Investigating Subtler Biases in LLMs: Ageism, Beauty, Institutional, and Nationality Bias in Generative Models

TL;DR

This paper examines subtle biases related to age, beauty, and nationality in large language models, introducing a new dataset and benchmark to measure these less-studied but impactful social biases.

Contribution

It presents a novel template-based dataset and benchmark for evaluating subtle social biases in LLMs, focusing on underexplored dimensions like age and beauty.

Findings

LLMs show correlations between social groups and positive/negative attributes.

The dataset reveals biases similar to human stereotypes like 'beauty is good'.

Benchmark can track progress in reducing subtle social biases.

Abstract

LLMs are increasingly powerful and widely used to assist users in a variety of tasks. This use risks the introduction of LLM biases to consequential decisions such as job hiring, human performance evaluation, and criminal sentencing. Bias in NLP systems along the lines of gender and ethnicity has been widely studied, especially for specific stereotypes (e.g., Asians are good at math). In this paper, we investigate bias along less-studied but still consequential, dimensions, such as age and beauty, measuring subtler correlated decisions that LLMs make between social groups and unrelated positive and negative attributes. We ask whether LLMs hold wide-reaching biases of positive or negative sentiment for specific social groups similar to the "what is beautiful is good" bias found in people in experimental psychology. We introduce a template-generated dataset of sentence completion tasks that asks the model to select the most appropriate attribute to complete an evaluative statement about a person described as a member of a specific social group. We also reverse the completion task to select the social group based on an attribute. We report the correlations that we find for 4 cutting-edge LLMs. This dataset can be used as a benchmark to evaluate progress in more generalized biases and the templating technique can be used to expand the benchmark with minimal additional human annotation.