Inferring reward prediction errors in patients with schizophrenia: a dynamic reward task for reinforcement learning.

Area of Science:

• Neuroscience
• Psychiatry
• Computational Psychiatry

Background:

• Dopamine system dysregulation is linked to reinforcement learning deficits and psychosis.
• Reward prediction error (RPE) signals, encoded by dopaminergic neurons, are crucial for learning and decision-making.
• Altered RPE processing may contribute to the development and manifestation of schizophrenia.

Purpose of the Study:

• To investigate belief formation and belief perseveration in schizophrenia patients during a reward prediction error task.
• To evaluate the utility of RPE measurement as a behavioral index for dopamine activity and psychotic symptoms.
• To compare RPE-related decision-making between patients with high and low psychosis and healthy controls.

Main Methods:

• Utilized a feedback-based dynamic reward task with 45 schizophrenia patients (26 high-psychosis, 19 low-psychosis) and 24 controls.
• Employed matching law analysis to assess reward sensitivities.
• Applied reinforcement learning modeling with Bayesian estimation to analyze trial-by-trial decision data.

Main Results:

• Both schizophrenia groups showed lower reward sensitivity compared to controls.
• Reinforcement learning models indicated that psychosis groups update reward values more rapidly.
• High-psychosis patients exhibited the lowest degree of choice perseveration, which trended towards a negative correlation with symptom severity (PANSS p1+p3 scores).

Conclusions:

• Schizophrenia is associated with altered reward prediction error processing, characterized by reduced reward sensitivity and accelerated value updating.
• Belief perseveration, particularly in high-psychosis individuals, may be a key feature of aberrant RPE processing in schizophrenia.
• The findings support RPE measurement as a promising avenue for understanding reward learning and decision-making deficits in psychosis.