Avatar

Kevin A Smith

Research Scientist in Brain and Cognitive Sciences

Massachusetts Institute of Technology

Biography

I am a Research Scientist in the Computational Cognitive Science laboratory with Joshua Tenenbaum at MIT.

I am interested in how people use physical reasoning for a variety of common-sense tasks such as prediction, inferences about object properties, or action planning. To support these capacities, we all have the ability to simulate how our environment will unfold based on the physics of the world. By using a combination of psychophysics and computational modeling, I study how we are able to perform this simulation, how we select between simulations and other forms of physical knowledge, and how we can draw on these predictions to make inferences about the world and plan our actions.

Interests

  • Physical Reasoning
  • Computational Cognitive Science

Education

  • PhD in Psychology, 2015

    UC San Diego

  • BA in Cognitive Science (Minor in Computer Science), 2005

    Dartmouth College

Research Topics

The Nature of Physical Representations and Simulations

How the mind efficiently represents the world for physical reasoning.

Integrating Simulation and Symbolic Strateiges

Choosing cognitive systems for thinking about physics.

Integrating Prediction and Action

Using physical reasoning to act on the world.

Developing Physical Common Sense

Formalizing how we learn about the physical world.

Physical Reasoning in Artificial Agents

Building physical reasoning for AI.

Talks

Some of my recent recorded talks include:

Teaching

I am not currently teaching or TA’ing any courses, but materials from prior courses can be found here, including:

  • A lecture on Introduction to Optimization from the 2016 Center for Brains, Minds, and Machines summer school.
  • A lecture on Probabilistic Programming from the 2018 Center for Brains, Minds, and Machines summer school.
  • Lecture notes from a Graduate Statistics course I TA’d with Ed Vul at UCSD. These were meant to supplement the theoretical lessons by introducing students to the R statistical programming language and how to implement statistical techniques using R.

Contact

  • k2smith@mit.edu
  • 77 Massachusetts Avenue, Cambridge, MA, 02139, United States
  • MIT Building 46-4053