STTR Phase I: Cloud-Based Pluggable Learning Analytics Engine for Educational Games

这个STTR第一阶段项目将进行基于云的可插拔数据分析引擎的研究和开发，以解决教育游戏市场？实时评估学习的需要。 如果从游戏中学习可以很好地量化，那么教育游戏将变得更加成功，这样买家就可以放心，使用游戏的时间是有成效的。然而，目前游戏制造商没有资格或资金提供此类分析所需的统计数据和认知评估。因此，该项目将建立一个商业可插入的第三方引擎原型，可以在不受干扰的情况下真实的跟踪学习者知识的增长，并为教育利益相关者提供定制的评估摘要和反馈。该原型将在代表不同人口群体的三所高中进行数据扫盲游戏的开发和测试。在商业环境中的测试将开始与两个成功的教育游戏公司合作。数据密集型评估技术的创新使用将通过在学习过程中提供精简和准确的信息，帮助美国目前苦苦挣扎的STEM教育。该项目还将帮助启动一项新业务，该业务有可能提升教育游戏和数字学习的市场价值。该STTR第一阶段项目采用了蒙特-卡罗贝叶斯知识追踪（MC-BKT）算法。该算法是最近在内部开发的，基于多年来在物理、教育和计算方面的研究所积累的技术，并首次实现了实时的个性化知识追踪。在之前的研究中，事后MC-BKT分析导致识别出多达七种与游戏片段中的知识增长相关的不同模式，与基于游戏屏幕和玩家话语的视频分析的人类判断相比，准确率为84%。该项目将进行研究，以测试MC-BKT算法的这种评估潜力是否可以在最初的研究之外扩展到涉及不同内容领域的游戏玩家，在更多的具有不同人口统计特征的教室（涉及约600名高中生）中，以及在真实的时间。基于研究成果，该项目将以基于云的可插拔引擎的形式围绕MC-BKT算法构建原型商业产品。两个流行的商业教育游戏以及该项目内部的各种游戏将被测试连接到引擎，用于实时测试知识追踪，学习问题检测以及向教师，家长，游戏设计师和学习者提供反馈。

This STTR Phase I project will carry out research and development on a cloud-based pluggable data analytics engine to address the educational game market?s need of real-time assessment for learning. Educational games will become much more successful if learning from games can be well quantified so that buyers will be assured that the time spent using games is productive. However, currently game makers are not qualified or funded to provide the statistics and cognitive assessment required for such analysis. This project will thus build a prototype of commercial pluggable third-party engine that traces the growth of the learner's knowledge in real time without interference and provides customized assessment summary and feedback to educational stakeholders. The prototype will be developed and tested with games that teach data literacy in three high schools representing diverse demographic groups. The testing in a commercial environment will begin in collaboration with two successful educational game companies. The innovative use of data-intensive assessment technology will aid in currently struggling STEM education in the United States by providing streamlined and accurate information while learning occurs. This project will also help launch a new business that has potential to boost the market value of educational games and digital learning.This STTR Phase I project utilizes the Monte-Carlo Bayesian Knowledge Tracing (MC-BKT) algorithm. This algorithm was recently developed in-house based on techniques distilled through years of research in physics, education, and computation, and makes it possible to perform individualized knowledge tracing in real-time for the first time. In prior research, post hoc MC-BKT analysis led to identification of up to seven distinct patterns associated with knowledge growth during game segments, with 84% accuracy as compared with human judgments based on video analysis of game screens and players' discourse. This project will conduct research to test whether this assessment potential of the MC-BKT algorithm can be extended beyond initial research to players with games involving different content domains, in a greater number of classrooms with diverse demographics (involving around 600 high school students), and in real time. Based on research results, this project will build a prototype commercial product around the MC-BKT algorithm in the form of a cloud-based pluggable engine. Two popular commercial educational games as well as various games internally sourced within this project will be test-connected to the engine for real-time testing of knowledge tracing, learning problem detection, and feedback delivery to teachers, parents, game designers, and learners.