Abstract
Optimal decision-making employs short-term rewards and abstract long-term information based on which of these is deemed relevant. Employing short-vs. long-term information is associated with different learning mechanisms, yet neural evidence showing that these two are dissociable is lacking. Here we demonstrate that long-term, inference-based beliefs are biased by short-term reward experiences and that dissociable brain regions facilitate both types of learning. Long-term inferences are associated with dorsal striatal and frontopolar cortex activity, while short-term rewards engage the ventral striatum. Stronger concurrent representation of reward signals by mediodorsal striatum and frontopolar cortex correlates with less biased, more optimal individual long-term inference. Moreover, dynamic modulation of activity in a cortical cognitive control network and the medial striatum is associated with trial-by-trial control of biases in belief updating. This suggests that counteracting the processing of optimally to-be-ignored short-term rewards and cortical suppression of associated reward-signals, determines long-term learning success and failure.
Original language | English |
---|---|
Article number | 1690 |
Journal | Nature Communications |
Volume | 8 |
Issue number | 1 |
DOIs | |
State | Published - 1 Dec 2017 |
Externally published | Yes |
Bibliographical note
Publisher Copyright:© 2017 The Author(s).
ASJC Scopus subject areas
- General Chemistry
- General Biochemistry, Genetics and Molecular Biology
- General
- General Physics and Astronomy