人工智能时代，在中小学开展人工智能课程是时代发展的要求，旨在培养学生高阶思维能力，提升信息素养。在众多教学形式中，协作学习更能适应新兴复杂的教学内容的要求。但在协作学习的研究中还存在一些问题，一方面，设计作为教育教学的一个重要因素，在协作学习的研究中往往被忽略，而且设计质量的优化不被重视；另一方面，在协作学习活动中，对设计质量的评价不够客观和全面，设计质量的优化也缺少数据依据。故需要在协作学习的研究中关注设计因素，关注多源客观的设计质量评价方法，同时关注数据驱动的设计质量优化。

本研究主要基于数据驱动的视角和以设计为中心的研究范式探索协作学习活动设计质量的评价指标体系和协作学习活动的优化策略。本研究在北京某初中的人工智能课堂中展开，设计了18个人工智能的协作学习活动，每个任务均分别在两个班开展第一轮和第二轮协作学习活动，共开展活动36次，并且每个班随机选择两个协作学习小组作为研究样本，共分析协作学习小组72个，历时9个月。本研究采用信息流的协作学习交互分析方法来分析交互过程，采用准实验法和访谈法验证优化策略的有效性。具体研究内容及结果如下：

首先，协作学习活动设计质量可从四方面进行评价，一是协作学习设计方案的评价，包括目标与任务的一致性、媒体多元性、目标设计的适应性、任务设计的适应性；二是设计与实施的一致性评价，包括知识点范围的一致性、知识建构程度的一致性、交互方式的一致性；三是协作学习交互过程的信息流属性评价，包括激活量、细化程度、聚焦程度；四是协作学习的结果性评价，即小组作品成绩。对各个指标分别设计计算方法，最终构建了一套多源、客观的协作学习活动设计质量的评价指标体系。

其次，本研究基于数据驱动的视角优化协作学习活动，优化的数据依据为评价指标的分析数据，从四个维度系统地提出了协作学习活动的优化策略。一是优化协作设计方案，优化策略包括优化任务设计、增加媒体资源类型、均衡不同难度的目标和任务数量等；二是提升设计与实施一致性，优化策略包括搭建知识类脚手架、关注遗漏的知识点、任务策略引导、明确角色职责等；三是优化在协作学习交互过程，优化策略包括搭建知识类和元认知脚手架、设置任务情景、搭建在线协同环境等；四是优化协作学习评价，优化策略包括制定评价规则、奖惩规则等。

最后，本研究通过比较两轮协作学习活动的评价数据、比较两个班的期末成绩以及对学生进行访谈的形式来验证协作学习活动的优化策略的有效性。结果显示第二轮的评价数据和期末成绩都显著高于第一轮，学生的访谈结果表明对协作学习活动的优化策略持肯定态度，整体表明数据驱动的协作学习活动的优化策略是有效的。

In the era of artificial intelligence, carrying out artificial intelligence courses in primary and secondary schools is a requirement of the development of the times, which aims to cultivate students' high-level thinking ability and improve information literacy. Among many instructional modes, collaborative learning can better meet the requirements of emerging and complex teaching content. But there are still some problems in the research of collaborative learning. On the one hand, the design is often overlooked in the study of collaborative learning, which is an important factor in education and teaching. And the optimization of design quality is not taken seriously. On the other hand, in collaborative learning activities, the evaluation of design quality is not objective and comprehensive, and the optimization of design quality also lacks data basis. Therefore, it is necessary to pay attention to design factors, multi-source and objective design evaluation methods in the study of collaborative learning, and data-driven design optimization.

This research aimed to explore the evaluation index system for design quality of collaborative learning activities and the optimization strategies of collaborative learning activities based on a data-driven perspective and a design-center research (DCR) method. This study was carried out in a real classroom teaching scenario of a junior high school artificial intelligence class in Beijing. A total of 18 collaborative learning activities for artificial intelligence were designed. Collaborative learning activities were conducted the first and second rounds in two classes, and a total of 36 activities were carried out. Two collaborative learning groups selected randomly in every class were regard as the research samples, and a total of 72 samples were collected in 9 months. The interactive analysis method of information flow collaborative learning was adopted to analyze the interactive process, and the quasi-experimental method as well as the interview method were adopted to verify the effectiveness of the optimization strategy. The results are as follows:

Firstly, the quality of collaborative learning activity design could be evaluated from four aspects. The first one is the evaluation of collaborative learning design schemes, including the consistency of goals and tasks, the diversity of media, the adaptability of goal design, and the adaptability of task design. The second one is the consistency of design and implementation, including the consistency of the scope of knowledge points, the consistency of the degree of knowledge construction, and the consistency of the interaction mode. The third one is the evaluation of the information flow attributes of the collaborative learning interaction process, including the amount of activation, the degree of refinement, and the degree of focus. And the last one is the evaluation of collaborative learning result, including the performance of the group work. The calculation method was designed for each index, and finally a set of multi-source and objective collaborative learning activity design quality evaluation index system was constructed.

Secondly, the collaborative learning activities were optimized from a data-driven perspective, and the data source was the analysis data of the evaluation indicators. The optimization strategies of collaborative learning activities were systematically proposed from four dimensions. The first one is optimizing the collaborative design scheme, and the optimizing strategies include optimizing task design, increasing the type of media resources, and balancing the goals of different difficulties. The second one is optimizing the consistency of design and implementation, and the optimizing strategies include building knowledge scaffolding, focusing on missing knowledge points, setting task strategy guidance, and clearing role responsibilities. The third one is optimizing the interactive process in collaborative learning, and the optimizing strategies include building knowledge and metacognitive scaffolding, setting task scenarios, and constructing online collaborative environments. The last one is optimizing collaborative learning results, and the optimizing strategies include formulating evaluation rules, setting reward and punishment rules.

Finally, whether the optimization of collaborative learning activities is effective were verified in this study. Comparing the evaluation data of the two rounds of collaborative learning activities, comparing the final grades of the two classes, and interviewing students were conducted. The results show that the indicator data and final grades in the second round were significantly higher than those in the first round. Interviews with students also showed a positive attitude towards the optimizing strategies in collaborative learning activities. All results showed that the optimization of data-driven collaborative learning activities was effective.