Harmful Speech Detection by Language Models - Gender-Queer Dialect Bias

Watch a 19-minute ACM conference presentation examining how language models demonstrate bias in detecting harmful speech when analyzing gender-queer dialects. Explore research findings from Rebecca Dorn, Lee Kezar, Fred Morstatter, and Kristina Lerman that reveal systematic biases in how AI systems process and flag potentially harmful content across different gender expressions and linguistic patterns.