The stereotypes and prejudice prevalent in text are liable to be absorbed and amplified by NLP models, even when they are implicit or unintentional. Developing automated methods to detect them could help mitigate this concern but requires overcoming technical challenges. Concepts like “bias” are difficult to define or collect annotated data for, and classifiers or analyses can easily capture confounding variables rather than the attributes of interest. In this talk, I will discuss the methodology we are developing to detect social biases in conversational (social media) and narrative (online articles) text. Our approach involves combining NLP models with methodology from causal inference and frameworks from social psychology in order to examine how people are addressed or described in text. In this work, we aim to move beyond classic NLP tasks like offensiveness classification or sentiment analysis and examine subtle implications in text influenced by social variables.
Anjalie is a PhD student at the Language Technologies Institute at Carnegie Mellon University, where she is advised by Yulia Tsvtekov. Her work focuses on the intersection of NLP and computational social science, including both developing NLP models that are socially aware and using NLP models to examine social issues like propaganda, stereotypes, and prejudice. She has presented her work in NLP and interdisciplinary conferences, receiving a nomination for best paper at SocInfo 2020, and she is also the recipient of a NSF graduate research fellowship and a Google PhD fellowship. Prior to graduate school, she received her undergraduate degree in computer science, with minors in Latin and ancient Greek, from Princeton University.