CRII: Cyberlearning: Automatic Discovery of Optimal Progressions of Language Content

教育中的一个关键挑战是确定向学生呈现概念的最佳顺序。这对于学习第二语言尤其重要，因为要学习的内容非常多，学习者很难保持专注。尽管教育和心理学研究者已经确定了排序学习材料的指导原则，但缺乏实用的技术来应用这些原则来自动构建有效的学习进度。为了解决这一差距，本项目将：(1)开发估计语言学习内容难度的技术，(2)确定各种专家制作的进度的重要特征，以及(3)通过实验找到适合个别学生的最佳进度。这些进步将提高免费向公众开放的在线互动学习工具的有效性。预期的步调一般原则的发现将使教育者受益，并为最大限度地参与适应性学习系统提供基础。该项目将进行三项调查，以建立适应性学习进程。首先，该项目将建立一个按难度组织的语言资料库。为此，PI将探索目标语言知识（如语法）的程序模型，确定哪些程序步骤是理解特定材料所必需的，并将这些材料组织成技能树。其次，该项目将开发一套控制参数，以捕获节奏和顺序的关键方面。这将允许PI从专家制作的进程中提取节奏策略，并将这些策略映射到其他进程中。第三，该项目将通过在线互动学习工具进行大规模多元实验，优化个人学习者的进度。

A key challenge in education is determining the optimal sequence of concepts to present to a student. This is particularly important for second language learning, where there is a huge amount of content to learn and it is difficult for learners to stay engaged. Although researchers in education and psychology have identified guiding principles for sequencing learning materials, there is a lack of pragmatic techniques that can apply these principles to build effective learning progressions automatically. To address this gap, this project will: (1) develop techniques for estimating the difficulty of language learning content, (2) identify important characteristics of various expert-crafted progressions, and (3) experiment to find optimal progressions for individual students. These progressions will enhance the effectiveness of online interactive learning tools that are freely accessible to the public. The anticipated discovery of general principles of pacing will benefit educators and provide the foundation for maximizing engagement in adaptive learning systems.The project will conduct three investigations in order to build adaptive learning progressions. First, the project will build a library of language materials that are organized by difficulty. To do this, the PI will explore procedural models of target linguistic knowledge such as grammar, identify which procedural steps are necessary for understanding specific materials, and organize these materials into skill trees. Second, the project will develop a set of control parameters that capture key aspects of pacing and sequencing. This will allow the PI to extract pacing strategies from expert-crafted progressions, and map these strategies onto other progressions. Third, the project will optimize progressions for individual learners by conducting large-scale multivariate experiments in online interactive learning tools.