NCS-FO: Integrating Non-Invasive Neuroimaging and Educational Data Mining to Improve Understanding of Robust Learning Processes

从小学的数学游戏到职场培训，基于计算机的学习应用正变得越来越普遍。有了这些程序，越来越有可能使用生成的数据，例如正确和不正确的解决问题的回答，来开发测试学生知识的方法，并根据学生的需要进行个性化指导。学生回答的日志可以捕获答案，但它们无法捕获在学生与软件交互之间的暂停期间发生的事情的关键信息。该项目由亚利桑那州立大学和伍斯特理工学院的一组研究人员领导，将探索使用轻量级大脑传感器和学生日志数据测量大脑活动，以了解学习过程中重要的心理活动。该研究将使用ASSISTments智能辅导系统来检查大学和社区大学学生的发展性数学学习。利用脑成像技术，项目团队将检查学生在学习任务暂停期间是否在深入思考问题或走神，并使用日志和大脑数据相结合来预测学习结果。这项工作将为神经成像、机器学习和个性化学习环境相结合的新方法奠定基础。通过更好地了解在使用辅导系统时停顿时学习发生的时间和方式，学习技术研究人员和开发人员将能够在辅导系统中创建适应性干预措施，从而更好地个性化个人需求。该项目由理解神经和认知系统的综合策略（NSF-NCS）资助，这是一个由计算机与信息科学与工程（CISE）、教育与人力资源（EHR）、工程（ENG）和社会、行为和经济科学（SBE）联合支持的多学科项目。该项目有三个目标：1)整合多个数据流以创建跨学科语料库；2)检测日志数据暂停时认知状态的实时变化；3)通过基于大脑和基于日志的认知状态推断来预测学习结果。为了实现这些目标，该团队将使用功能性近红外光谱神经成像技术收集大脑数据，并从与规则学习和走神相关的受控、易于理解的任务中收集行为数据，以及从真实的学习任务中收集行为数据。涉及大脑活动记录的认知神经科学研究传统上需要具有高度受限刺激、时间和任务要求的范式，而在复杂的现实环境（如辅导系统）中的研究很少与这些范式一致。在认知任务中大脑活动的特征将被用来推断学生在真实学习任务中的认知。此外，大脑特征将与日志数据特征相结合，创建机器学习模型，准确预测学生的学习成果，并在学生使用交互式学习环境后进行后测。该项目对STEM学习的贡献将包括提高对学生如何在数字学习环境中响应教学事件构建知识的理解，构建更好的预测模型，预测学生何时从使用个性化学习环境中学习，以及绘制学习过程与暂停长度和上下文之间的映射。这个项目也将有助于理解如何结合神经成像数据和日志数据的分析。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。

From elementary school math games to workplace training, computer-based learning applications are becoming more widespread. With these programs, it becomes increasingly possible to use the data generated, such as correct and incorrect problem-solving responses, to develop ways to test for student knowledge and to personalize instruction to student needs. The logs of student responses can capture answers, but they fail to capture critical information about what is happening during pauses between student interactions with the software. This project, led by a team of researchers at Arizona State University and Worcester Polytechnic Institute, will explore the use of measurements of brain activity from lightweight brain sensors alongside student log data to understand important mental activities during learning. The study will examine developmental math learning in college and community college students using the ASSISTments intelligent tutoring system. Using brain imaging, the project team will examine whether students are thinking deeply about the problem or mind-wandering during pauses in the learning tasks and use the combined log and brain data to make predictions about learning outcomes. This work will build a foundation for new methods of combining neuroimaging, machine learning, and personalized learning environments. With a better understanding of when and how learning occurs during pauses in tutoring system use, learning technology researchers and developers will be able to create adaptive interventions within tutoring systems that are better personalized to the needs of the individual. This project is funded by Integrative Strategies for Understanding Neural and Cognitive Systems (NSF-NCS), a multidisciplinary program jointly supported by the Directorates for Computer and Information Science and Engineering (CISE), Education and Human Resources (EHR), Engineering (ENG), and Social, Behavioral, and Economic Sciences (SBE).This project has of three goals: 1) Integrating multiple data streams for the creation of an interdisciplinary corpus; 2) Detecting real-time changes in cognitive states during pauses in log data; and 3) Predicting learning outcomes from brain-based and log-based inferences of cognitive states. In addressing these goals, the team will collect brain data, using functional near-infrared spectroscopy neuroimaging, and behavioral data from controlled, well-understood tasks related to rule learning and mind wandering and from authentic learning tasks. Cognitive neuroscience research involving recordings of brain activity traditionally requires paradigms with highly constrained stimuli, timing, and task requirements, whereas research in complex real-world environments such as tutoring systems rarely align with these paradigms. Features of the brain activity during the cognitive tasks will be used to make inferences about student cognition during authentic learning tasks. In addition, brain features will be combined with log data features to create machine learning models that make accurate predictions of student robust learning outcomes, to be assessed using a posttest given after students use the interactive learning environment. Contributions of this project to STEM learning will include improved understanding of how students build knowledge in response to instructional events within digital learning environments, the construction of better predictive models of when students learn from the use of personalized learning environments, and a mapping between learning processes and the length and context of pauses. This project will also contribute to understandings of how to combine analyses of neuroimaging data and log data.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.