Junzhe Zhang

Assistant Professor 

Electrical Engineering and Computer Science


Areas of Expertise 

  • Causal Inference
  • Reinforcement Learning
  • Fairness Analysis
  • Generative Models 

My research focuses on causal inference theory and its applications to artificial intelligence, reinforcement learning, and machine learning. I am particularly interested in understanding the principles of robust decision-making in the context of distribution shifts, including challenges of confounding bias, selection bias, and external validity, and using that understanding to develop more efficient, robust, and fair decision-making systems (i.e., agents). Here are a few questions I’ve been exploring: 

  • How do we evaluate the causal effects of interventions from biased data? How do we draw robust counterfactual claims in the presence of distribution shifts? 
  • How can robust off-policy learning methods be developed to evaluate candidate policies from biased data? How can we extrapolate informative knowledge from offline data to accelerate a future online learning process? 
  • How can ethical concepts such as fairness and discrimination be incorporated into AI decision systems? How can we audit these systems to detect potential discriminatory behaviors? 

Selected Publications:  

Partial Counterfactual Identification from Observational and Experimental Data 
Junzhe Zhang, Jin Tian, Elias Bareinboim. 
ICML-2022. In Proceedings of the 39th International Conference on Machine Learning. 

Causal Imitation Learning with Unobserved Confounders 
Junzhe Zhang, Daniel Kumor, Elias Bareinboim. 
NeurIPS-2020. In Proceedings of the 34th Annual Conference on Neural Information Processing Systems. 

Near-Optimal Reinforcement Learning in Dynamic Treatment Regimes 
Junzhe Zhang, Elias Bareinboim. 
NeurIPS-2019. In Proceedings of the 33rd Annual Conference on Neural Information Processing Systems. 

Fairness in Decision-Making — The Causal Explanation Formula 
Junzhe Zhang, Elias Bareinboim. 
AAAI-2018. In Proceedings of the 32nd AAAI Conference on Artificial Intelligence. 

Transfer Learning in Multi-Armed Bandits: A Causal Approach 
Junzhe Zhang, Elias Bareinboim. 
IJCAI-2017. In Proceedings of the 26th International Joint Conference on Artificial Intelligence.