Background: Prior findings on ambiguous risk-taking in schizophrenia spectrum disorders are mixed. However, research on reversal learning, which requires adaptation to changing contingencies, consistently shows reduced flexibility and reward updating. We used a novel adaptation of the Balloon Analogue Risk Task (BART) that reverses risk probabilities and conducted exploratory analyses linking latent computational parameters of risk-taking with schizotypal traits.

Methods: An online sample of adults (n=100) completed our reversal BART. The task presented blocks of three balloon colors with different explosion probabilities. Halfway through the task, the colors with highest and lowest risk switched probabilities. Data were fit hierarchically by pre and post reversal block to three computational models: Scaled Target Learning, Exponential Weight Mean Variance, and Four Parameter. Participants also completed the Cardiff Anomalous Perceptions Scale, Peters Delusion Inventory (PDI), and Schizotypal Personality Questionnaire (SPQ). Correlation analyses examined relationships between model parameters and questionnaire scores.

Results: After reversal, higher SPQ scores were associated with lower Scaled Target Learning reward learning rates on cashed trials. Inverse temperature from the Exponential Weight Mean Variance model was positively associated with PDI scores both before and after reversal.

Conclusions: Our preliminary findings suggest that higher schizotypal traits are linked to reduced reward learning when a once punishing contingency becomes safer, and that more stochastic behavior overall is linked to heightened delusional ideation. This indicates difficulty updating previously negative associations and may reflect trait-level mechanisms relevant to belief updating in ambiguous risk.