Exploring Gender Bias in Large Language Models: An In-depth Dive into the German Language

TL;DR

This paper introduces five German datasets to evaluate gender bias in large language models, revealing language-specific challenges and emphasizing the need for tailored bias measurement methods across languages.

Contribution

It provides new German datasets for gender bias evaluation in LLMs and highlights language-specific issues not addressed by existing English-focused methods.

Findings

German gender bias evaluation faces unique challenges.

Ambiguous male occupational terms influence bias measurements.

Neutral nouns can affect gender perception in German.

Abstract

In recent years, various methods have been proposed to evaluate gender bias in large language models (LLMs). A key challenge lies in the transferability of bias measurement methods initially developed for the English language when applied to other languages. This work aims to contribute to this research strand by presenting five German datasets for gender bias evaluation in LLMs. The datasets are grounded in well-established concepts of gender bias and are accessible through multiple methodologies. Our findings, reported for eight multilingual LLM models, reveal unique challenges associated with gender bias in German, including the ambiguous interpretation of male occupational terms and the influence of seemingly neutral nouns on gender perception. This work contributes to the understanding of gender bias in LLMs across languages and underscores the necessity for tailored evaluation frameworks.