Whose Opinions Do Language Models Reflect?

Shibani Santurkar, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, Tatsunori Hashimoto
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:29971-30004, 2023.

Abstract

Language models (LMs) are increasingly being used in open-ended contexts, where the opinions they reflect in response to subjective queries can have a profound impact, both on user satisfaction, and shaping the views of society at large. We put forth a quantitative framework to investigate the opinions reflected by LMs – by leveraging high-quality public opinion polls. Using this framework, we create OpinionQA, a dataset for evaluating the alignment of LM opinions with those of 60 US demographic groups over topics ranging from abortion to automation. Across topics, we find substantial misalignment between the views reflected by current LMs and those of US demographic groups: on par with the Democrat-Republican divide on climate change. Notably, this misalignment persists even after explicitly steering the LMs towards particular groups. Our analysis not only confirms prior observations about the left-leaning tendencies of some human feedback-tuned LMs, but also surfaces groups whose opinions are poorly reflected by current LMs (e.g., 65+ and widowed individuals).

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-santurkar23a, title = {Whose Opinions Do Language Models Reflect?}, author = {Santurkar, Shibani and Durmus, Esin and Ladhak, Faisal and Lee, Cinoo and Liang, Percy and Hashimoto, Tatsunori}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {29971--30004}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/santurkar23a/santurkar23a.pdf}, url = {https://proceedings.mlr.press/v202/santurkar23a.html}, abstract = {Language models (LMs) are increasingly being used in open-ended contexts, where the opinions they reflect in response to subjective queries can have a profound impact, both on user satisfaction, and shaping the views of society at large. We put forth a quantitative framework to investigate the opinions reflected by LMs – by leveraging high-quality public opinion polls. Using this framework, we create OpinionQA, a dataset for evaluating the alignment of LM opinions with those of 60 US demographic groups over topics ranging from abortion to automation. Across topics, we find substantial misalignment between the views reflected by current LMs and those of US demographic groups: on par with the Democrat-Republican divide on climate change. Notably, this misalignment persists even after explicitly steering the LMs towards particular groups. Our analysis not only confirms prior observations about the left-leaning tendencies of some human feedback-tuned LMs, but also surfaces groups whose opinions are poorly reflected by current LMs (e.g., 65+ and widowed individuals).} }
Endnote
%0 Conference Paper %T Whose Opinions Do Language Models Reflect? %A Shibani Santurkar %A Esin Durmus %A Faisal Ladhak %A Cinoo Lee %A Percy Liang %A Tatsunori Hashimoto %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-santurkar23a %I PMLR %P 29971--30004 %U https://proceedings.mlr.press/v202/santurkar23a.html %V 202 %X Language models (LMs) are increasingly being used in open-ended contexts, where the opinions they reflect in response to subjective queries can have a profound impact, both on user satisfaction, and shaping the views of society at large. We put forth a quantitative framework to investigate the opinions reflected by LMs – by leveraging high-quality public opinion polls. Using this framework, we create OpinionQA, a dataset for evaluating the alignment of LM opinions with those of 60 US demographic groups over topics ranging from abortion to automation. Across topics, we find substantial misalignment between the views reflected by current LMs and those of US demographic groups: on par with the Democrat-Republican divide on climate change. Notably, this misalignment persists even after explicitly steering the LMs towards particular groups. Our analysis not only confirms prior observations about the left-leaning tendencies of some human feedback-tuned LMs, but also surfaces groups whose opinions are poorly reflected by current LMs (e.g., 65+ and widowed individuals).
APA
Santurkar, S., Durmus, E., Ladhak, F., Lee, C., Liang, P. & Hashimoto, T.. (2023). Whose Opinions Do Language Models Reflect?. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:29971-30004 Available from https://proceedings.mlr.press/v202/santurkar23a.html.

Related Material