Goals vs. habits in the human brain: Cognitive and computational mechanisms

描述（由申请人提供）：习惯可以对行为施加强大的控制。有时习惯是不适应的，比如滥用药物。然而，在许多情况下，习惯可以是适应性的，使生物体能够快速有效地做出健康的选择。近年来，通过对啮齿动物的研究，在理解支持从努力的、目标导向的行为转变为更自动的、习惯性行为的神经机制方面取得了实质性进展。这项工作强调了多巴胺及其纹状体靶点在习惯中的作用，并开始提出控制其强度的原则。然而，令人惊讶的是，啮齿动物研究在理解人类目标导向与习惯行为之间的平衡方面只取得了有限的进展，阻止了科学界将这些知识应用于临床环境和日常生活。我们假设，造成这种差距的主要原因是缺乏丰富和精确的行为标记，在实验室环境中的习惯性行为在人类。该研究计划旨在解决这一差距。我们提出了一系列的功能成像（fMRI）和神经心理学研究，专注于计算表征不同类型的学习和影响其相对强度的因素。由于习惯和目标导向行为都依赖于过去的经验，我们可以利用最近在表征人类学习和记忆大脑系统方面的进展，这些进展已经导致了行为和神经信号多个方面的丰富，定量表征。利用这些习惯形成的分级和动态特征--具体来说，通过在试错学习过程中以及随后对形成的记忆进行探索的过程中检查选择和选择相关的功能磁共振成像信号--我们提出了一系列问题，旨在揭示大脑平衡目标导向行为和习惯行为的机制。我们的具体目标是：（1）确定形成灵活记忆表征的能力如何调节目标导向系统和习惯系统;（2）了解反馈的时间如何调节系统之间的平衡;（3）确定反馈的可靠性如何调节目标导向系统和习惯系统。对于每一个目标，我们测试健康受试者与功能磁共振成像，以确定多个大脑区域的动态贡献的不同方面的行为。帕金森患者的平行研究补充了功能磁共振成像研究，并提供了多巴胺能输入纹状体在习惯学习中的因果作用的证据。这项研究的结果将促进对控制习惯出现的行为和神经生物学机制以及对决策和行动的影响的理解。确定用于学习的多个大脑系统如何支持行为从目标导向到习惯控制的转变，将为未来潜在治疗干预的转化工作奠定基础，这些干预可以自适应地将行为转变为健康和疾病中健康习惯的形成。

DESCRIPTION (provided by applicant): Habits can exert powerful control over behavior. Sometimes habits are maladaptive, such as in substance abuse. In many cases, however, habits can be adaptive, enabling organisms to engage in healthy choices rapidly and efficiently. In recent years there has been substantial progress in understanding the neural mechanisms that support the transition from effortful, goal-directed to more automatic, habitual behavior through research in rodents. This work has emphasized the role of dopamine and its striatal targets in habits and has begun to suggest the principles by which their strength is controlled. Surprisingly, however, the rodent research has led to only limited progress in understanding the balance between goal-directed vs. habitual behaviors in humans, preventing the scientific community from being able to apply this knowledge to clinical settings and everyday life. We hypothesize that the main reason for this gap is the lack of rich and precise behavioral markers for habitual behavior in laboratory settings in humans. The proposed research program aims to address this gap. We propose a series of functional imaging (fMRI) and neuropsychological studies that focus on computational characterization of distinct types of learning and the factors that impact their relative strength. Because both habits and goal-directed behaviors depend on past experience, we can leverage recent advances in characterizing brain systems for learning and memory in humans that have already led to a rich, quantitative characterization of multiple aspects of behavior and neural signaling. Using these graded and dynamic signatures of habit formation - specifically, by examining choices and choice-related fMRI signals during trial-and-error learning, and during subsequent probes of the memories formed - we propose to ask a series of questions collectively aimed at uncovering the mechanisms by which the brain balances goal-directed and habitual behaviors. Our specific aims are: (1) To determine how the capacity to form flexible memory representations modulates goal-directed and habitual systems (2) To understand how the timing of feedback modulates the balance between the systems; (3) To determine how the reliability of feedback modulates goal-directed and habitual systems. For each aim, we test healthy subjects with fMRI to determine the dynamic contribution of multiple brain regions to different aspects of behavior. Parallel studies with Parkinson's patients complement the fMRI studies and provide evidence about the causal role of dopaminergic inputs to the striatum in habit learning. Results from this research will advance understanding of the behavioral and neurobiological mechanisms that control the emergence of habits and the implications for decisions and actions. Determining how multiple brain systems for learning support the transition of behavior from goal-directed to habitual control will lay the foundation for future translational work on potential treatment interventions that can adaptively shift behavior towards the formation of healthy habits in both health and disease.