The Reward-Complexity Trade-off in Schizophrenia

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry
• Reinforcement Learning

Background:

• Action selection relies on policies mapping states to actions.
• Policy complexity, or memory needed, increases with state-dependence.
• Capacity limits on policy complexity create a reward-complexity trade-off.

Purpose of the Study:

• To empirically characterize the reward-complexity trade-off in schizophrenia patients and healthy controls.
• To investigate cognitive effort abnormalities in schizophrenia.

Main Methods:

• Empirical characterization of reward and policy complexity.
• Comparison between schizophrenia patients and healthy controls.
• Modeling using a cost-sensitive actor-critic framework.

Main Results:

• Schizophrenia patients exhibit lower average policy complexity.
• Schizophrenia patients' policies are more biased from the optimal trade-off curve.
• Both groups lie on the same empirical reward-complexity trade-off curve.

Conclusions:

• Schizophrenia is associated with altered reward-complexity trade-offs.
• Findings illuminate cognitive effort abnormalities in schizophrenia.
• Cost-sensitive actor-critic models can explain these differences.