Jessica A. F. Thompson

Postdoctoral researcher, Human Information Processing Lab, University of Oxford
Soon: Lecturer, School of Psychology and Neuroscience, University of Glasgow (from December 2026)

Jessica-Thompson_headshot_2018.jpg

Department of Experimental Psychology

University of Oxford

Life and Mind Building

South Parks Road, Oxford

OX1 3EL, UK

I study learning in humans and machines, in individuals and in groups.

My research is built on questions about how people learn, teach, communicate, and work together to solve problems. I build neural network models of the cognitive and neural mechanisms that support these capacities, and I build AI tools to improve human learning and cooperation. I also study large language models from a cognitive science perspective, and investigate the impact of AI use in science.

I am currently a postdoctoral researcher in Chris Summerfield’s group at the University of Oxford, and in December 2026 I will join the School of Psychology and Neuroscience at the University of Glasgow as a Lecturer. I did my PhD at the University of Montreal, at the International Laboratory for Brain, Music & Sound Research (BRAMS) and Mila (the Quebec AI Institute), where I compared representations in artificial neural networks and the human auditory system.

Where I’m heading

Cognitive science has traditionally studied the abilities of individuals in isolation. But in real life, we rarely do anything truly alone: we are always learning and working together. Humanity’s greatest achievements come from our capacity to learn from and with each other and to organize ourselves around shared goals, and many of the defining crises of our time reflect our failure to do so.

In my own lab, I will continue to study the computational principles of human learning while centering the fact that all human learning is socially embedded. I want to know what separates the wisdom of crowds from the madness of mobs. As Nick Chater and George Loewenstein write in It’s On You, paraphrasing Mancur Olson, “the concentrated interests of the few are so powerful because they can most easily be aligned and coordinated to carry through a coherent plan of action. But coordinated action by the many, once achieved, is unstoppable.” In a hyperconnected, globalized world, our actions have far-reaching effects that are difficult to anticipate. We did not evolve to reason in such a world. I want to do research that supports the collective intelligence and coordinated action of the many, perhaps with the help of new technologies.

Interested in working with me? I’m looking for students and collaborators. See join.

news

Oct 08, 2026 I’m excited to share that I will join the School of Psychology and Neuroscience at the University of Glasgow as a Lecturer in December 2026. I’m looking for students and collaborators. Get in touch!
May 20, 2026 I gave an oral presentation, Using behavioural cloning to accelerate human language learning, at the Behavioral Clones Workshop of the Machine+Behavior Conference in Berlin. Joint work with Jirko Rubruck, Mohammed Amara, Vincent Adam, Nicholas Marsh, and Chris Summerfield.
Feb 05, 2024 I will present my postdoctoral research at COSYNE 2024 in Lisbon.
Dec 03, 2023 I have begun an NSERC postdoctoral fellowship, which allows me to continue my research in Oxford.
Dec 02, 2023 I am now also a Junior Research Fellow at Somerville College, University of Oxford.

latest posts

selected publications

  1. Book chapter
    AI Surrogacy in Psychological Research
    William D’Alessandro and Jessica Thompson
    In The Role of Artificial Intelligence in Science: Methodological and Epistemological Studies, 2026
    Forthcoming
  2. ICLR
    Reward models inherit value biases from pretraining
    Brian Christian, Jessica Thompson, Elle Michelle Yang, and 4 more authors
    In International Conference on Learning Representations (ICLR), 2026
  3. FAccT
    Reward Model Interpretability via Optimal and Pessimal Tokens
    Brian Christian, Hannah Rose Kirk, Jessica Thompson, and 2 more authors
    In Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency (FAccT), 2025
  4. Neuron
    Zero-shot counting with a dual-stream neural network model
    Jessica Thompson, Hannah Sheahan, Tsvetomira Dumbalska, and 3 more authors
    Neuron, 2024
  5. J Neurophysiol
    Forms of explanation and understanding for neuroscience and artificial intelligence
    Jessica Thompson
    Journal of Neurophysiology, 2021