University of Toronto & University of Michigan

The University of Toronto and University of Michigan entry represents a formal cross-institutional AI safety research collaboration between two prominent NLP research groups: Prof. Zhijing Jin's Jinesis AI Lab at the University of Toronto and Prof. Rada Mihalcea's Language and Information Technologies (LIT) Lab at the University of Michigan. The collaboration received grant funding through the Survival and Flourishing Fund's 2025 S-Process for general support of their joint AI safety research program. Zhijing Jin is an Assistant Professor at the University of Toronto (since July 2025), a Canada CIFAR AI Chair at the Vector Institute, founder of EuroSafeAI, a faculty affiliate at UC Berkeley's CHAI, and a member of the Schwartz Reisman Institute. She received her PhD from the Max Planck Institute for Intelligent Systems and ETH Zurich in 2024, and was a 2022 Future of Life Institute Technical PhD Fellow. Her research focuses on causal inference for NLP, multi-agent LLM safety, and AI for causal science. The Jinesis AI Lab, founded in 2025, operates three teams across Toronto (UofT and Vector Institute), Europe (MPI, ETH, and EuroSafeAI), and a global team of international researchers. Rada Mihalcea is the Janice M. Jenkins Collegiate Professor of Computer Science and Engineering at the University of Michigan, Director of the Michigan Artificial Intelligence Lab since 2017, and leader of the LIT Lab. She is an ACM Fellow, AAAI Fellow, ACL Fellow, and former ACL President (2021). Her research spans computational linguistics, NLP, and computational social science, with particular focus on cross-cultural LLMs, safe and trustworthy AI, and AI for social good. She has authored over 500 publications. The collaboration's AI safety work centers on multi-agent LLM simulations including GovSim, SanctSim, and MoralSim. Their NeurIPS 2024 paper "Cooperate or Collapse: Emergence of Sustainability in a Society of LLM Agents" introduced GovSim, a simulation platform that studies how AI agents collectively manage shared resources, revealing that stronger reasoning capabilities can paradoxically make models more prone to selfish strategies. They also collaborate on causal reasoning in LLMs, with their ICLR 2024 paper "Can Large Language Models Infer Causation from Correlation?" examining fundamental reasoning capabilities. The researchers won two Best Paper Awards at NeurIPS 2024 Workshops for work on multilingual alignment and gender bias in LLMs.