A One-Shot Shift from Explore to Exploit in Monkey Prefrontal Cortex

Area of Science:

• Neuroscience
• Cognitive Science
• Animal Behavior

Background:

• Animal learning is typically slow, relying on cumulative changes driven by reward prediction errors.
• However, rapid learning occurs when the abstract structure of a problem is known, allowing quick association of new information with existing roles.
• The frontal cortex is implicated in this rapid, structure-based learning process.

Purpose of the Study:

• To investigate information seeking and utilization within a known problem structure.
• To examine neural activity in the frontal cortex during rapid, one-shot learning.
• To understand the explore/exploit decision-making process in monkeys within a familiar task.

Main Methods:

• Monkeys were trained on an explore/exploit task involving object selection and reward feedback.
• Neural activity in the frontal cortex was recorded during task performance.
• Behavioral data, including choice patterns and learning speed, were analyzed.

Main Results:

• Many frontal cortex cells exhibited an explore/exploit preference that aligned with one-shot learning.
• A significant switch from an explore to an exploit state occurred after a single learning trial.
• This binary switch was not explained by continuous changes in expectancy or prediction error.
• Explore/exploit preferences were independent for object selection and feedback receipt stages.

Conclusions:

• Frontal cortex activity may control distinct explore and exploit processes within a known task structure.
• The observed neural activity supports a rapid, one-trial switching mechanism between exploring and exploiting.
• This suggests that frontal cortex plays a critical role in efficient learning and decision-making when task structure is understood.