Tianmin Shu

Tianmin Shu

I am a postdoc at MIT advised by Josh Tenenbaum and Antonio Torralba. My research goal is to engineer and reverse engineer social intelligence. I develop (1) cognitively inspired machine learning and AI methodologies for building socially inteligent systems, and (2) principled computational models that help uncover the cognitive mechanisms underlying human social intelligence. I received my Ph.D. in Statistics from University of California, Los Angeles under the supervision of Song-Chun Zhu, and have interned at Facebook AI Research and Salesforce Research.

Email: tshu [at] mit.edu

Google Scholar / Thesis


03/2023: Invited talk in the Social Cognitive Neuroscience Lab at the University of Iowa.

12/2022: Invited talk in the Computational Cognition, Vision, and Learning Group at Johns Hopkins University.

10/2022: Invited talk at Robotics Seminar, University of New Hampshire.

10/2022: Co-organized ECCV Workshop on Machine Visual Common Sense.

07/2022: Invited talk at Columbia University, Johns Hopkins University, and University of Maryland.

07/2022: Co-organized RSS Workshop on Social Intelligence in Humans and Robots.

01/2022: Invited talk in VITA Lab at EPFL.

11/2021: Our work on machine social common sense was covered by ScienceNews for Students and VentureBeat; MIT News also covered our work on social modeling.

10/2021: Invited talk at AI Seminar, USC ISI.

06/2021: Co-organized ICRA Workshop on Social Intelligence in Humans and Robots.

05/2021: Co-organized ICLR 2021 Social "Social AI Virtual Gathering."

02/2021: Invited talk at Sony CSL, Paris.

11/2020: Invited talk at Virutal Computational Neuroscience (VCN) Journal Club hosted by Stanford, MIT/Harvard, and Princeton.

Reseach Highlights

Social Scene Understanding

Multi-agent Cooperation

Imitation Learning

Intuitive Psychology