Are Social Sentiments Inherent in LLMs? An Empirical Study on Extraction of Inter-demographic Sentiments

TL;DR

This empirical study investigates whether large language models inherently encode social group sentiments by comparing LLM responses with social survey data across nationality, religion, and ethnicity.

Contribution

The paper provides an empirical evaluation of LLMs' ability to capture inter-group sentiments, highlighting their alignment with real-world social survey results.

Findings

LLMs show higher correlation with survey data for larger social groups.

Responses from LLMs reflect inter-group sentiments consistent with social surveys.

The study validates the potential of LLMs to mirror social biases and sentiments.

Abstract

Large language models (LLMs) are supposed to acquire unconscious human knowledge and feelings, such as social common sense and biases, by training models from large amounts of text. However, it is not clear how much the sentiments of specific social groups can be captured in various LLMs. In this study, we focus on social groups defined in terms of nationality, religion, and race/ethnicity, and validate the extent to which sentiments between social groups can be captured in and extracted from LLMs. Specifically, we input questions regarding sentiments from one group to another into LLMs, apply sentiment analysis to the responses, and compare the results with social surveys. The validation results using five representative LLMs showed higher correlations with relatively small p-values for nationalities and religions, whose number of data points were relatively large. This result indicates that the LLM responses including the inter-group sentiments align well with actual social survey results.