Skip to content
Kahn Lab Computational Cognitive Neuroscience

Publications

Papers and preprints

A complete list is also available on Google Scholar, and in the CV.

Preprints

2026

2025

Humans rationally balance detailed and temporally abstract world models

Kahn, A. E., Daw, N. D.

Communications Psychology 3 1–11 · 2025

Participants relied on a mixture of temporally abstract prediction and detailed mental simulation, and adjusted the balance between the two in a manner consistent with the costs each one carried.

  • planning
  • successor representation
  • reinforcement learning

Network structure influences the strength of learned neural representations

Kahn, A. E., Szymula, K., Loman, S., Haggerty, E. B., Nyema, N., Aguirre, G. K., Bassett, D. S.

Nature Communications 16 994 · 2025

When the graph underlying a sequence was modular rather than lattice-like, representations in visual areas reflected that structure more strongly, over timescales of minutes rather than sessions.

  • graph learning
  • fMRI
  • network science

Trial-by-trial learning of successor representations in human behavior

Kahn, A. E., Bassett, D. S., Daw, N. D.

PLOS Computational Biology 21 e1013696 · 2025

Much evidence suggests people use a successor representation, but little speaks to how they acquire one. We measure its acquisition trial by trial and compare temporal difference learning against alternative accounts.

  • successor representation
  • reinforcement learning
  • planning

2024

2023

2022

2021

2020

Human information processing in complex networks

Lynn, C. W., Papadopoulos, L., Kahn, A. E., Bassett, D. S.

Nature Physics 1–9 · 2020

The information a network conveys to a human observer depends not only on its topology, but on the errors people make in perceiving it.

  • graph learning
  • network science

2019

2018

Network constraints on learnability of probabilistic motor sequences

Kahn, A. E., Karuza, E. A., Vettel, J. M., Bassett, D. S.

Nature Human Behaviour 2 936–947 · 2018

Participants learned the same transition statistics more readily when the graph generating them was modular, indicating sensitivity to structure beyond pairwise associations.

  • graph learning
  • network science

2017

2015