Neuronal substrates underlying the construction of value in humans

我们的个人偏好是如何产生的？我们都有各种各样的偏好和个人品味，但为什么我们会喜欢一些东西而不喜欢另一些呢？例如，为什么一个人喜欢薯片而不喜欢爆米花，喜欢流行音乐而不喜欢西部乡村音乐，喜欢棒球而不喜欢篮球，或者喜欢伦勃朗的具象画而不喜欢杰克逊·波洛克的抽象现代艺术？我们在这个项目中要解决的主要问题是我们的偏好是如何在大脑中形成的？研究人员的目的是测试一种理论，即大脑依赖于物体的“特征”，即构成物体的元素——比如食物中的成分，或者一幅画中的颜色、形状和纹理——来决定有多喜欢或不喜欢某件东西。假设是，脑细胞（神经元）对物体的这些基本特征做出反应，而其他神经元将这些特征整合起来，形成一个整体的偏好判断。为了验证这一点，研究人员直接从人脑中记录下神经元的反应，同时人们对各种物品（包括食物、视觉艺术甚至衣服）的喜欢程度做出简单的判断。这项研究之所以成为可能，是因为脑外科手术的安全技术进步，以及正在接受治疗的难治性癫痫患者的慷慨。在脑外科手术切除癫痫组织之前，作为治疗的一部分，患者的大脑中暂时放置了小电极来定位癫痫发作。这些患者通常愿意自愿参加研究，这就提供了一个难得的机会，可以直接测量神经元的电活动。了解我们大脑中的神经元如何形成我们的偏好，可以帮助我们更好地理解人们在现实世界中是如何做出决定的。虽然我们知道价值信号是在大脑中编码的，但一个基本问题仍然存在，即这些信号最初是如何被计算出来的。在这个项目中要测试的主要假设是，刺激的价值信号，从食物到消费品再到艺术，都是由大脑以动态的方式通过整合刺激的组成特征或属性来构建的。虽然许多先前的研究已经使用神经成像技术（如功能磁共振成像或fMRI）检查了价值构建，但只能通过该技术的时空分辨率获得见解，而该技术不能测量单个细胞。因此，对于刺激特征是如何由单个细胞表示并整合到计算整体刺激值的，我们知之甚少。在这里，这一限制可以通过在人类癫痫患者执行三种不同任务探测价值构建时使用颅内记录来克服。这项研究的重点是大脑中价值网络的三个关键部分的神经元群：杏仁核、外侧眶额皮层和腹内侧前额叶皮层。这些方法的空间分辨率使我们能够深入了解单个特征是如何在神经元水平上表示的，而时间分辨率使我们能够识别时间动态，通过这些动态，特征被主动整合以产生总体价值信号。杏仁核和外侧眶额皮质的特征被假设为整合在腹内侧前额皮质中产生价值信号。此外，该研究旨在解决随着总体目标的变化而改变价值信号的手段，将特征直接由环境变化调节的模型与特征神经元和价值编码神经元之间的连接由目标环境调节的替代模型进行比较，但特征表征保持不变。这个项目有望对人类大脑中价值信号产生的神经计算，以及我们个人偏好的起源和神经基础，以及指导人类决策的基本作用，提供基本的机制见解。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

How do our personal preferences arise? We all have a multitude of preferences and a sense of our own individual taste – but why do we come to like some things and not like others? For instance, why might a person like potato chips but not like popcorn, pop music rather than country-western, baseball rather than basketball, or come to like a representational painting by Rembrandt, but dislike abstract, modern art of Jackson Pollock? The major question we are addressing in this project is how are our preferences formed in the brain? The researchers aim to test a theory that the brain relies on “features” of objects which are the elements that make up an object – such as the ingredients in a food item, or the color, shapes and textures in a painting, in order to make a decision about how much something is liked or not. The hypothesis is that the brain cells (neurons) respond to these elementary features of an object, and that other neurons integrate over these features to form an overall preference judgment. To test for this, the responses of neurons are recorded directly from the human brain, while people make simple judgements about how much they like a variety of items including food, visual art and even clothing items. This research is made possible because of safe technological advances in brain surgery, and the generosity of patients who are being treated for intractable epilepsy. Prior to brain surgery to remove the epileptic brain tissue, patients have small electrodes temporarily placed into their brain to locate their seizures as part of their treatment. These patients are often willing to volunteer to take part in research studies, yielding a rare opportunity to measure electrical activity from neurons directly. Understanding how neurons in our brain form our preferences could help us better understand how people make decisions in the real world.While value signals are known to be encoded in the brain, a fundamental question remains about how such signals come to be computed in the first place. The major hypothesis to be tested in this project is that value signals for stimuli ranging from foods to consumer goods through to art, are constructed by the brain in a dynamic fashion by integrating over the component features or attributes of a stimulus. While a number of prior studies have examined value construction using neuroimaging techniques (such as functional magnetic resonance imaging or fMRI), insights can only be gained over the spatiotemporal resolution of the technique, which cannot measure single cells. As a consequence, little is known about how stimulus features are represented by individual cells and integrated to compute an overall stimulus-value. Here, this limitation can be overcome through the use of intracranial recordings in human epilepsy patients while they perform three different tasks probing value construction. The research focuses on neuronal populations in the three key parts of the value network in the brain: the amygdala, lateral orbitofrontal, and ventromedial prefrontal cortex. The spatial resolution of these methods enables insight into how individual features are represented at the neuronal level, while the temporal resolution allows us to identify the temporal dynamics by which features are actively integrated to yield an overall value signal. Features in the amygdala and lateral orbitofrontal cortex are hypothesized to be integrated to yield value signals in the ventromedial prefrontal cortex. Furthermore, the research aims to address the means by which value signals are altered following a change in one’s overall goals, comparing a model in which features are modulated directly by a change in context, against an alternative in which the connectivity between feature neurons and value coding neurons are modulated by goal context, yet feature representations remain unchanged. This project promises to yield fundamental mechanistic insights into the neural computations underlying the production of value signals in the human brain, and the origin and neural basis of our individual preferences, which in turn serve a fundamental role in guiding human decision-making.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.