Hirohito M. Kondo ^1* Takeyuki Oba ² Takahiro Ezaki ³ Takanori Kochiyama ⁴ Yasuhiro Shimada ⁵ Hideki Ohira ²

• ¹ Chukyo University, Nagoya, Aichi, Japan
• ² Nagoya University, Nagoya, Aichi, Japan
• ³ The University of Tokyo, Bunkyo, Tōkyō, Japan
• ⁴ ATR-Promotions, Kyoto, Japan
• ⁵ National Institute of Information and Communications Technology (Japan), Koganei, Tôkyô, Japan

Decision-making under risk is a common challenge. It is known that risk-taking behavior varies between contexts of reward and punishment, yet the mechanisms underlying this asymmetry in risk sensitivity remain unclear. This study used a monetary task to investigate neurochemical mechanisms and brain dynamics underpinning risk sensitivity. Twenty-eight participants engaged in a task requiring selection of visual stimuli to maximize monetary gains and minimize monetary losses. We modeled participant trial-and-error processes using reinforcement learning. Participants with higher subjective utility parameters showed risk preference in the gain domain (r = -0.59) and risk avoidance in the loss domain (r = -0.77). Magnetic resonance spectroscopy (MRS) revealed that risk avoidance in the loss domain was associated with γ-aminobutyric acid (GABA) levels in the ventral striatum (r = -0.42), but not in the insula (r = -0.15). Using functional magnetic resonance imaging (fMRI), we tested whether risk-sensitive brain dynamics contribute to participant risky choices. Energy landscape analyses demonstrated that higher switching rates between brain states, including the striatum and insula, were correlated with risk avoidance in the loss domain (r = -0.59), a relationship not observed in the gain domain (r = -0.02). These findings from MRS and fMRI suggest that distinct mechanisms are involved in gain/loss decision making, mediated by subcortical neurometabolite levels and brain dynamic transitions.