Whose Opinions Do Language Models Reflect?

by   Shibani Santurkar, et al.

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 https://github.com/tatsu-lab/opinions_qa.


page 8

page 32

page 33

page 34

page 35

page 36

page 40


Aligning Language Models to User Opinions

An important aspect of developing LLMs that interact with humans is to a...

Towards Measuring the Representation of Subjective Global Opinions in Language Models

Large language models (LLMs) may not equitably represent diverse global ...

Using Open-Ended Stressor Responses to Predict Depressive Symptoms across Demographics

Stressors are related to depression, but this relationship is complex. W...

When the Majority is Wrong: Leveraging Annotator Disagreement for Subjective Tasks

Though majority vote among annotators is typically used for ground truth...

Evidence of Demographic rather than Ideological Segregation in News Discussion on Reddit

We evaluate homophily and heterophily among ideological and demographic ...

AI-Augmented Surveys: Leveraging Large Language Models for Opinion Prediction in Nationally Representative Surveys

How can we use large language models (LLMs) to augment surveys? This pap...

Please sign up or login with your details

Forgot password? Click here to reset