Policy adjustment in a dynamic economic game.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Psychiatry

Background:

• Sequential decision-making in non-stationary environments is challenging.
• Previous neuroimaging studies focused on simpler tasks with fixed reward contingencies.
• Real-world learning requires adapting to changing reward probabilities based on choice history.

Purpose of the Study:

• To investigate brain and behavioral responses during history-dependent decision-making.
• To model learning signals in non-stationary reward environments.
• To identify neural correlates of adaptive strategy shifts.

Main Methods:

• Functional magnetic resonance imaging (fMRI) in human subjects.
• A continuous decision-making task with history-dependent reward contingencies.
• Behavioral analysis and computational modeling (actor-critic model).

Main Results:

• A reward prediction error signal was identified in ventral striatal structures.
• A simple actor-critic model explained behavior and brain responses in stable conditions.
• Novel reward structures engaged prefrontal cortex (inferior frontal gyrus, anterior insula), exceeding the simple model's scope.

Conclusions:

• Ventral striatum signals reward prediction errors in history-dependent tasks.
• Prefrontal cortex involvement is crucial for adapting to new, non-stationary reward structures.
• Striatal and prefrontal interactions are key for navigating complex, real-world decision environments.