A Neurobiologically-based Neural Network Model of Risky Decision-making

描述（由申请人提供）：风险决策导致普遍的负面健康结果（例如，酒精和药物滥用，危险的性决定，事故）。这些人的一个核心特征（例如，男男性接触者（MSM）的风险在于，他们继续从事具有非常有益的短期后果，但极其负面的长期后果的行为，包括医疗，社会和法律的问题。为什么他们很难做出正确的选择？越来越多的研究表明，有动机的人类决策是三个系统之间动态相互作用的结果：（1）一个相对自动的食欲系统，被称为“冲动系统”，（2）一个执行控制系统，被称为“反射系统”[11]，以及（3）一个神经系统，它将内感受信号转化为一种欲望的感觉，或可能有助于推动个体获得奖励的冲动[5，12]，抑制自我控制所需的认知资源。 不幸的是，我们对这些复杂的神经系统如何相互作用，以及如何与各种社会和环境因素相互作用，以产生冒险行为，何时，为谁，以及为什么，缺乏系统的理解。这一差距阻碍了预防和干预科学的更快发展。适当的计算工具可以帮助解决这一关键障碍，并更好地推进累积科学，但它们目前缺乏。 该项目旨在通过开发可推广的计算工具来解决这一差距：这些系统相互作用的经验证的神经生物学神经网络模型可以改变我们推进理论和有效干预的能力。为此，一个由社会科学家、神经科学家和计算神经科学家组成的团队将（a）开发基于生物学的计算模型，利用和整合现有的神经网络模型，这些模型将行为视为从接近中涌现出来的，避免动机结构[13，14]，并在不同的规模水平上，模拟激励处理和学习的潜在生物学基础的神经网络模型，执行功能和决策[15，16];（B）通过预测180名年轻MSM子样本的神经和行为反应来测试、验证和完善模型（性风险，性风险甲基苯丙胺用户，和非风险）从一个完整的NIDA成像研究的风险决策;（c）通过重点测试评估其可推广性，并与更多的NIDA数据子样本进行交叉验证;以及（d）进行探索性计算分析，目的在于同时预测MSM在虚拟日期模拟中随时间的顺序神经和行为动力学，并使用该模型来探索什么样的干预措施，何时以及对谁来说可能更有效地减少冒险行为。对这些神经系统及其相互作用的更深入了解，将改变我们推进理论、设计有效的降低风险干预措施、增强社会健康和福祉的能力，同时降低经济成本。

DESCRIPTION (provided by applicant): Risky decision-making leads to pervasive negative health outcomes (e.g., alcohol and drug abuse, risky sexual decisions, accidents). One central characteristic of such individuals (e.g., risky men who have sex with men (MSM)) is that they continue to engage in behaviors with very rewarding short-term consequences, but extremely negative long-term consequences, including medical, social and legal problems. Why do they have such difficulties making the right choices? A growing body of research suggests that motivated human decision-making is the result of a dynamic interplay among three systems: (1) a relatively automatic appetitive system, which has been called the "Impulsive System", (2) an executive control system, which has been called the "Reflective System"[11], and (3) a neural system that translates interoceptive signals into what one experiences as a feeling of desire, or urge [5,12] that may help propel individuals towards reward, and inhibit cognitive resources needed for self-control. Unfortunately, we lack a systematic understanding of how these complex neural systems interact with each other and with various social and contextual factors to produce risk-taking, when, for whom, and why. This gap impedes more rapid advancements in prevention and intervention science. Adequate computational tools could help address this critical barrier, and better advance a cumulative science, but they are currently lacking. This project aims to address this gap by developing generalizable computational tools: A validated neurobiologically based, neural network model of the interaction of these systems could transform our ability to advance theory and effective interventions. To this end, a team of social scientists, neuroscientists, and computational neuroscientists will (a) develop biologically-based computational models that leverage and integrate existing neural network models that view behavior as emergent from approach and avoid motivational structures[13,14] and, at a different level of scale, neural network models that simulate the underlying biological basis of incentive processing and learning, executive function, and decision-making [15,16]; (b) test, validate, and refine the model by predicting the neural and behavioral responses of a subsample from 180 young MSM (sexually risky, sexually risky methamphetamine users, and non-risky) from a completed NIDA imaging study on risky decision-making; (c) assess its generalizability via focused tests, and cross validate with additional NIDA data subsamples; and (d) conduct exploratory computational analyses aimed at concurrently predicting MSM's sequential neural and behavioral dynamics in a virtual date simulation over time, and using the model to explore what interventions, when, and for whom might more effectively reduce risk-taking. A deeper understanding of these neural systems and their interactions, will transform our ability to advance theory, design effective risk-reduction interventions and enhance societal health and well-being, while reducing economic costs.