SourceWrite: Real-time, biometric, intention-informed scaffolding of source-based writing processes

写作是一项复杂的任务，它包括几个组成过程：阅读源材料，设定目标，计划内容，将想法转化为语言，阅读已经写好的文本，编辑，等等。学生使用的过程，以及顺序，影响他们的书面作文的质量。到学生上大学的时候，大多数人都会发展出自己的写作过程。这些措施的效果各不相同。当面对要求严格的学科写作任务时，特别是那些需要综合多种来源的写作任务时，学生的既定写作过程往往是次优的。这对于攻读科学、技术、工程和数学（STEM）学位的学生来说尤其令人担忧。必修的大学水平作文课旨在帮助学生提高写作技能。然而，在这些课程中，学生通常只收到关于他们已经写过的文本的反馈，而不是关于他们写作时使用的过程。这是因为写作教师无法接触到学生写作的每时每刻。在这个项目中，研究人员将开发一个名为“SourceWrite”的智能辅导系统，它将自动跟踪学生在写作过程中所做的事情，推断他们为什么这样做，然后提供个性化的建议和帮助，所有这些都是在学生仍在写作过程中的真实的时间内完成的。具体来说，研究人员将开发自动写作过程分析方法，将联合收割机生物统计数据（写作时间和眼球运动）与自然语言处理相结合，以推断学生在写作过程中的意图。这些方法将允许自动，实时预测写作过程模式以及这些模式将如何影响文本的最终质量。这将在真实的时间内实现，在文本组成期间，在文本已经完全产生之前。为了实现这一目标，该项目将把（数据驱动的）写作分析研究与（理论驱动的）文本生成心理语言学结合起来，这两个方向传统上是分开进行的。学习和教学创新将在设计，实施和评估一种新的教育干预，将提供智能支持，学生，因为他们从事与他们的来源和生产学术文本，在大学作文课程的背景下。通过一系列基于设计的研究迭代，随后进行随机对照评估，该项目将为这种新的教学法建立设计原则，并确定其对培养大学生写作能力的有效性。该奖项反映了NSF的法定使命，并被认为值得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来支持。

Writing is a complex task which comprises several component processes: reading source materials, setting goals, planning content, translating ideas into language, reading already-written text, copyediting, and so forth. Which processes a student uses, and in what sequence, affects the quality of their written composition. By the time students reach college, most will have developed their own individual mixture of writing processes. These will vary in effectiveness. When faced with demanding disciplinary writing tasks, especially those that require synthesizing multiple sources, students' established writing processes often turn out to be suboptimal. This is a particular concern for students studying for Science, Technology, Engineering, and Mathematics (STEM) degrees. Required college-level composition classes are designed to help students improve their writing skills. However, in these classes, students usually receive feedback only about the texts they have already written, not about the processes they use when they write. This is because writing instructors do not have access to the moment-by-moment actions by which students' texts are produced. In this project, the researchers will develop an intelligent tutoring system called "SourceWrite" that will automatically track what the student is doing during the composition process, infer why they are doing it, and then provide individualized advice and assistance, all in real time while the student is still in the process of composing their text.Specifically, the researchers will develop methods for automatic writing-process analysis that will combine biometric data (keystroke timings and eye movements) with natural language processing to infer the student's intentions during composition. These methods will permit automatic, real-time predictions about writing-process patterns and how these will affect the ultimate quality of the text. This will be achieved in real time, during text composition, before the text has been fully produced. To achieve this end, this project will bring together research in (data-driven) writing analytics with (theory-driven) psycholinguistics of text production, two directions that have traditionally been followed separately. The learning and teaching innovation will be in designing, implementing, and evaluating a novel educational intervention that will provide intelligent support to students as they engage with their sources and produce academic text, in the context of a college composition course. Through a series of design-based research iterations followed by a randomized, controlled evaluation, this project will establish design principles for this new pedagogy and determine its effectiveness for developing college students' writing ability.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.