Computational and circuit mechanisms of decision making

摘要 知觉决策的神经生物学阐明了高等认知的基本神经机制 功能，对它的理解将启发治疗神经和精神疾病的新策略 影响思想、感知和意识。调查的重点是在 获得感官证据并致力于一个主张、行为选择或计划。进展是 通过在猴子的前额叶和顶叶联合皮质中发现持续的神经活动促进了这一研究。 通过要求猴子通过眼球运动来交流它的决定，决定的过程是可以观察到的 作为对一种或另一种行为的进化神经承诺。现在人们对神经的了解很多。 证据积累的基础机制，速度和准确性之间的权衡， 评估对决策的信心，并纳入偏见。然而，目前还不清楚这些是如何 机制可以应用于更复杂的决策，这些决策不会被解释为来自单个 行动选择的证据来源。进展的一个关键限制是关于流动的知识的空白 当证据到行动的路径被间接和灵活地控制时，电路之间的信息。海流 Proposal通过开发新的行为任务来解决这个问题，在这些任务中，从感觉到决策的路径 行动被转移或精心设计，它利用新兴工具来测量和操纵 为了描述服务中并行、串联和多路传输的神经元群体之间的相互作用 计算。目标1阐明了两个顶区之间的上下文相关的相互作用 通过眼睛或手臂的运动来传达的决定。同步多通道神经记录来自 内侧和外侧顶叶内侧区域将暴露出证据的连续遗传或并行处理模式， 决定和终止。目标2阐明了在以下情况下决策表示的动态变化 决策到行动的映射在决策过程中会发生变化。神经记录是从 猴子的顶叶皮质在一项由介入性眼动任务暂停的知觉决定中。 如果成功，目标2将在凝视变化的视觉稳定性和完整性之间建立联系 一项跨越意图变化的决定。目标3调查了当两个 证据流支持关于同一对象的两种截然不同的决定，即双重决定。行为 来自人类和猴子的证据表明，感官证据可以并行获得，但 按顺序合并到双重决定中。因此，目标3有望阐明 这一系列的整合，从而开始解释为什么精神操作需要这么长的时间。团结在一起 拟议的研究将开辟计算和机械询问电路相互作用的新领域 决策和认知控制服务。

Abstract The neurobiology of perceptual decision-making elucidates fundamental neural mechanisms of higher cognitive function, the understanding of which will inspire new strategies to treat neurological and psychiatric diseases affecting thought, perception and awareness. The inquiry focuses on processes that intervene between the acquisition of sensory evidence and commitment to a proposition, behavioral choice, or plan. Progress was facilitated by the discovery of persistent neural activity in prefrontal and parietal association cortex of the monkey. By requiring a monkey to communicate its decision with an eye movement, the decision process is observable as an evolving neural commitment to one action or another. Much is now understood about the neural mechanisms that underlie the accumulation of evidence, the tradeoff between speed and accuracy, the assessment of confidence in a decision, and the incorporation of bias. However, it is unknown how these mechanisms can apply to more complex decisions that are not construed as a dedicated chain from a single source of evidence to action selection. A critical limitation to progress is a gap in knowledge about the flow of information between circuits when the path from evidence to action is indirect and flexibly controlled. The current proposal addresses this problem by developing new behavioral tasks in which the path from sensation to decision to action is diverted or elaborated, and it exploits emerging tools to measure and manipulate neural activity in order to characterize interactions between populations of neurons in the service parallel, serial and multiplexed computations. Aim 1 elucidates context-dependent interactions between two parietal areas that mediate decisions communicated by an eye or arm movement. Simultaneous multichannel neural recording from the medial and lateral intraparietal areas will expose patterns of serial inheritance or parallel processing of evidence, the decision and termination. Aim 2 elucidates the dynamical changes in the representation of a decision when the decision to action mapping changes during the decision itself. Neural recordings are obtained from the parietal cortex of monkeys during a perceptual decision that is suspended by an intervening eye movement task. If successful, Aim 2 will forge a connection between the stability of vision across changes of gaze and the integrity of a decision across changes of intention. Aim 3 investigates a processing bottleneck that arises when two streams of evidence support two distinct decisions about a single object, that is, a double decision. Behavioral evidence from humans and monkeys indicates that sensory evidence can be acquired in parallel, but is incorporated sequentially into the double-decision. Aim 3 thus promises to elucidate the neural mechanisms of this serial incorporation and thus begin to explain why mental operations take the time they do. Together the proposed research will open new areas of computational and mechanistic interrogation of circuit interactions in the service of decision-making and cognitive control.