Societal Bias
Societal biases, reflecting prejudices present in training data, are a significant concern in large language models (LLMs) and other AI systems, particularly impacting fairness and equity in applications ranging from recruitment to healthcare. Current research focuses on detecting and mitigating these biases, employing techniques like adversarial training, data augmentation, and prompt engineering across various model architectures including BERT and LLMs, with a growing emphasis on multilingual and culturally-sensitive datasets. This work is crucial for ensuring responsible AI development and deployment, preventing the amplification of harmful stereotypes and promoting more equitable outcomes across diverse populations.
Papers
Evaluating Machine Perception of Indigeneity: An Analysis of ChatGPT's Perceptions of Indigenous Roles in Diverse Scenarios
Cecilia Delgado Solorzano, Carlos Toxtli Hernandez
"Im not Racist but...": Discovering Bias in the Internal Knowledge of Large Language Models
Abel Salinas, Louis Penafiel, Robert McCormack, Fred Morstatter