Probing Covert Racism in Language Models Through the Lens of Dialect Prejudice

Join a thought-provoking seminar exploring how language models perpetuate covert racism through dialect prejudice, particularly focusing on biases against African American English speakers. Delve into groundbreaking research presented by Dr. Valentin Hofmann from the Allen Institute for AI that reveals how AI language models exhibit Raciolinguistic stereotypes more severe than documented human prejudices. Examine the relationship between overt and covert stereotypes in language models, understand their potential harmful impacts, and evaluate the effectiveness of existing bias mitigation techniques like human feedback training. Learn how dialect prejudice influences language models' reasoning capabilities and explore the broader implications for NLP, linguistics, and computational social science. Gain insights from Dr. Hofmann's extensive research experience at prestigious institutions including the University of Oxford, Stanford University, and DeepMind.