Overestimated Volatility Underpins Aberrant Decision Making in Bipolar I Disorder

Background Bipolar disorder (BD) is marked by extreme mood fluctuations, yet the computational mechanisms underlying this instability remain poorly understood. Recent computational accounts suggest that mood instability may arise from altered reward prediction error (RPE) processing, yet little is known about the factors that modulate this process. Volatility inference—the ability to track environmental change—may be particularly relevant as it modulates the influence of prediction errors on belief updating. However, it remains unclear whether BD is characterized by altered volatility inference.Method We investigated whether BD showed altered volatility inference across distinct environmental structures using decision-making tasks and computational modeling. Individuals with bipolar I disorder (BD-I; N = 32) and healthy controls (HC; N = 40) completed two tasks: a probability-tracking task (PTT) with predetermined contingency changes across stable and volatile phases and a probabilistic reversal learning task (PRL) with performance-contingent reversals.Results The BD-I group showed reduced performance and switched choices more frequently than the HC group across both tasks. Computational modeling further revealed heightened volatility beliefs in the BD-I group, reflected by a higher prior belief about volatility in the PTT and higher tonic volatility in the PRL, together with greater behavioral stochasticity across both tasks. These computational alterations were not associated with current manic or depressive symptom severity.Conclusions These findings suggest that BD-I shows deficits in adapting to volatile environments, potentially arising from trait-like computational alterations characterized by heightened expectations of environmental change and noisy belief-to-action mapping, which may provide candidate mechanisms through which decision-making abnormalities contribute to mood instability.