Measuring Stereotype and Deviation Biases in Large Language Models

TL;DR

This paper investigates stereotype and deviation biases in large language models by analyzing their associations of demographic traits and comparing generated content with real-world distributions.

Contribution

It introduces a method to measure two types of biases in LLMs and provides empirical evidence of significant biases across multiple models.

Findings

All examined LLMs exhibit both stereotype and deviation biases.

Biases are present across multiple demographic groups.

Biases may lead to potential harms in LLM-generated content.

Abstract

Large language models (LLMs) are widely applied across diverse domains, raising concerns about their limitations and potential risks. In this study, we investigate two types of bias that LLMs may display: stereotype bias and deviation bias. Stereotype bias refers to when LLMs consistently associate specific traits with a particular demographic group. Deviation bias reflects the disparity between the demographic distributions extracted from LLM-generated content and real-world demographic distributions. By asking four advanced LLMs to generate profiles of individuals, we examine the associations between each demographic group and attributes such as political affiliation, religion, and sexual orientation. Our experimental results show that all examined LLMs exhibit both significant stereotype bias and deviation bias towards multiple groups. Our findings uncover the biases that occur when LLMs infer user attributes and shed light on the potential harms of LLM-generated outputs.