The neural computations underlying human social interaction recognition

项目摘要 感知和理解社会互动的能力对日常生活至关重要，并且在 自闭症通过简单的一瞥，我们可以毫不费力地识别出人们是否在互动， 是合作还是竞争，以及它的交际意图。然而，人们对神经基础知之甚少。 这些能力。我们最近发现，在后上级颞沟（pSTS）的区域，选择性 在观看社交互动时参与。这一发现，加上新的方法和建模， 的进步，创造了一个机会，调查社会互动的神经计算基础。我们 将测量功能磁共振成像，脑电图，和计算模型对控制和自然刺激的反应， 研究社会互动感知和理解的神经基础。我们的主要假设是， pSTS是一个社会的视觉和概念表示之间的关键计算连接点， 它通过两种不同的计算机制提取这些表示：自下而上 模式识别（来自身体和运动选择性大脑区域的视觉信息）与自上而下的认知 过程（基于来自心智理论网络的输入）。目标1将测试神经层次结构 从视觉原语到抽象概念的社会交互表征。使用条件丰富， 多模态功能磁共振成像实验，我们将测试的工作假设，社会交往的处理 沿着“第三视觉路径”分层：用身体和运动选择性视觉表示社会原语 区域，pSTS中社会互动的多模态表示，以及沿着pSTS的更高层次的社会特征沿着 STS和心理理论网络。目标2将确定整个社会互动中信息流的方向 网络通过将EEG记录与我们在Aim 1中获得的fMRI数据相结合，我们可以研究 信息在大脑区域之间流动，以确定社交互动的不同方面（从视觉到视觉）是否 从概念到概念）是以自下而上与自上而下的方式提取的。我们假设社会互动 检测和目标兼容性（即，合作与竞争）将通过自下而上的方式在pSTS早期进行编码 信息流从视觉区域。相反，我们假设其他社会评价将被代表 基于来自心理理论网络的额外输入，目标3将识别神经 社会互动表征的基础计算。我们会把我们的神经记录和底部的- 向上（判别式）和自上而下（生成式）计算模型，直接操作神经 上面概述的计算理论，以理解在社会互动中进行的计算 脑网络拟议的研究将为用于识别的神经计算提供新的见解 社交互动了解这些机制在典型的发展中的成年人是必不可少的第一步 来揭示这些计算在自闭症中是如何改变的。

Project Summary The ability to perceive and understand social interactions is crucial to daily life, and characteristically altered in autism. From a brief glance, we can effortlessly recognize whether people are interacting, whether the interaction is cooperative or competitive, and its communicative intent. However, little is known about the neural basis of these abilities. We recently identified a region in the posterior superior temporal sulcus (pSTS) that is selectively engaged when viewing social interactions. This discovery, coupled with novel methodological and modeling advances, creates an opportunity to investigate the neurocomputational underpinnings of social interactions. We will measure fMRI, EEG, and computational modeling responses to both controlled and naturalistic stimuli to investigate the neural basis of social interaction perception and understanding. Our central hypotheses are that the pSTS is a key computational junction between the visual and conceptual representations of a social interaction, and that it extracts these representations via two different computational mechanisms: bottom-up pattern recognition (from visual information in body and motion-selective brain regions) vs. top-down cognitive processes (based on input from the theory of mind network), respectively. Aim 1 will test for a neural hierarchy of social interaction representations from visual primitives to abstract concepts. Using a condition-rich, multimodal fMRI experiment, we will test the working hypothesis that social interactions are processed hierarchically along a ‘third visual pathway’: with social primitives represented in body and motion-selective visual regions, multimodal representations of social interactions in the pSTS, and higher-level social features along the STS and theory of mind network. Aim 2 will identify the direction of information flow across the social interaction network. By combining EEG recordings with our fMRI data from Aim 1, we can investigate the relative timing of information flow across brain regions to determine whether different aspects of a social interaction (from visual to conceptual) are extracted in a bottom-up versus top-down manner. We hypothesize that social interaction detection and goal-compatibility (i.e., cooperation vs. competition) will be coded early in the pSTS via bottom-up information flow from visual regions. In contrast, we hypothesize that other social evaluations will be represented significantly later based on additional input from the theory of mind network. Aim 3 will identify the neural computations underlying social interaction representations. We will compare our neural recordings with bottom- up (discriminative) and top-down (generative) computational models, which directly operationalize the neural computational theories outlined above, to understand the computations carried out across the social interaction brain network. The proposed studies will provide novel insights into the neural computations used to recognize social interactions. Understanding these mechanisms in typically developing adults is an essential first step towards uncovering how these computations are altered in autism.