Impaired Expected Value Computations Coupled With Overreliance on Stimulus-Response Learning in Schizophrenia

Area of Science:

• Cognitive Neuroscience
• Psychiatry
• Computational Psychiatry

Background:

• Schizophrenia is associated with decision-making deficits.
• These deficits may stem from overreliance on stimulus-response systems and impaired expected value representation.
• This study investigates conditions that exacerbate decision-making impairments in schizophrenia.

Purpose of the Study:

• To test computational frameworks of decision-making in schizophrenia.
• To identify conditions where maladaptive expected value representations are most evident.
• To delineate specific decision-making impairments in schizophrenia.

Main Methods:

• A modified reinforcement learning paradigm was used.
• Participants included medicated individuals with schizophrenia and healthy volunteers.
• Computational modeling (actor-critic and Q-learning) analyzed learning strategies.

Main Results:

• Individuals with schizophrenia showed performance impairments correlating with reduced expected value learning (Q-learning).
• A contextual choice bias in schizophrenia was observed.
• This bias was explained by context-dependent reward prediction errors in the actor-critic model.

Conclusions:

• Decision-making impairments in schizophrenia increase with computational demands on expected value.
• Contextual choice bias suggests overreliance on stimulus-response learning due to maladaptive value representation.
• Findings clarify conditions triggering decision-making deficits in schizophrenia.