Attention-deficit/hyperactivity disorder (ADHD) is often accompanied by excessive risky behavior, and an impulsive trait has been proposed to be associated with risk-taking. However, the aspect of the cognitive process that impulsivity influences is not well understood. We hypothesized that impulsivity could be related to an overhasty shifting of beliefs during risk learning, thereby resulting in enhanced risk-taking behavior. In this study, we tested our hypothesis using the Bayesian modeling approach and predicted overhasty learning by a data-driven approach. We used an openly available task-based functional magnetic resonance imaging (fMRI) dataset. Participants with adult ADHD (n = 42) completed the balloon analog risk task (BART). By fitting our computational model that encapsulates the degree of overhasty learning, we estimated the degree of learning bias and investigated its relationship with Barratt Impulsiveness Scale (BIS) outcomes. Moreover, we created a connectome-based predictive model (CPM) based on fMRI data to predict the degree of risk-learning bias. The degree of overhasty learning in ADHD patients was significantly correlated with the BIS score (r = 0.424, p = 0.009). The CPM predicted the ‘learning bias’ parameter using negatively correlated edges (r = 0.341, p = 0.041; q² = 0.092). The ‘hub nodes’ in the predictive network were in the frontal lobe, including the orbitofrontal area. Our findings suggest that impulsivity in ADHD patients is associated with overhasty updating of beliefs during risk learning. Weak functional connectivity to the both dorso-lateral prefrontal and orbitofrontal lobes is predictive of the degree of overhasty learning.

注意缺陷/多动障碍 (ADHD) 通常伴随着过度的冒险行为，并且有人提出冲动性特征与冒险行为有关。然而，冲动影响的认知过程的方面还不是很清楚。我们假设冲动性可能与风险学习期间过快的信念转变有关，从而导致冒险行为的增强。在这项研究中，我们使用贝叶斯建模方法测试了我们的假设，并通过数据驱动的方法预测了过度学习。我们使用了一个公开可用的基于任务的功能磁共振成像 (fMRI) 数据集。成人 ADHD 参与者 (n = 42) 完成了气球模拟风险任务 (BART)。通过拟合我们包含过快学习程度的计算模型，我们估计了学习偏差的程度，并研究了它与 Barratt Impulsiveness Scale (BIS) 结果的关系。此外，我们基于 fMRI 数据创建了一个基于连接组的预测模型 (CPM)，以预测风险学习偏差的程度。ADHD患者过快学习的程度与BIS评分显着相关（r  = 0.424，p = 0.009）。CPM 使用负相关边（r  = 0.341，p  = 0.041；q ²  = 0.092）预测“学习偏差”参数。预测网络中的“枢纽节点”位于额叶，包括眶额区。我们的研究结果表明，多动症患者的冲动与风险学习过程中对信念的过快更新有关。与背外侧前额叶和眶额叶的弱功能连接可以预测过快学习的程度。