2018 Conference on Cognitive Computational Neuroscience

5-8 September 2018    Philadelphia, Pennsylvania

Technical Program

Paper Detail

Paper: PS-2A.35
Session: Poster Session 2A
Location: Symphony/Overture
Session Time: Friday, September 7, 17:15 - 19:15
Presentation Time:Friday, September 7, 17:15 - 19:15
Presentation: Poster
Publication: 2018 Conference on Cognitive Computational Neuroscience, 5-8 September 2018, Philadelphia, Pennsylvania
Paper Title: Neural mechanisms of human decision-making
Manuscript:  Click here to view manuscript
DOI: https://doi.org/10.32470/CCN.2018.1219-0
Authors: Kai Krueger, Seth Herd, Ananta Nair, eCortex Inc, United States; Jessica Mollick, Yale University, United States; Randall O'Reilly, University of Colorado Boulder, United States
Abstract: This work presents a neural network model and theory of cognitive decision-making. It attempts to explain the interactions of cortical and subcortical mechanisms and how these may lead to some of the behavioral properties of flexible complex decision-making. Specifically, we model the interactions among several layers of cortex, basal ganglia and the dopamine system. Key to our theory is that all cognitive decisions result from corticostriatal-thalamus loops akin to those heavily studied in animal motor action selection. Relevant areas of cortex propose a plan of action using associative mechanisms, driven by reinforcement learning; then other cortical areas use that information from sensory input, contextual information and internal goal states to make a prediction about outcome. That prediction is used by striatum to make a go/nogo decision on that plan. Cortical areas thereby learn in a supervised way from actual observed outcomes, whereas the basal ganglia learns its go/nogo decision based on dopaminergic reinforcement signals. By breaking up complex decisions into sequential, simple go/nogo decisions, the same canonical decision-making circuit, as used in basic action selection, can scale up to flexible complex decisions. Furthermore, we postulate that model-free and model-based decision-making are different modes of the same canonical decision-making circuit.