This page was translated automatically from Turkish. Read the original

This page is an AI-assisted Turkish summary of the study presented at the 26th World Congress of Psychiatry. The study belongs to the presenting authors; please refer to the original source when citing it.

EV0908

Instructional control of reinforcement learning in schizophrenia: a neurocomputational investigation

INSTRUCTIONAL CONTROL OF REINFORCEMENT LEARNING IN SCHIZOPHRENIA: A NEUROCOMPUTATIONAL INVESTIGATION.

Patients with schizophrenia-spectrum disorders exhibited a higher temperature parameter than healthy controls in a reinforcement learning task, indicating that their choices were less strongly tied to learned values and were more exploratory or noisy.

The aim of the study was to compare instructional control of behavior in a reinforcement learning task between patients with schizophrenia-spectrum disorders (SZ) and healthy controls (HC) using neurocomputational methods. The study included 101 SZ patients (56 men / 45 women, aged 37.59 ± 13.48 years) and 114 healthy controls (42 men / 72 women, aged 35.77 ± 17.43 years). Participants chose between probabilistically reinforced stimuli and were falsely told that a particular stimulus had the highest (or lowest) probability of reinforcement; during the learning phase, stimulus pairs were presented with reinforcement probabilities of 80%/20%, 70%/30%, and 60%/40%. Choice and avoidance patterns were assessed during the test phase, and behavioral data were also analyzed using a Bayesian model. Consistent with previous reports and the confirmation bias model, this manipulation resulted in overvaluation of the instructed stimulus in both the SZ and HC groups, even after participants had experienced its actual value. Bayesian modeling revealed a difference between the two groups in the temperature parameter (T), which determines how strongly estimated values guide choices; T values were higher in SZ patients than in healthy controls (2.36 ± 1.50 versus 1.69 ± 1.49), indicating that choices were more loosely tied to value differences and more exploratory, and that beliefs or instructions translated less strongly into action. Overall, compared with healthy controls, SZ patients made more random or noisy choices, were less consistent in choosing the highest-value option, translated learned values into behavior less effectively, and exhibited greater exploration or variability in their choices. These findings suggest altered sensitivity to explicit instructions during reinforcement learning in schizophrenia.

All WCP 2026 summaries