PeopleofSuperintelligence

Researcher

Samuel Gershman

Also known as Sam Gershman, Samuel J. Gershman

Samuel Gershman is a Harvard professor whose work connects cognitive science with reinforcement-learning agents and artificial intelligence. He earned a BA in Neuroscience and Behavior at Columbia, completed a Ph.D. in Psychology and Neuroscience at Princeton, completed a postdoctoral fellowship at MIT, and joined Harvard in 2015. His work ranges from a Bayesian and reinforcement-learning account of associative learning to EMPA’s fast learning across Atari-style games and PlayTrain’s LLM-generated JavaScript environments. He won a 2024 Schmidt Sciences Polymath award, and in 2025 his lab worked on a computer model of Stentor, a single-celled organism.

Timeline

  1. 1990s

    An early interest in computers

    Growing up in Chicago in the 1990s, he is intrigued by the idea that computers could think, which prompts questions about how humans think.

  2. 2007–2013

    Degrees at Columbia and Princeton

    He earns a BA in Neuroscience and Behavior from Columbia University in 2007 and a Ph.D. in Psychology and Neuroscience from Princeton University in 2013, where he works with Ken Norman and Yael Niv.

  3. 2013–2015

    Postdoctoral fellowship at MIT

    From 2013 to 2015, he is a postdoctoral fellow in MIT’s Department of Brain and Cognitive Sciences, working with Josh Tenenbaum and Nancy Kanwisher.

  4. 2015

    Associative learning at Harvard

    He joins Harvard in 2015 and publishes “A Unifying Probabilistic View of Associative Learning.” The paper combines Bayesian and reinforcement-learning theories of associative learning in one framework.

  5. 2021

    EMPA and learning across tasks

    In 2021, he coauthors “Human-Level Reinforcement Learning through Theory-Based Modeling, Exploration, and Planning,” which introduces EMPA, an agent that learns models of video games and plans through simulation. EMPA matches human learning efficiency on 90 Atari-style games, learning new games in minutes. A separate study finds evidence that people apply previously learned policies to novel scenarios.

  6. 2024

    A Schmidt Sciences Polymath award

    In 2024, he wins one of six Schmidt Sciences Polymath awards.

  7. 2025

    Modeling memory in Stentor

    In 2025, his lab works to develop a computer model that can think like Stentor, a single-celled organism. The project asks how cells learn and remember without neurons or synapses.

  8. 2026

    The PlayTrain framework

    In 2026, he coauthors PlayTrain, an RL framework that uses LLMs to generate JavaScript games from a minimal human prompt and run the games in a standard 'gym' environment. The framework lets researchers train reinforcement-learning agents on those games.

  9. As of October 2026

    Harvard professor and Kempner faculty

    As of October 2026, he is a Professor of Psychology and Brain Science at Harvard and Kempner Associate Faculty. He also serves on the Kempner Faculty Steering Committee.

Career

  1. NowHarvard UniversityAssociate Faculty, Professor of Psychology and Brain Science
  2. 2013–2015Massachusetts Institute of TechnologyPostdoctoral fellow

Education

  1. 2013Princeton UniversityPhD in Psychology and Neuroscience
  2. 2007Columbia UniversityB.A. in Neuroscience and Behavior

From the biography.

Checked career sources: news.harvard.edu, journals.plos.org, kempnerinstitute.harvard.edu, news.harvard.edu.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Samuel Gershman yet. Send us a free-licensed photo.

  1. How Do Single-Celled Organisms Learn and Remember | Harvard Magazine harvardmagazine.com. Checked 27 October 2026.
  2. Samuel Gershman - Kempner Institute kempnerinstitute.harvard.edu. Checked 27 October 2026.
  3. Psychology Professor Samuel J. Gershman wins Polymath award — Harvard Gazette news.harvard.edu. Checked 27 October 2026.
  4. A Unifying Probabilistic View of Associative Learning | PLOS Computational Biology journals.plos.org. Checked 27 October 2026.
  5. Human-Level Reinforcement Learning through Theory-Based Modeling, Exploration, and Planning | arXiv arxiv.org. Checked 27 October 2026.
  6. Multi-task reinforcement learning in humans | Nature Human Behaviour nature.com. Checked 27 October 2026.
  7. PlayTrain: An Efficient Reinforcement Learning Framework for LLM-Generated Adaptable JavaScript Games | arXiv arxiv.org. Checked 27 October 2026.
  8. On cellular complexity and the future of biological intelligence: Q&A with Sam Gershman — Harvard Gazette news.harvard.edu. Checked 27 October 2026.