Neurostimulation Reveals Context-Dependent Arbitration Between Model-Based and Model-Free Reinforcement Learning.

Area of Science:

• Neuroscience
• Cognitive Science
• Reinforcement Learning

Background:

• Humans utilize both model-based (MB) and model-free (MF) reinforcement learning systems.
• The neural mechanisms governing the arbitration between MB and MF learning remain poorly understood.

Purpose of the Study:

• To provide causal evidence for a neural mechanism that arbitrates between MB and MF reinforcement learning.
• To investigate the role of the left ventrolateral prefrontal cortex in this arbitration process.
• To determine if this arbitration mechanism is context-dependent.

Main Methods:

• Applied excitatory and inhibitory transcranial direct current stimulation (tDCS) over the left ventrolateral prefrontal cortex.
• Utilized tasks with varying contexts that favored either MB or MF control.
• Measured behavioral shifts in control between MB and MF learning systems.

Main Results:

• Causal evidence for a neural arbitrator in the left ventrolateral prefrontal cortex was established.
• Excitatory and inhibitory tDCS induced bidirectional shifts in control between MB and MF learning.
• The arbitration mechanism's sensitivity, indicated by switching frequency, was modulated by tDCS.
• Observed effects were dependent on task context, demonstrating context-specific arbitration.

Conclusions:

• A specific neural mechanism in the left ventrolateral prefrontal cortex causally arbitrates between MB and MF reinforcement learning.
• This arbitration is not fixed but flexibly adapts to environmental demands and task contexts.
• The findings illuminate how the brain dynamically selects between different learning strategies.