How instructed knowledge modulates the neural systems of reward learning

Area of Science:

• Neuroeconomics
• Cognitive Neuroscience
• Decision Science

Background:

• Reinforcement learning models explain reward-based decision-making behavior and neural activity.
• Traditional models often overlook symbolic communication and instructed knowledge as learning mechanisms.
• Humans utilize efficient communication for learning, minimizing errors compared to pure trial-and-error.

Purpose of the Study:

• To investigate how instructed knowledge of cue-reward probabilities affects behavioral performance.
• To examine the neural correlates of learning when explicit information is provided versus learned through reinforcement.
• To explore the role of the dorsolateral prefrontal cortex (DLPFC) in integrating instructed information during value-based learning.

Main Methods:

• Participants engaged in economic decision-making tasks with varying levels of instructed cue-reward probability information.
• Behavioral performance metrics were recorded to assess decision accuracy and efficiency.
• Blood-oxygen-level-dependent (BOLD) responses were measured using functional magnetic resonance imaging (fMRI) focusing on reward-related brain regions and the DLPFC.

Main Results:

• Instructed knowledge significantly improved behavioral performance in decision-making tasks.
• Reinforcement learning-related BOLD responses were diminished in the nucleus accumbens, ventromedial prefrontal cortex, and hippocampal complex.
• Reduced BOLD responses in valuation regions were functionally correlated with increased activation in the dorsolateral prefrontal cortex (DLPFC).

Conclusions:

• Instructed knowledge provides an efficient alternative to pure reinforcement learning for acquiring action values.
• The DLPFC plays a crucial role in dynamically modulating responses in valuation regions based on the utility of instructed information.
• These findings highlight the brain's flexibility in employing different learning strategies, integrating explicit knowledge to optimize decision-making.