Christopher Summerfield

Distinct roles of hippocampus and neocortex in symbolic compositional generalization

The paper investigates how the hippocampus and neocortex contribute to symbolic compositional generalization by analyzing BOLD data during a spatial inference task.

A recurrent neural network model of prefrontal brain activity during a working memory task

This paper trains recurrent neural networks to model how retro-cues transform memory representations in macaque prefrontal cortex during a working memory task.

Goal-seeking compresses neural codes for space in the human hippocampus and orbitofrontal cortex

The paper shows that goal-directed behavior compresses neural representations of space in the hippocampus and orbitofrontal cortex, with compressed representations predicting successful learning.

Predictive and Interpretable: Combining Artificial Neural Networks and Classic Cognitive Models to Understand Human Learning and Decision Making

The paper combines artificial neural networks with classic cognitive models to create interpretable hybrid models that reveal human learning mechanisms similar to reinforcement learning.

Rich and lazy learning of task representations in brains and neural networks

The paper compares 'lazy' and 'rich' coding strategies in neural networks and biological brains for multitasking, finding evidence for rich coding in human and macaque neural data.

Neural knowledge assembly in humans and deep networks

The paper shows that human brains and neural networks can rapidly reorganize knowledge when new information is introduced, using BOLD signals and a gradient descent model.

Model sharing in the human medial temporal lobe

The paper shows humans share neural state transition models between similar contexts during planning, with medial temporal lobe activity reflecting this sharing and stronger updates after positive outcomes.

Neural state space alignment for magnitude generalization in humans and recurrent networks

This study shows that both humans and recurrent neural networks develop aligned neural number lines to generalize magnitude concepts across contexts.

Hippocampal place cells encode global location but not changes in environmental connectivity in a 4-room navigation task

Selective integration during sequential sampling in posterior neural signals

The paper tests a selective integration model of decision-making using EEG, finding that neural signals encode locally preferred attributes with higher gain.

Causal inference in the multisensory brain

This study identifies a spatio-temporal cascade of multisensory integration in the human brain, showing prefrontal cortex guides flexible integration based on sensory representations.

Human noise blindness drives suboptimal cognitive inference

This study shows humans adapt near-optimally to sensory noise but exhibit suboptimal cognitive judgments under integration noise, captured by a Bayesian model blind to this noise.

Neural structure mapping in human probabilistic reward learning

The paper shows that human neural representations of novel visual stimuli align with magnitude-based mental number line representations during reward-guided learning.

Focused learning promotes continual task performance in humans

This study shows that focused training improves continual task performance in humans by enabling factorised task representations, which reduces interference between tasks.

A network for computing value homeostasis in the human medial prefrontal cortex

The paper shows that the medial prefrontal cortex network computes value homeostasis by tracking imbalance and redress in goal-related assets during decision-making.

A unified theory of distraction in human perceptual, cognitive and economic decision-making

This paper presents a unified model of distraction that explains perceptual, cognitive, and economic decision-making phenomena and predicts new behavioral outcomes.

Robust Averaging Protects Decisions from Noise in Neural Computations

The paper shows robust averaging improves decision-making performance in the presence of late noise by reducing the impact of noisy neural computations during decision-making.