About the Talk:

Language models (LMs) are increasingly being used in open-ended contexts, where the opinions reflected by LMs in response to subjective queries can have a profound impact, both on user satisfaction, as well as shaping the views of society at large. In this work, we put forth a quantitative framework to investigate the opinions reflected by LMs -- by leveraging high-quality public opinion polls and their associated human responses. Using this framework, we create OpinionsQA, a new 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 demographic 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). Our code and data are available at this https URL.

Link to the full paper: Whose Opinions Do Language Models Reflect?

Full list of authors: Shibani Santurkar, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, Tatsunori Hashimoto

About the Speaker:

Shibani Santurkar is a researcher at OpenAI working in building safe and reliable machine learning models. Shibani received a PhD in Computer Science from MIT in 2021, where she was advised by Aleksander Mądry and Nir Shavit. Subsequently, she was a postdoctoral researcher at Stanford University with Tatsu Hashimoto, Percy Liang and Tengyu Ma. She is a recipient of the Google Fellowship and an Open Philanthropy early-career grant.