Hua Shen

Date: January 26th, 2026

Time: 16:00 (CET)

Title:  Bidirectional Human–AI Alignment: From Value Gaps to Bidirectional Opinion Influence

Abstract: The integration of LLMs into society necessitates a broader understanding of alignment—one that accounts for bidirectional influence between humans and AI. This talk introduces a bidirectional human-AI alignment framework and presents a series of studies to understand and measure it. Our research operationalizes this concept through three critical lenses: 1) Value Misalignment: We introduce "ValueCompass", a method to quantify contextual value alignment across cultures, and expose systematic value-action gaps in LLMs; 2) Perceptual Manipulation: We document user experiences with LLM dark patterns that manipulate belief and behavior; and 3) Dynamic Influence: We provide empirical evidence of bidirectional opinion dynamics in conversation, where both agent and human stances co-adapt. Together, our work provides new lenses to measure alignment, exposes critical risks, and charts a path for developing truly human-centered, responsible AI systems that are truly aligned through mutual understanding and adaptation.

About the Presenter:  Hua Shen is an Assistant Professor of Computer Science at NYU Shanghai, jointly affiliated with New York University. Her work anchors in HCI and intersects with multiple AI fields, such as Natural Language Processing,  Speech Processing, Machine Learning, and Data Science. Particularly, her research focuses on bidirectional human-AI alignment, aiming to empower humans to interactively explain, evaluate, and collaborate with AI, while incorporating human feedback and values to improve AI systems. She was selected as the 2023 Rising Stars of Data Science. Her research has been recognized with multiple awards, including Best Paper at AIED '24, Best Demo at CSCW’23, Best Paper Honorable Mention Award at IUI’23, and 2023 Google Research Science Conference Scholarships. She received her Ph.D. from Penn State and was previously a postdoctoral scholar at the University of Washington and the University of Michigan. 

learn more about the presenter at: https://hua-shen.org/