Collaborative Research: Precision Learning: Data-Driven Experimentation of Learning Theories using Internet-of-Videos

该项目旨在研究如何在 ASSISTments 系统的背景下帮助学生更有效地学习。 ASSISTments 是一个在线系统，可为学生提供帮助，并为教师提供实时评估数据。 ASSISTments 目前支持 100,000 名学生，他们已完成超过 1200 万道数学问题。该系统利用教师输入和人工智能为试图解决数学问题的学生提供帮助。该项目将通过结合视频建议（例如卡恩学院制作的视频建议）来增加教师和机器学习提供的帮助，以满足学生的需求。该实验将采用三本开放教育资源教科书的内容，这些教科书是公开许可并免费提供给学校的。更具体地说，研究人员将识别大量涉及教科书中数学技能的视频，并提取这些视频的特征，包括语言复杂性、语速和其他特征。这些视频和功能在呈现给学生之前将由教师通过 Mechanical Turk 流程进行可用性检查。此外，该项目还将开发一套用于精准学习的新技术，包括细粒度视频特征提取、从异构原始数据中学习学生特征、因果建模以及公平意识和因果关系增强的优化个性化推荐。该研究将增进对个性化学习相关基本问题的理论理解，并使数据驱动的学习理论实验成为可能。因果建模将使研究人员能够了解与学习效果相关的视频特征。该项目是美国国家科学基金会利用数据革命大创意活动的一部分，由本科教育部和学习研究部共同资助。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

This is a project to study what works to help students learn more effectively in the context of the ASSISTments system. ASSISTments is an online system that provides both assistance to students and real time assessment data to teachers. ASSISTments now supports 100,000 students who have completed more than 12 million mathematics problems. The system uses teacher input and artificial intelligence to provide assistance to students who are attempting to solve mathematics problems. This project will increase the assistance provided by the teacher and machine learning by incorporating video suggestions, such as those produced by the Kahn academy, targeted to the needs of the student. The experimentation will take content from three Open Educational Resource textbooks that are openly licensed and free to schools.More specifically, the researchers will identify a large collection of videos that address mathematics skills in the textbooks and will extract features of these videos including language complexity, speaking rate, and other features. These videos and features will be checked by both teachers and through a Mechanical Turk process for usability before they are presented to students. Additionally, the project will develop a suite of novel technologies for precision learning including fine grained video feature extraction, student feature learning from heterogeneous raw data, causal modeling, and fairness aware and causal relationship enhanced optimized personalized recommendation. The research will advance theoretical understanding of fundamental issues related to personalized learning and will enable data-driven experimentation of learning theories. Causal modeling will enable the researchers to learn the features of video that are correlated with learning effectiveness. This project is part of the National Science Foundation's Harnessing the Data Revolution Big Idea activity and is co-funded by the Division of Undergraduate Education and the Division of Research on Learning.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.