Nora Chen

I recently graduated from the University of California, San Diego, majoring in Cognitive Science and specializing in Language and Culture. During my undergraduate career, I worked as a research assistant in Dr. Judith Fan's lab, studying how humans intuitively grasp physical principles and how this understanding can be modeled in artificial intelligence systems. 

Currently I reside as a Research Fellow at Harvard University in the Computation, Cognition, and Development (Ullman) Lab.

Concurrently, I have joined the Cognitive Development and Learning (Gopnik) Lab at UC Berkeley as a research intern.

Additionally, I serve as an ad hoc reviewer for the AI and Ethics Journal.

Research: I am broadly interested in exploring the cognitive underpinnings that allow people to understand and make fast, robust inferences about diverse, unknown physical phenomena. To facilitate these capabilities, I research how latent physical properties can be inferred from observed data and how humans use this information to improve predictive models of the world. Through this, I aim to shed light on the cognitive processes enabling robust inferential performance and inform the design of artificial intelligence systems that are better equipped to cope with the dynamic, uncertain nature of the physical world and more attuned to its nuanced complexities. 

Furthermore, my interests extend into visual neuroscience, particularly investigating the complex interplay between attention, perception, and physical reasoning (e.g., how predictive coding mechanisms in the visual hierarchy might be modulated by attention during physical inference tasks). Here, I'm interested in exploring how the brain might leverage attention to dynamically adjust its internal physics models, through selective amplification of prediction errors in early visual areas or biasing competition among multiple physical interpretations in higher-order regions. 

GitHub / Email / CV / Google Scholar / LinkedIn / Writing Samples

Physical Intuition and the Role of Counterfactual Reasoning: Insights from Jenga and Fluid Dynamics Simulations

[PDF]

We show that people can use counterfactual reasoning to make inferences about the physical world, even when kinetic information is degraded. Furthermore, we extend and focus on the specific physical properties that lead people to determine the viscosity of a liquid in our three fluid simulations.