Reimagining NLP Practices Towards Justice - From "Democratization" to Personal Names

This 57-minute seminar presented by Arjun Subramonian from UCLA at the USC Information Sciences Institute examines problematic practices in natural language processing (NLP) through a justice-oriented lens. Explore two critical issues: the misuse of "democratization" terminology in language technologies and the problematic association of personal names with sociodemographic characteristics. Learn how current "democratization" rhetoric in NLP can misrepresent power distribution and public control of AI, and discover recommendations for moving beyond superficial access toward truly democratic technologies. Understand the validity issues and ethical concerns inherent in linking personal names to sociodemographic attributes, including systematic errors, construct validity problems, potential harms, differential impacts, and cultural insensitivity. Gain valuable guiding questions and normative recommendations to avoid these pitfalls in NLP work. The speaker, a Computer Science PhD candidate at UCLA focusing on fairness and ethics in machine learning and NLP, brings insights as a core organizer of Queer in AI and recipient of multiple prestigious fellowships and awards.