A shared temporal window of integration across cognitive control and reinforcement learning paradigms: A...

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry

Background:

• Cognitive control enables goal-directed behavior by overriding automatic responses.
• Reinforcement learning involves acquiring actions based on feedback and rewards.
• Recent theories suggest potential overlaps between cognitive control and reinforcement learning.

Purpose of the Study:

• To investigate shared learning mechanisms between cognitive control and reinforcement learning.
• To determine if a similar time window of integration exists in control learning and reinforcement learning rate.
• To explore the relationship between control learning and reinforcement learning parameters.

Main Methods:

• Correlational analysis of a large public dataset (n=522).
• Utilized data from a probabilistic selection task, a probabilistic Wisconsin Card Sorting Task (WCST), and a Stroop task.
• Examined time scale of control indices and learning rate parameters.

Main Results:

• Found significant correlations between control indices time scale and learning rate in the probabilistic WCST.
• Observed no correlation between the learning rate parameters of the two reinforcement learning tasks.
• Indicated task-specific value updating processes in reinforcement learning.

Conclusions:

• Suggests a shared learning mechanism underlies both cognitive control and reinforcement learning.
• Highlights that value updating in reinforcement learning can be task-specific.
• Proposes that tasks like Stroop and WCST may share updating mechanisms due to feature-specific requirements.