Representations Yielding Task-Dependent Flexibility

描述(申请人提供)：本研究的主要目的是促进对认知灵活性发展的理解。孩子们可能会表现出惊人的失败，无法灵活地使他们的行为适应不断变化的环境。例如，孩子们经常坚持，当先前的行为不再合适时，他们会重复，即使面对重复和明确的信息来改变路线。相比之下，成熟的人类表现出独特的能力，能够灵活地使他们的行为适应不断变化的环境，即使没有明确的信息表明如何或何时这样做。发展文献中的许多工作都集中在这一过程的一个方面--儿童积极维持为他们提供的目标的能力不断增强。虽然这项工作使人们对支持儿童灵活行为的机制有了丰富的理解，但它在很大程度上未能解决发展中的两个关键转变。首先，在儿童在主动维护目标表征方面逐渐提高之前，他们表现出一种质的不同的、反应性的认知控制形式，只在需要的时候才提取目标。第二，即使儿童擅长维持为他们提供的目标，以引导外源性(外部提示)灵活性，但当他们必须为自己选择目标时，他们在内源性(内在驱动)灵活性方面表现出持续的限制和发展。这项拟议的研究建立在一个统一的、基于生物学的计算建模框架上，通过对儿童和神经网络模型的平行研究，调查这些不同组成部分的灵活性的发展。这一框架侧重于主动、抽象的知识表征在推动从被动控制到主动控制、从外源控制到内源控制的转变中所起的作用。这项工作将推进儿童如何在灵活行为服务中发展、维持和更新表征的明确机械模型，并将为提高灵活性的应用程序提供信息-从寻找婴儿到儿童规则的使用，再到提高自己驱动灵活行为的能力。

DESCRIPTION (provided by applicant): The primary goal of this research is to advance understanding of the development of cognitive flexibility. Children can show stunning failures to flexibly adapt their behavior to changing circumstances. For example, children often perseverate, repeating prior behaviors when they are no longer appropriate, even in the face of repeated and explicit information to change course. In contrast, mature humans show unique abilities to flexibly adapt their behavior to changing circumstances, even in the absence of explicit information about how or when to do so. Much of the work in the developmental literature has focused on one aspect of this process - children's increasing abilities to actively maintain goals that are provided for them. While this work has led to a rich understanding of the mechanisms supporting children's flexible behaviors, it has largely failed to address two critical transitions in development. First, prior to children's gradual improvements in proactive maintenance of goal representations, they show a qualitatively different, reactive, form of cognitive control, retrieving goals only as they are needed in the moment. Second, even after children excel at maintaining goals that are provided for them to guide exogenous (externally-cued) flexibility, they show continued limitations and developments in endogenous (internally-driven) flexibility, when they must select goals for themselves. The proposed research builds on a unified, biologically-based computational modeling framework to investigate developments in these diverse components of flexibility, through parallel studies with children and neural network models. This framework focuses on the role of active, abstract knowledge representations in driving transitions from reactive to proactive control, and from exogenous to endogenous control. This work will advance explicit mechanistic models of how children develop, maintain, and update representations in the service of flex- ible behavior, and will inform applications for improving flexibility - from infant search to children's rule use to increasing abilities to drive flexible behavior on one's own.