Our final sessions for LAK began with a session on Supporting Learning and Achievement, starting with a discussion on how we can help learners frame their learning experiences to achieve greater success. This first presentation described two studies, one explored a range of NLP-based analyses to try to identify clusters of learners that behaved in different ways. They conducted an analysis of student text contributions to see whether they could identify sensible intervention points. The second explored plan making in MOOCs, where students in the experimental group were asked to develop written plans, and contrasted with a control group, across three courses from different disciplines, with a positive result that more students in the experimental group achieved a completion certificate. They explored the open text of the plans, and found predictors in terms of the maturity and complexity of their plans with their overall course achievement.
This is an exciting result, as I can see potential for planning strategies to address several areas of concern in MOOCs. An earlier discussion of MOOS at L@S identified that few courses explicitly talk about how learners should communicate and collaborate – it might be possible to bring in specific planning strategies for this. Further, we know from our own work in teacher professional development, and the work of others, that professionals in MOOCs have trouble transferring course learnings to their professional setting – perhaps this is also an area that could be assisted with planning? In our teacher professional learning MOOCs, we explicitly model the professional setting, and embed activities that merge the course and professional setting. This certainly does help in address that latter problem, however I think this could be an interesting way of improving this further.
The second paper explores what self-regulated learning strategies are used by students in MOOCs to engage in online resources that we can encourage. Their previous work has found a higher correlation between online interactive practice (‘doing’ practices, such as MCQs, drag and drop, and other interactive activities), and successful outcome, than other forms of online activity, such as video watching and browsing. Overall, they’ve found a correlation between online activities, particularly online activities with feedback, with learner success.
The final talk explored how we can trigger higher order thinking processes in learners within a MOOC context. They base their work in the ICAP framework – more learning occurs in constructive modes, where they are both active and generating new outputs. Using text contributions from a MOOC course, and in course access data, they developed a manual ICAP-based coding manual, to see if coded higher order thinking behaviours was associated with more learning. They found similar results to the earlier paper, in that students who are more deeply engaged with online materials, including discussion forums, learn more. The authors also explored a further study, that used topic modelling to see what topics generate most discussion in the forum, which can be used to help identify what kinds of materials generate opportunities for discussion and learning.
This reminded me of an interesting short paper presentation from L@S, which explored some initial work on whether students who share URLs in their discussion forum posts are different learners, or generate different kinds of discussion. I thought this would be worth exploring further, particularly using sentiment analysis and SNA to explore clustering in discussion threads.
Forecasting Student Achievement in MOOCs with Natural Language Processing,
Is the Doer Effect a Causal Relationship? How Can We Tell and Why It’s Important,
Towards triggering higher-order thinking behaviors in MOOCs,