Liancheng (Krystal) Gong

Hi! I’m Krystal, a first-year Ph.D. student in Information Studies at the University of Maryland, College Park, where I am advised by Professor Julia Mendelsohn and a member of the CATS Lab (Computational Analysis of Text and Society). Prior to starting my Ph.D., I worked as a research assistant with Professor Daniel Hopkins at the University of Pennsylvania and as a part-time research assistant with Professor Li Zhang at Drexel University. I received my B.S. in Data Science from NYU Shanghai and M.S.E. in Data Science from the University of Pennsylvania.

My research focuses on trustworthy large language models and computational social science. I study large-scale social media data to understand how linguistic and structural features shape communication, persuasion, and information flow in digital environments, including political communication and media narratives. I am also interested in the interpretability of language models, with the goal of developing methods that make their outputs more reliable, transparent, and verifiable.

Here are some directions that I am working on or am excited about:

• Persuasion & Communication Modeling. I develop computational frameworks (e.g., persuasion indices) to quantify how linguistic and structural features shape persuasive effects in large-scale social media data.

• LLM Mechanisms & Interpretability. I study the internal mechanisms of large language models, using techniques such as probing and activation analysis to understand how models generate persuasive or strategic language, and how to make these processes more transparent and reliable.

• Structured Analysis of Social Text. I build methods to transform unstructured text into structured, analyzable representations, enabling large-scale and reproducible studies of online discourse, information flow, and human–AI interaction.

If any of this resonates, feel free to reach out. I’d love to chat, collaborate, and connect.

Personal Interests: Outside of research, I enjoy rock climbing and skiing.