Cortical mechanisms for reinforcement learning in competitive games

Area of Science:

• Neuroscience
• Behavioral Economics
• Primatology

Background:

• Game theory models optimal strategies in social interactions.
• Animal and human behavior often deviates from game theory predictions.
• Rhesus monkey behavior in zero-sum games shows deviations explained by reinforcement learning.

Purpose of the Study:

• Investigate neural mechanisms underlying decision-making in rhesus monkeys during competitive games.
• Identify functional specialization within the primate frontal cortex for action selection and outcome monitoring.
• Determine how neural signals relate to choices, opponent actions, and reward history.

Main Methods:

• Computer-simulated zero-sum games with rhesus monkeys.
• Electrophysiological recordings from dorsolateral prefrontal cortex (dlPFC) and anterior cingulate cortex (ACC).
• Analysis of neural activity in relation to animal's choices, opponent's choices, and reward history.

Main Results:

• Monkey behavior approximated a reinforcement-learning algorithm, explaining deviations from game theory.
• dlPFC neurons encoded previous choices (own and opponent) and reward history.
• ACC neurons predominantly encoded reward history.
• Demonstrated functional specialization between dlPFC and ACC in processing decision-related information.

Conclusions:

• Neural signals in dlPFC and ACC support outcome monitoring and action selection during competitive games.
• Temporally extended signals in dlPFC may link choices to delayed outcomes.
• Opponent's actions and reward history information in dlPFC and ACC inform decision-making strategies.