The mutual influence of attention and learning during knowledge acquisition

描述(由申请者提供)：有效的学习包括提取关键的信息模式来捕捉我们经验的本质，并使用这些信息来构建有用的知识，使其能够在新的情况下预测行为。人类有多个学习系统与不同的大脑区域相关联，但这些学习系统如何相互作用还不是很清楚。此外，现有的理论在很大程度上忽略了信息采样行为是如何在学习过程中受到经验和目标的指导的。本研究将从一个新的理论视角，结合功能磁共振成像(FMRI)和眼动跟踪来研究注意和学习的机制以及它们在类别学习过程中的相互作用。关键的假设是，在学习过程中，个人必须选择要采样的信息，这将受到个人的知识和当前目标的指导。将开发一类计算模型，将类别学习重新塑造为一个动态的决策过程，在该过程中，注意力作为信息处理出现，由学习者的目标、能力限制和当前知识引导。这些模型通过展望未来并评估预期利润的未来行动来计划一系列行动(例如，眼球运动)。实验一将检验这些模型的关键预测，即类别学习中的顺序抽样行为是通过注意和知识成分的交互作用来调节的。将这些模型与完美和不完美学习者的眼动和分类行为相适应，将表征促进最佳类别学习的信息采样和学习过程的本质。在实验2和实验3中，模型组件将与类别学习中涉及的特定神经生物学学习系统联系起来。实验2将开发和验证一种将正式模型与学习大脑联系起来的新方法，方法是使用fMRI数据的多变量大脑模式来 在相互竞争的认知模型中进行裁决。这种方法的关键推理是，如果一个模型代表了类别学习的真实本质，那么该模型在学习过程中状态的连续测量应该反映在学习大脑的逐次尝试测量中。实验3将使用这种新的模型选择方法来表征前额叶、腹侧纹状体和内侧颞叶之间的相互作用如何支持成功的类别学习。了解支持注意力和知识的神经生物学机制将提供一种预测和促进有效学习的手段。此外，这项工作有可能为诊断工具的开发提供信息，这些工具可以准确地表征表现出学习缺陷的临床人群的认知障碍，例如精神分裂症、抑郁症和阿尔茨海默病患者。

DESCRIPTION (provided by applicant): Effective learning involves extracting key patterns of information that capture the essence of our experiences and using this information to build useful knowledge that enables predictive behavior in novel situations. Humans have multiple learning systems associated with different brain regions, yet how these learning systems interact is not well understood. Also, existing theories largely ignore how information sampling behaviors are guided by experience and goals during learning. The research presented in this proposal will use a novel theoretical perspective in combination with functional magnetic resonance imaging (fMRI) and eye tracking to investigate the mechanisms of attention and learning and their interactions during category learning. The key hypothesis is that during learning, individuals must choose what information to sample, which will be guided by the person's knowledge and current goals. A class of computational models will be developed that recasts category learning as a dynamic decision process in which attention emerges as information processing guided by the learner's goals, capacity limitations, and current knowledge. These models plan a course of action (e.g., eye movements) by looking ahead and evaluating future actions for expected profit. Experiment 1 will test the key prediction of these models that sequential sampling behavior during category learning is mediated by the interaction of attention and knowledge components. Fitting these models to eye movement and classification behavior from perfect and imperfect learners will characterize the nature of information sampling and learning processes that promote optimal category learning. In Experiments 2 and 3, model components will be linked to the specific neurobiological learning systems implicated in category learning. Experiment 2 will develop and validate a novel method of linking formal models to the learning brain by using multivariate brain patterns of fMRI data to adjudicate among competing cognitive models. The key reasoning of this method is that if a model represents the true nature of category learning, continuous measures of that model's states during learning should be reflected in trial-by-trial measures of the learning brain. Experiment 3 will employ this novel model selection method to characterize how interactions between prefrontal cortex, ventral striatum, and the medial temporal lobe support successful category learning. Understanding the neurobiological mechanisms that support attention and knowledge will provide a means of predicting and promoting effective learning. Moreover, this work has the potential to inform the development of diagnostic tools that precisely characterize cognitive impairments in clinical populations that exhibit learning deficits, such as individuals with schizophrenia, depression, and Alzheimer's disease.