Avatar

Thomas Alexandre Langlois

Postdoctoral Associate

Massachusetts Institute of Technology (MIT) & NYU

UT Austin

Princeton University

University of California, Berkeley

Biography

How do human minds resolve the difficult computational trade-offs involved in generating flexible and accurate inferences about the world while remaining so efficient? My research activities center on identifying the efficient coding principles that govern how biological minds form accurate yet efficient perceptual and semantic representations using information theory. I am also interested in developing tools to examine how minds deviate from leading AI systems with respect to resolving the same computational trade-offs. Finally, I develop behavioral experiments based on statistical sampling techniques to uncover the subjective priors that enable human minds to solve complex cognitive and perceptual tasks.

For a representative example of my work probing visual memory priors, see my work in PNAS. For another example of my work exploring the neural (biological) basis of Bayesian inference during visual decision making, see my more recent work in PNAS. I also investigate how biological vision differs from computer vision systems. For a representative example of this work, see my oral presentation and paper in NeurIPS 2021. I am currently exploring the information-theoretic tradeoffs that govern how humans communicate about visual percepts using the Information Bottleneck (IB) Principle.

I completed my Ph.D. in Thomas Griffiths’ Computational Cognitive Science Lab at UC Berkeley in August of 2018. Prior to completing my Ph.D., I completed an M.S. in Computer Science (EECS), also at UC Berkeley. I then joined Princeton University as a postdoctoral researcher, where I developed experimental methods to estimate subjective probability distributions in visual memory. Next, I completed a postdoctoral fellowship at UT Austin in the Center for Perceptual Systems (CPS), where I worked in Dr. Robbe Goris’ group. I am currently a postdoctoral associate in the Brain and Cognitive Sciences Department (BCS) at MIT, where I am working with Dr. Roger Levy and Dr. Noga Zaslavsky on applications of the Information Bottleneck (IB) Principle. I am also a visiting scholar affiliated with the Department of Psychology at NYU.

Interests

  • Bayesian models of cognition
  • Information theory
  • Perception and language
  • Artificial Intelligence
  • Visual neuroscience

Education

  • PhD in Psychology (Cognition), 2018

    University of California, Berkeley

  • MS in Computer Science (EECS), 2018

    University of California, Berkeley

  • BA in Psychology (Cognitive Science), BA in Art History & Studio Art, 2008

    Georgetown University

Recent Publications

(2025). Bayesian Inference by Visuomotor Neurons in Prefrontal Cortex. In Proceedings of the National Academy of Sciences (PNAS) 122(13).

Preprint PDF Code

(2021). Passive Attention in Artifical Neural Networks Predicts Human Visual Selectivity. In Advances in Neural Information Processing Systems (NeurIPS), 35. Accepted (Oral), ArXiv preprint: 2107.07013.

PDF

(2017). Uncovering visual priors in spatial memory using serial reproduction. In Proceedings of the 39th Annual Conference of the Cognitive Science Society.

PDF