Reinforcement-Learning-Informed Queries Guide Behavioral Change.

Area of Science:

• Cognitive Science
• Computational Psychiatry
• Neuroscience

Background:

• Reinforcement learning (RL) parameters correlate with psychopathology symptoms.
• Cognitive-behavioral therapy (CBT) can alter these RL parameters.
• Task environment modifications influence learning in nonclinical samples.

Purpose of the Study:

• To investigate how verbal queries affect learning processes in a probabilistic reward-learning task.
• To explore the potential of CBT-guided restructuring and RL models for identifying therapeutic targets in depression.
• To determine if querying specific RL components alters computational learning parameters.

Main Methods:

• A parallel-arm design with 1,299 online participants.
• Participants completed a probabilistic reward-learning task.
• Experimental manipulation involved repeated verbal queries about task components (11 arms) versus an active control.

Main Results:

• Querying participants about RL-related task components significantly altered computational-model-defined learning parameters.
• The effects on learning parameters were specific to the targeted component of the query.
• These learning alterations were consistent across varying levels of depression symptom severity.

Conclusions:

• Verbal queries can specifically modulate learning processes defined by computational models.
• These findings support the development of novel learning-based strategies for therapeutic intervention in mood psychopathology.
• This research identifies potential new therapeutic targets within learning mechanisms for symptom change.