| |
 Toward Human-Aligned Intelligence:
Modeling Behavior, Expression, and Connections Speaker (s):
Hanjia LYU
PhD student
Department of Computer Science
University of Rochester
| |
About the Talk When machine learning models are deployed in settings where outcomes hinge on human actions, interpretations, and interactions, achieving strong performance requires models that can reason about human behavior, communication, and social context. Building such human-aligned intelligence calls for systems that comprehend the richness of human expression and the structure of human interactions. In this talk, I will present my research on developing human-aligned systems that reason in human-like ways by augmenting and structuring data representations. These representations capture how people express themselves through language, visuals, and behavior, as well as how people and information are embedded within social and semantic networks. I will highlight key challenges inherent in real-world human behavioral data and introduce computational methods designed to address them, enabling models to more reliably align with human intent and behavior. I will also demonstrate how model behavior shifts under different data conditions. Finally, I will discuss opportunities for advancing human-aligned systems that are not only accurate, but also trustworthy and supportive of human agency. About the Speaker Hanjia Lyu is a Ph.D. student in Computer Science at the University of Rochester, advised by Professor Jiebo Luo. His research sits at the intersection of artificial intelligence and human behavior, with a focus on building human-aligned systems that understand, reason, and act in ways consistent with human decision-making. He is a recipient of the 2024 Google PhD Fellowship. He has also collaborated with industry research labs, contributing to large-scale projects at the interface of foundation models and human behavior, with real-world applications in recommender systems and social networks. |
