Collaborative Research: Fostering Collaborative Computer Science Learning with Intelligent Virtual Companions for Upper Elementary Students

越来越多的人认识到孩子们可以而且应该学习计算机科学。计算机科学的核心原则之一是它是一门协作学科，然而孩子们并没有与生俱来的协作能力。该项目将为小学高年级学生提供学习计算机科学和建立强大协作实践的机会。利用虚拟学习伙伴的承诺，该项目将解决三个重点问题。首先，该项目将收集不同小学高年级教室计算机科学协作学习的数据集。其次，该项目将设计、开发并迭代改进其智能虚拟学习伙伴，这些虚拟学习伙伴可以通过交互式在线编码工具在搭建的计算机科学学习环境中支持成对的学生。第三，该项目将产生关于儿童如何在计算机科学学习中合作，以及如何最好地支持他们与智能虚拟学习伙伴合作的研究结果和证据。将有三个系列的可交付成果：学习活动和专业发展，有虚拟学习伙伴的智能学习环境，以及进一步提高围绕计算机科学协作学习的学术和实践水平的研究证据。该项目将在北卡罗来纳州达勒姆县和佛罗里达州阿拉楚阿县两个州的高度多样化的小学中进行。该项目由探索研究PreK-12计划支持，该计划为STEM创新和方法的研究和开发提供资金。该项目解决了一个研究问题，“我们如何使用智能虚拟学习伙伴来支持小学高年级学生在计算机科学方面的学习和协作？”最初的数据集将通过在小学的计算机实验室中收集对话和解决问题的活动，为学生在课堂计算机科学学习任务中的合作方法提供一个基本的真实衡量标准。该项目将收集三角定性数据，以更好地了解围绕计算机科学二元学习的课堂动态影响。该项目的技术创新在于支持学生的方式：小学教室里的每对孩子都将与一组最先进的智能虚拟学习伙伴互动。这些伙伴将通过实时适应学生的协作和解决问题的模式，为这对学生提供量身定制的支持，从而增强课堂体验。虚拟学习伙伴将通过彼此之间的对话，为健康合作的关键方面建模，包括自我解释、提出问题、将挑战归因于任务而不是彼此的缺陷，以及通过吸收想法建立共同点。该项目将比较计算机科学学习的结果，以两种方式衡量：单独的前测试与后测试，以及合作产生的解决方案的质量。项目团队将通过目标协作策略的对话分析，以及计算机科学的兴趣和自我效能来度量协作实践。该项目将利用多层次模型设计来研究虚拟学习伙伴对学生成绩的影响。使用语音、对话记录、代码工件分析，以及手势和面部表情的多模态分析，团队将进行顺序分析，以确定对学生特别有益的虚拟学习同伴交互，并将我们的开发工作集中在扩展和细化这些交互上。他们还将确定学生没有参与的支持，并决定是否取消或重新设置它们。协作过程数据的分析将再次与定性的课堂数据相结合，这些数据来自实地笔记、焦点小组和对学生和教师的半结构化访谈。出现的主题将指导环境和学习活动的后续改进。

There is growing recognition that children can, and should, learn computer science. One of the central tenets of computer science is that it is a collaborative discipline, yet children do not start out with an intrinsic ability to collaborate. The project will provide the opportunity for upper elementary students to learn computer science and build strong collaboration practices. Leveraging the promise of virtual learning companions, the project will address three thrusts. First, the project will collect datasets of collaborative learning for computer science in diverse upper elementary school classrooms. Second, the project will design, develop, and iteratively refine its intelligent virtual learning companions, which support dyads of students in a scaffolded computer science learning environment with an interactive online coding tool. Third, the project will generate research findings and evidence about how children collaborate in computer science learning, and how best to support their collaboration with intelligent virtual learning companions. There will be three families of deliverables: learning activities and professional development, an intelligent learning environment with virtual learning companions, and research evidence that furthers the state of scholarship and practice surrounding the collaborative learning of computer science. The project will situate itself in highly diverse elementary schools in two states, Durham County, North Carolina and Alachua County, Florida. This project is supported by the Discovery Research PreK-12 program, which funds research and development of STEM innovations and approaches. The project addresses the research question, "How can we support upper elementary-school students in computer science learning and collaboration using intelligent virtual learning companions?" The initial dataset will provide a ground-truth measure of students' collaboration approaches to classroom computer science learning tasks through instrumenting computer labs in elementary schools for collecting dialogue and problem-solving activity. The project will collect triangulating qualitative data to better understand impactful classroom dynamics around dyadic learning of computer science. The technical innovation of the project is the way in which student dyads are supported: each pair of children within the elementary school classroom will interact with a dyad of state of-the-art intelligent virtual learning companions. These companions will enhance the classroom experience by adapting in real time to the students' patterns of collaboration and problem solving to provide tailored support specifically for that pair of students. The virtual learning companions will model crucial dimensions of healthy collaboration through their dialogue with one another, including self-explanation, question generation, attributing challenges to the task and not to deficits in each other, and establishing common ground through uptake of ideas. The project will compare outcomes of computer science learning as measured in two ways: individual pre-test to post-test, and quality of collaboratively produced solutions. The project team will measure collaborative practices through dialogue analysis for the target collaboration strategies, as well as interest and self-efficacy for computer science. The project will utilize a multilevel model design to study the effect of the virtual learning companions on student outcomes. Using speech, dialogue transcripts, code artifact analysis, and multimodal analysis of gesture and facial expression, the team will conduct sequential analyses that identify the virtual learning companion interactions that are particularly beneficial for students, and focus our development efforts on expanding and refining those interactions. They will also identify the affordances that students did not engage with and determine whether to eliminate or re-cast them. The analytics of collaborative process data will once again be augmented with qualitative classroom data from field notes, focus groups, and semi-structured interviews with students and teachers. The themes that emerge will guide subsequent refinement of the environment and learning activities.