Large-scale neural network computations and multivariate representations during approach-avoidance conflict...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Decision Science

Background:

• Real-world decisions often involve balancing potential rewards against threats.
• Understanding approach-avoidance conflict (AAC) is crucial, yet mechanisms are poorly understood.
• Separate studies explored reward and threat decision-making, but not their interaction.

Purpose of the Study:

• Investigate decision-making under approach-avoidance conflict (AAC).
• Examine how the brain integrates reward and threat information.
• Identify neural correlates of conflict processing and choice biases.

Main Methods:

• Developed a novel task with concurrent reward and threat learning.
• Employed computational learning models and reinforcement learning.
• Utilized independent component analysis (ICA) and multivariate pattern analysis (MVPA).

Main Results:

• A modified reinforcement learning model predicted choices by integrating reward and threat values.
• Specific brain networks differentially encoded reward vs. threat prediction errors and values.
• The left frontoparietal network uniquely encoded the degree of conflict.
• Reward and threat outcomes were decoded in salience and inferior frontal networks, respectively.
• Relative neural activation predicted approach vs. avoidance decisions.

Conclusions:

• Decision-making in AAC integrates reward and threat values via an internal policy.
• Brain networks differentially process reward and threat signals, with unique roles in conflict encoding.
• Neural representations of potential outcomes bias choices in AAC scenarios.