bias while retaining crucial contextual information for high-fidelity text generation, thereby pushing forward the performance-fairness Pareto frontier.
APA
Liang, P.P., Wu, C., Morency, L. & Salakhutdinov, R.. (2021). Towards Understanding and Mitigating Social Biases in Language Models. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:6565-6576 Available from https://proceedings.mlr.press/v139/liang21a.html.

Related Material