基于多源学习活动流的分析规则设计及其应用

Design and Application of Analysis Rules based on Multi-Source Learning Activity Stream

多源学习活动流的提出,可有效反映学习者穿梭于不同学习生态空间的数字活动轨迹,而将这些数据汇聚在一起加以追溯和分析,可全面理解学习者整个学习生态的需求和问题,从而有效支持智慧教育下的精准教与学。在借鉴ADL xAPI的活动描述模型、CAM的分层结构和Paradata的多用户视角的基础上,多源学习活动流的描述模型以及汇聚层次被加以讨论,补充了在以往学习分析研究中少有考虑到的数据记录的多源性以及汇聚分析的层次问题。首先,六类数据要素(Actor、Action、Object、Tool、Session和Source)逐渐递增组合、22条"情境-应用-数据"数据分析规则被提出,并按照个体性情境、任务性情境、社会性情境、时空性情境和环境性情境加以归类,以指导多源学习活动流数据分析的具体应用。然后,借助由Slack(一款App聚合协作学习平台)和Trello(一款学习项目管理平台)所搭建的多源学习环境,将提出的数据分析规则应用于实践:实验对象35名大二学生被分成10组,所收集的6179条学习活动行为数据使用Actor{1,n}-Action{1,n}/Tool{1,n}-Source等规则,从小组和个人的不同层次,解读学习者的行为模式偏好和学习状态变化,初步验证了所提分析规则的可用性。最终得到的研究成果指向解决今后必然越发复杂的数字学习生态中的学习分析问题,以期为其他研究者和教育一线工作者,提供一定的分析思路引导和框架指南。

The multi-source learning activity stream can effectively reflect the digital activity trajectory of the learner in different learning spaces. The data of multi-source learning activity stream can be collected,traced and analyzed to fully understand learners’ needs and problems from the perspective of the entire learning ecology,so as to support precision teaching and learning under smart education. Based on the activity behavior description model of ADL xAPI,the hierarchical structure of CAM and the multiuser perspective of Paradata,the description model and the convergence level of multi-source learning activity stream was discussed.Six types of data elements(including actor,action,object,tool,session and source) were gradually combined. From individual context,task context,social context,spatiotemporal context and environmental context,22 "context-application-data" data analysis rules were proposed. These rules establish the mapping relationship between data and context,and can be used to guide the specific applications. Then,with the multi-source learning environment built with Slack and Trello,the application practice of the data analysis rules was carried out to reveal learners’ behavior preferences and learning state changes. 35 sophomores were divided into 10 groups and the records of their learning activities were 6179. The data was analyzed using the proposed analysis rules,such as Actor {1,n}-Action{1,n}/Tool{1,n}-Source,from the levels of groups and individuals,and the results preliminarily verified the usability of the proposed analysis rules. The research points to solving the problem of learning analysis in the digital learning ecology that is bound to become more complicated in the future,and hopes to provide certain idea and guidance.