Probing the neural computations underlying goal-directed decision-making in humans with single-neuron recordings

用单神经元探讨人类目标导向决策背后的神经计算 录音 MPI：John P.O‘Doherty博士和Ueli Rutishauser博士 项目总结 灵活的目标导向决策是高阶适应性生物智能的核心方面。一个数字 精神障碍涉及目标导向决策的损害，但目前甚至缺乏基本的 理解人类神经元水平的目标导向行为选择是如何实现的阻碍了我们的 准确定位这些神经精神功能障碍的能力。特别是，它完全不知道目标是如何实现的，以及 为了追求它们而需要从刺激和行动中选择的刺激和行动被表示在单个水平上 神经元，也不知道如何从可用的可能目标中选择目标。 在这里，我们将描述人类腹内侧额前核(VmPFC)、背侧 前扣带回(DACC)和前补充运动区(Pre-SMA)在这些突起中。我们将首先测试 长期以来的提议从未在人类神经元水平上测试过刺激的价值是特别的 VmPFC中的神经元代表动作的价值，而背侧皮质中的神经元更多地代表动作的价值 例如SMA前的领域。然后，我们将讨论目标本身是如何表示的。在真实的世界见面, 包括人类在内的动物在采取任何行动之前都需要选择一个目标。因此，有一种等级制度 先选择目标再选择行动的过程。我们假设vmPFC在 目标-评估和选择，而背内侧区域的神经元，包括SMA前和dACC，将发挥更多的 在评估和选择实现所选目标的行动方面发挥作用。大多数决策研究都集中在 无论是在行动上还是在刺激选择上，但不要首先解决目标是如何被选择的。我们将使用 定制行为任务，使我们能够区分这些不同的目标和与行动相关的计算 并通过透镜分析从这些大脑区域同时收集的单个神经元数据 计算强化学习模型。这项提议的意义在于，我们将第一次获得 时间，全面了解人类PFC在目标导向决策中的功能 细胞分辨率，同时揭示了目标表征和选择的神经机制 到目前为止，几乎没有人对此进行过研究。因此，拟议的项目在以下方面具有重要意义 将对理解不同的人类PFC亚区在 目标导向的决策。

Probing the neural computations underlying goal-directed decision-making in humans with single-neuron recordings MPIs: Dr. John P. O’Doherty and Dr. Ueli Rutishauser PROJECT SUMMARY Flexible goal-directed decision-making is a core aspect of higher-order adaptive biological intelligence. A number of psychiatric disorders involve impairments in goal-directed decision-making, yet the current lack of even a basic understanding of how goal-directed action selection is implemented at the neuronal level in humans hinders our ability to pinpoint these neuropsychiatric dysfunctions. In particular, it is completely unknown how goals, and the stimuli and actions that need to be selected from in order to pursue them, are represented at the level of single neurons, nor how goals get selected from available possible goals. Here we will characterize the functional contribution of human ventromedial prefrontal (vmPFC), dorsal anterior cingulate (dACC) and pre-supplementary motor area (pre-SMA) in these processes. We will first test the longstanding proposal never tested at the neuronal level in humans that the value of stimuli is especially represented by neurons in vmPFC, while the value of actions are more represented by neurons in dorsal cortical areas such as the pre-SMA. We will then address how goals themselves are represented. In the real world, animals including humans need to select a goal before any action is performed. Thus, there is a hierarchical process of goal selection followed by action selection. We hypothesize that the vmPFC plays a specific role in goal-valuation and selection, while neurons in dorsomedial areas including pre-SMA and dACC will play more of a role in valuing and selecting the actions that implement the chosen goal. Most decision-making studies focus either on action or stimulus selection, but don’t address how goals get selected in the first place. We will use bespoke behavioral tasks to allow us to distinguish between these different goal and action-related computations and analyze single neuron data simultaneously collected from these brain areas through the lens of computational reinforcement-learning models. The significance of this proposal is that we will gain for the first time, a comprehensive understanding of the functions of the human PFC in goal-directed decision-making at cellular resolution while shedding light on the neural mechanisms of goal-representation and selection which hitherto has been virtually unstudied. Consequently, the proposed project is highly significant in terms of the potential impact that will be made toward understanding the distinct role of different human PFC subregions in goal-directed decision-making.