Gender Bias in Machine Translation and The Era of Large Language Models
Eva Vanmassenhove
TL;DR
The chapter investigates how machine translation propagates gender bias, driven by cross-linguistic gender marking, data biases, and statistical artifacts. It surveys conventional NMT and GPT-based MT literature, highlighting debiasing attempts and their limitations, and then reports a preliminary English–Italian translation study using GPT-3.5 that reveals persistent bias even under explicit prompting. The findings underscore the need for continued research, gender-inclusive data, and hybrid approaches that integrate linguistic knowledge with MT systems. The authors advocate for broad, interdisciplinary collaboration to develop fair, unbiased, and inclusive language technologies with accountable deployment practices.
Abstract
This chapter examines the role of Machine Translation in perpetuating gender bias, highlighting the challenges posed by cross-linguistic settings and statistical dependencies. A comprehensive overview of relevant existing work related to gender bias in both conventional Neural Machine Translation approaches and Generative Pretrained Transformer models employed as Machine Translation systems is provided. Through an experiment using ChatGPT (based on GPT-3.5) in an English-Italian translation context, we further assess ChatGPT's current capacity to address gender bias. The findings emphasize the ongoing need for advancements in mitigating bias in Machine Translation systems and underscore the importance of fostering fairness and inclusivity in language technologies.