A blog about programming (usually scientific python), mathematics (usually game theory) and learning (usually student centred pedagogic approaches).
Review of an arxiv preprint about student interaction with video
This short post will give a brief overview describing my thoughts about a preprint that appeared recently entitled "Examining the relationship between student performance and video interactions": arxiv.org/abs/1807.01912.
The paper tackles an interesting area of research and one that I was not fully aware of. One of the nice things about it, is the early paragraphs which highlight some details about the literature. As an example I was not aware of the work of Brinton and Chiang who in "MOOC performance prediction via clickstream data and social learning networks" apparently study relationship between how students interact with videos and perform on tests carried out immediately after watching the videos.
The main contribution of the paper considered here is that they study the relationship between student-video interaction and performance on assessment at a later date. This makes me think that it is attempting to offer a better understanding of the effect on actual retention (and possibly even learning).
"For each laboratory assessment, students were required to make a presentation as a way to demonstrate their understanding of the covered material. The presentations were then peer-evaluated and graded."
This is done by collecting data using the coursera system which is pretty cool as it seems to capture all sorts of details (for example: number of pauses etc...).
Finally two different linear regression models are applied and they conclude by saying that a linear model is unable to predict performance. The authors go on to describe a number of improvements that could be made.
This finding about interactions with video are interesting. I suggest they indicate that consumption of video is still a passive exercise and what indicates actual performance (or indeed learning) would be some measure of the active process that these videos scaffold.
I'd have two things I'd be interested in seeing:
- It would be cool if the data could be made openly available (perhaps, as this is a preprint, this will be taken care of by the journal).
- The pedagogic framework this course sits in, is described as very active: the videos are used to ensure class time can be spent doing worthwhile things. I wonder if it would be at all possible to supplement the data set with information about behaviour during those sessions (percent attendance, degree of participation etc...).
TLDR Very cool research, a great paper with cool results.main.md