Societal AI banner

Societal AI

Research Project: Value Compass

Abstract:

Large Language Models (LLMs) have achieved remarkable breakthroughs, yet their growing integration into everyday life may pose societal risks, especially when they generate unethical content. Incorporating diverse human values and ethical principles into these powerful generative models is therefore critical. This can enhance AI safety, ensures responsible development, respects the diversity of cultural and individual values, and helps prevent bias. This project adopts an interdisciplinary approach, integrating AI research with philosophical, psychological, and social science perspectives on values and ethics. We focus on three core challenges in value alignment: (1) Clarity: defining alignment goals that are unambiguous, precise, and reflective of comprehensive human values; (2) Adaptability: designing algorithms to cover diverse contexts, evolving model capabilities, and changing societal norms beyond basic safety concerns; and (3) Transparency: developing interpretable alignment framework. Ultimately, this research aims to pave the way for a harmonious future where humans and machines can coexist and collaborate productively.

Team:

  • Xing Xie, Partner Research Manager, Microsoft Research Asia
  • Xiaoyuan Yi, Senior Researcher, Microsoft Research Asia
  • Jing Yao, Researcher, Microsoft Research Asia
  • Beibei Shi, Senior Research PM, Microsoft Research Asia
  • Scarlett Li, Principal Research PM Manager, Microsoft Research Asia
  • Yang Ou, Senior Designer, Microsoft Research Asia

Collaborators:

  • Xiting Wang, Assistant Professor, Renmin University of China
  • Peng Zhang, Associate Professor, School of Computer Science, Fudan University
  • Linus Huang, Assistant Professor, Division of Humanities, Hong Kong University of Science and Technology

Interns and Developers:

  • Yifan Gong, Undergraduate Student, Algorithm Research, Hunan University, College of Computer Science and Electronic Engineering (Internship: 2023.10-2024.05)
  • Shitong Duan, Master Student, Algorithm Research, Fudan University, School of Computer Science (Internship: 2023.04-2023.10)
  • Xingqi Wang, PhD Student, Algorithm Research, Tsinghua University, Department of Computer Science and Technology (Internship: 2023.02-2023.09)
  • Yan Liu, PhD Student, Algorithm Research, The University of Edinburgh, School of Informatics (Internship: 2023.11-2024.04)
  • Yuhan Zeng, Undergraduate Student, Project Management, Zhejiang University, Chu Kochen Honors College (Internship: 2024.02-2024.07)
  • Tiantian Xue, Developer, Tech Development, Microsoft Research Asia
  • Geli Guo, UI & UX Designer, Tech Development, Microsoft Research Asia

Representative Publications:

Open-source contributions:

Other achievements: