Learning and Intelligent Systems: Neurophysiological, Computational, and Educational Studies of Sequence Learning and Cognitive Planning

该项目由学习和智能系统倡议资助。该项目旨在通过检查思考事件序列的能力来促进对大脑如何产生智能行为的理解。无论是做一顿精致的饭菜还是仅仅拨打一个电话号码，都必须记住特定时间顺序的多个事件，这是一种有助于规划复杂思想和行动的工作记忆形式。该项目将通过跨学科方法研究这一问题。为了直接探索大脑机制，将对清醒行为的动物进行神经生理学实验。基于计算机的幼儿实验将被用来发现儿童如何学习顺序行为，并测试如何优化这种学习。行为研究将对人类婴儿如何学习序列进行研究。认知和神经建模将用于发现大脑设计和机制，将动物神经生理数据与人类认知数据联系起来。 这种跨学科的方法有望产生超出任何单一方法范围的见解。在两种不同的动物物种（猴子和大鼠）、人类婴儿和幼儿身上进行实验，将有助于识别不同物种之间保存的智能机制。动物研究将分析大量单个神经元的活动，以研究多个神经元如何相互作用以产生智能行为。神经建模将探索支配这些行为的规律，并将做出将大脑与行为联系起来的预测。通过这些对我们如何学习和记忆事件序列的分析，将为研究高级认知操作（如规划和推理）的神经基础奠定基础;开发更好的教育软件;并将模型应用于解决需要模拟人类智能的算法的突出技术问题。

This project is being funded through the Learning and Intelligent Systems (LIS) Initiative. The project aims to advance understanding of how the brain generates intelligent behavior by examining the capacity to think about sequences of events. Whether cooking an elaborate meal or merely dialing a phone number, multiple events in a specific temporal order must be kept in mind, in a form of working memory that helps in planning complex thoughts and actions. This problem will be studied in the project through an interdisciplinary approach. To directly probe brain mechanisms, neurophysiological experiments will be performed on awake behaving animals. Computer-based experiments with young children will be used to discover how children learn sequential behaviors and to test how to optimize such learning. Behavioral studies will be done on how human infants learn sequences. Cognitive and neural modeling will be used to discover brain designs and mechanisms to link the animal neurophysiological data to the human cognitive data. This interdisciplinary approach promises to produce insights that are beyond the scope of any single approach. Conducting experiments in two different animal species (monkeys and rats), in human infants, and in young children will permit identification of intelligent mechanisms that are preserved across different species. The animal studies will analyse the activity of large groups of single neurons to study how multiple neurons interact to generate intelligent behaviors. Neural modeling will probe the laws that govern these behaviors, and will make predictions that link brain to behavior. Through these analyses of how we learn and remember sequences of events, a foundation will be built for studying the neural basis of high-level cognitive operations such as planning and reasoning; for developing better educational software; and for applying models towards the solution of outstanding technological problems that require algorithms which emulate human intelligence.