Evaluating Biases in Context-Dependent Health Questions

TL;DR

This paper investigates biases in large language models when answering context-dependent healthcare questions, revealing a bias favoring young adult females through curated demographic-based questions.

Contribution

It introduces a dataset of healthcare questions with demographic context and analyzes model biases based on age, sex, and location attributes.

Findings

Models exhibit biases favoring young adult females.

Biases are evident across age, sex, and location attributes.

Contextual information influences model responses.

Abstract

Chat-based large language models have the opportunity to empower individuals lacking high-quality healthcare access to receive personalized information across a variety of topics. However, users may ask underspecified questions that require additional context for a model to correctly answer. We study how large language model biases are exhibited through these contextual questions in the healthcare domain. To accomplish this, we curate a dataset of sexual and reproductive healthcare questions that are dependent on age, sex, and location attributes. We compare models' outputs with and without demographic context to determine group alignment among our contextual questions. Our experiments reveal biases in each of these attributes, where young adult female users are favored.