Abstract

The recent advancements in generative AI, largely driven by training larger models with more data, unlock exciting new opportunities. However, these scaling-centric approaches overlook the human aspect of language, where meaning goes beyond just words and involves people and social context. This limitation leads to models that generate generic responses, potentially biased towards dominant social groups. In this talk, I present my research on developing socially-informed, human-centered AI. I demonstrate how incorporating social context, such as community and speaker information, can improve AI models’ adaptability, effectiveness, and fairness across all stages of development, from data to model and evaluation. I conclude by outlining my research vision for building AI models that foster a safer, inclusive, and personalized human experience.

Bio

Chan Young Park is a final-year PhD candidate in the School of Computer Science at Carnegie Mellon University, advised by Professor Yulia Tsvetkov. She is currently a visiting PhD student at the University of Washington. Her research focuses on the intersection of NLP, computational social science, and AI ethics. Her work has been published in top conferences and journals, including PNAS, ICLR, ACL, EMNLP, and WWW, and was featured in MIT Tech Review and the Washington Post. She received the ACL Best Paper Award and Wikimedia Foundation Research Award of the Year 2023 and was selected as a University of Chicago Rising Stars in Data Science. Chan Young is also a recipient of the K&L Gates Presidential Fellowship and Korea Foundation for Advanced Studies PhD Fellowship.

Host

Chris Brinton, cgb@purdue.edu