Russel Sage Foundation

Over the past few years, discriminatory speech on social media has grown significantly. Discriminatory speech refers to speech that targets people on the basis of traits that shape their identity and tie them to a category that has been persistently discriminated against, oppressed, or exploited. The principal investtigators aim to bridge the gaps in prior literature by analyzing the language of implicit discriminatory discourse, with the goal of developing insights and models to understand implicit discriminatory speech at scale. This project will combine PI Yang’s expertise in developing novel natural language processing algorithms to model semantics of the human language, such as bias, social support exchange, persuasion, and racism, and Co-PI ElSherif’s expertise in building novel datasets and algorithms for analyzing online hate speech. The researchers focus on two questions: (1) what is implicit discriminatory speech and (2) how can we design generic and scalable data-driven techniques to detect it.