Model-based influences on humans' choices and striatal prediction errors.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Psychiatry

Background:

• The mesostriatal dopamine system is linked to model-free reinforcement learning.
• Functional magnetic resonance imaging (fMRI) blood-oxygen-level-dependent (BOLD) signals in the ventral striatum correlate with model-free prediction errors.
• Behavioral studies indicate that human decision-making also involves model-based planning.

Purpose of the Study:

• To investigate the interaction between model-based and model-free influences on human choice behavior.
• To examine the purity of the ventral striatal BOLD signal as a sole indicator of model-free learning.
• To explore the computational architecture underlying high-level human decision-making.

Main Methods:

• A multistep decision task was designed to differentiate model-based and model-free influences on choice.
• Behavioral data was analyzed to quantify the contributions of both learning systems.
• fMRI BOLD signals in the ventral striatum were recorded and analyzed in relation to choice behavior.

Main Results:

• Human choices reflected a combination of both model-based and model-free learning influences.
• The ventral striatal BOLD signal mirrored these influences, reflecting both model-free and model-based predictions.
• The proportions of model-based and model-free signals matched those that best explained choice behavior.

Conclusions:

• The findings challenge the concept of a distinct, separate model-free learning system.
• Results suggest a more integrated computational architecture for human decision-making.
• The ventral striatal BOLD signal is not a pure report of model-free learning but reflects integrated value computations.