Understanding the dropout phenomenon has advanced from viewing it as a sign of deficient quality to viewing it as an explicit sign of individual choice, which highlights the importance of investigating how dropouts learn in massive open online courses (MOOCs). Nevertheless, the short, limited and heterogeneous behaviours of individual dropouts create challenges for understanding how dropouts learn over time. Taking a systematic network perspective, this study used clickstream data to build a flow network model of collective attention to investigate how dropouts learn in XuetangX's Introduction to Psychology (2018 autumn). The results showed that the quantification of behavioural data presented a stereotypical image of dropouts, but the network analytics presented a rather different picture of how dropouts learn. Recognising the distinct roles of introductory learning resources could prevent dropping out and improve the accuracy of prediction models. Interestingly, the assessments embedded in the MOOCs performed a scaffolding role in guiding dropouts to learn. Thus, redesigning quizzes or examinations in open and flexible learning environments to construct a minimum cost network of collective attention is vital to making this online space cost effective for learners at risk.

对辍学现象的理解已从将其视为质量不足的标志而发展为将其视为个人选择的明确标志，这突显了调查辍学如何在大规模开放在线课程（MOOC）中学习的重要性。然而，个别辍学学生的短暂，有限和异类的行为对理解辍学学生如何随着时间的学习提出了挑战。从系统的网络角度出发，本研究使用点击流数据建立了一个集体关注的流网络模型，以研究辍学者如何在XuetangX的《心理学概论》（2018年秋季）中学习。结果表明，对行为数据的量化给出了辍学的刻板印象，但是网络分析对辍学的学习方式却呈现出截然不同的印象。认识到入门学习资源的独特作用可以防止辍学并提高预测模型的准确性。有趣的是，嵌入在MOOC中的评估在指导辍学者学习中发挥了脚手架的作用。因此，在开放灵活的学习环境中重新设计测验或考试，以构建集体关注的最低成本网络，对于使这一在线空间对于处于风险中的学习者而言具有成本效益至关重要。