Neural basis of compositional control

Area of Science:

• Neuroscience
• Cognitive Science
• Behavioral Economics

Background:

• Naturalistic goal-directed behavior involves continuous actions toward dynamic goals.
• Control theory provides a framework for understanding choices in continuous contexts, analogous to microeconomics for discrete choices.
• Continuous behavior involves blends of goals, with strategic reweighting of goal-specific control policies representing choice.

Purpose of the Study:

• To investigate the algorithmic and neural underpinnings of continuous choice behavior.
• To examine human behavior and brain activity during a continuous prey-pursuit task.

Main Methods:

• Developed a novel control-theoretic decomposition to analyze continuous behavior.
• Recorded brain activity (neural correlates) in humans performing a continuous prey-pursuit task.
• Correlated neural activity with behavioral strategies and policy changes.

Main Results:

• Pursuit strategies were modeled as a meta-controller mixing lower-level controllers for specific goals.
• Anterior cingulate cortex (ACC) neurons predicted significant shifts in policy mixtures.
• Hippocampal neurons encoded and updated latent policy states for planning, while orbitofrontal cortex (OFC) encoded task value structure.

Conclusions:

• A tripartite functional division was observed: hippocampus as state-estimating controller, ACC as meta-controller, and OFC providing value context.
• This division offers insights into the neural basis of continuous, goal-directed decision-making.
• Findings support a hierarchical control structure for complex behaviors.