Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Yuen, Raghuram, Vethavikashini Chithrra, Mattern, Justus, Mihalcea, Rada, Jin, Zhijing
Formato: Preprint
Publicado: 2022
Materias:
Acceso en línea:https://arxiv.org/abs/2212.10678
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911073255292928
author Chen, Yuen
Raghuram, Vethavikashini Chithrra
Mattern, Justus
Mihalcea, Rada
Jin, Zhijing
author_facet Chen, Yuen
Raghuram, Vethavikashini Chithrra
Mattern, Justus
Mihalcea, Rada
Jin, Zhijing
contents Generated texts from large language models (LLMs) have been shown to exhibit a variety of harmful, human-like biases against various demographics. These findings motivate research efforts aiming to understand and measure such effects. This paper introduces a causal formulation for bias measurement in generative language models. Based on this theoretical foundation, we outline a list of desiderata for designing robust bias benchmarks. We then propose a benchmark called OccuGender, with a bias-measuring procedure to investigate occupational gender bias. We test several state-of-the-art open-source LLMs on OccuGender, including Llama, Mistral, and their instruction-tuned versions. The results show that these models exhibit substantial occupational gender bias. Lastly, we discuss prompting strategies for bias mitigation and an extension of our causal formulation to illustrate the generalizability of our framework. Our code and data https://github.com/chenyuen0103/gender-bias.
format Preprint
id arxiv_https___arxiv_org_abs_2212_10678
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Causally Testing Gender Bias in LLMs: A Case Study on Occupational Bias
Chen, Yuen
Raghuram, Vethavikashini Chithrra
Mattern, Justus
Mihalcea, Rada
Jin, Zhijing
Computation and Language
Machine Learning
Generated texts from large language models (LLMs) have been shown to exhibit a variety of harmful, human-like biases against various demographics. These findings motivate research efforts aiming to understand and measure such effects. This paper introduces a causal formulation for bias measurement in generative language models. Based on this theoretical foundation, we outline a list of desiderata for designing robust bias benchmarks. We then propose a benchmark called OccuGender, with a bias-measuring procedure to investigate occupational gender bias. We test several state-of-the-art open-source LLMs on OccuGender, including Llama, Mistral, and their instruction-tuned versions. The results show that these models exhibit substantial occupational gender bias. Lastly, we discuss prompting strategies for bias mitigation and an extension of our causal formulation to illustrate the generalizability of our framework. Our code and data https://github.com/chenyuen0103/gender-bias.
title Causally Testing Gender Bias in LLMs: A Case Study on Occupational Bias
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2212.10678