HMNTL Master Calendar

Electrical and Computer Engineering Department Seminar

by Liwei Jiang

PhD Candidate, Paul G. Allen School of Computer Science and Engineering, University of Washington

Thursday, April 2, 2026, 11:00 am-12:00 pm

B02 CSL Auditorium or Online via Zoom

Title: Humanistic, Pluralistic, and Coevolutionary AI Safety and Alignment

Abstract: As AI systems scale to unprecedented global impact, alignment has shifted from a largely theoretical concern to an urgent sociotechnical imperative. How can we ensure that these systems align with human morality, values, and needs, even as humanity continues to grapple with their complexity, contradictions, and evolution?
 
 In this talk, I present my research on humanistic, pluralistic, and coevolutionary AI safety and alignment, which seeks to translate abstract human values into practical, scalable, and future-proof technical solutions in this unprecedented era of large language models (LLMs). I begin by establishing foundations for computational morality, enabling models to reason about human values and norms that are contextual, conflicting, and culturally diverse. I then demonstrate how to operationalize abstract value alignment through the development of holistic safeguards for LLMs. Finally, I argue that alignment cannot remain static. As both humans and AI systems evolve, alignment must become a dynamic, coevolutionary process. I present emerging paradigms in which humans and machines are synergistically co-evolved to enable scalable oversight and robust safety guarantees. Ultimately, my work aims to redefine alignment from a defensive goal for preventing harm into an enabling infrastructure, one that unlocks new frontiers of human–AI collective intelligence, allowing humans and AI to achieve together what neither can accomplish alone.

Liwei Jiang is a final-year Ph.D. candidate in the Paul G. Allen School of Computer Science & Engineering at the University of Washington, advised by Prof. Yejin Choi. She was previously a graduate student researcher at NVIDIA and the Allen Institute for Artificial Intelligence (Ai2). Her research focuses on humanistic, pluralistic, and coevolutionary AI safety and alignment, where she spearheads research on moral and pluralistic value reasoning in language models and develops data-, algorithm-, and system-level solutions to socio-technical challenges in AI safety, security, and large language model alignment. Her work has received Best Paper Awards at NeurIPS 2025, NAACL 2022, and CHI 2024, as well as Outstanding Paper Awards at EMNLP 2023 and the AIA Workshop at COLM 2025, and has been featured in The New York Times, Nature Outlook, IEEE Spectrum, Wired, and other major media. She co-organizes workshops including MP2 (NeurIPS 2023) and SoLaR (NeurIPS 2024; COLM 2025), and co-leads the Guardrails and Security for LLMs tutorial at ACL 2025.