Artificial intelligence () has brought new ways for implementing learning analytics in computer-supported collaborative learning (CSCL). However, there is a lack of literature reviews that focus on AI-driven learning analytics applications and tools in CSCL contexts. To fill the gap, this systematic review provides an overview of the goals, characteristics, and effects of existing AI-driven learning analytics applications and tools in CSCL. According to the screening criteria, out of the 2607 initially identified articles between 2004 and 2023, 26 articles are included for final synthesis. Our results show that existing tools primarily focus on students’ cognitive engagement. Existing tools primarily utilize communicative discourse, behavioral, and evaluation data to present results and visualizations. Despite various formats of feedback are provided in existing tools, there is a lack of design principles to guide the tool design and development process. Moreover, although AI techniques have been applied for presenting statistical information, there is a lack of providing alert or suggestive information in existing tools or applications. Compared with the positive impacts on collaborative learning, our results indicate a lack of support for instructional interventions in existing tools. This systematic review proposes the following theoretical, technological, and practical implications: (1) the integration of educational and learning theories into AI-driven learning analytics applications and tools; (2) the adoption of advanced technologies to collect, analyze, and interpret multi-source and multimodal data; and (3) the support for instructors with actionable suggestions and instructional interventions. Based on our findings, we provide further directions on how to design, analyze, and implement AI-driven learning analytics applications and tools within CSCL contexts.

中文翻译：

---

计算机支持的协作学习中人工智能驱动的学习分析应用程序和工具：系统回顾


人工智能 () 为在计算机支持的协作学习 (CSCL) 中实施学习分析带来了新方法。然而，缺乏关注 CSCL 背景下人工智能驱动的学习分析应用程序和工具的文献综述。为了填补这一空白，本系统综述概述了 CSCL 中现有人工智能驱动的学习分析应用程序和工具的目标、特征和效果。根据筛选标准，在2004年至2023年期间初步确定的2607篇文章中，最终纳入26篇文章进行综合。我们的结果表明，现有工具主要关注学生的认知参与度。现有工具主要利用交流话语、行为和评估数据来呈现结果和可视化。尽管现有工具提供了各种格式的反馈，但缺乏指导工具设计和开发过程的设计原则。此外，尽管人工智能技术已应用于呈现统计信息，但现有工具或应用程序缺乏提供警报或提示信息。与协作学习的积极影响相比，我们的结果表明现有工具缺乏对教学干预的支持。本系统综述提出了以下理论、技术和实践意义：（1）将教育和学习理论整合到人工智能驱动的学习分析应用程序和工具中； （二）采用先进技术收集、分析和解释多源、多模态数据； (3) 通过可行的建议和教学干预为教师提供支持。 根据我们的发现，我们提供了有关如何在 CSCL 环境中设计、分析和实施人工智能驱动的学习分析应用程序和工具的进一步指导。