Gender Bias
Gender bias in artificial intelligence (AI) models, particularly large language models (LLMs) and machine learning systems, is a significant area of concern, focusing on identifying and mitigating the perpetuation of societal stereotypes. Current research investigates bias across various modalities, including text generation, machine translation, image generation, and speech processing, employing techniques like adversarial training, counterfactual analysis, and prompt engineering to reduce bias in model outputs. Understanding and addressing this bias is crucial for ensuring fairness, equity, and trustworthiness in AI applications across diverse sectors, from healthcare and finance to education and employment.
Papers
Gender-Neutral Large Language Models for Medical Applications: Reducing Bias in PubMed Abstracts
Elizabeth Schaefer, Kirk Roberts
Addressing speaker gender bias in large scale speech translation systems
Shubham Bansal, Vikas Joshi, Harveen Chadha, Rupeshkumar Mehta, Jinyu Li
LLMs Reproduce Stereotypes of Sexual and Gender Minorities
Ruby Ostrow, Adam Lopez