In the previous episode of Big Data in Education we argued that “wave of data” in education can shed light several dynamics of the learning process. In classical settings, learning takes place within the classroom actors (i.e. teachers and learners) and, for this reason, it is considered by many a black box to the outside world.
The use of Predictive Learning Analytics can change this by making learning more understandable for both teachers and learners and more accountable for the outer world. A data-driven approach can help to move away from using summative assessment as the only metrics for learning success. At the same time, if data is exploited correctly, can help to tailor formative and feedback-rich assessment which is generally more valuable than final grades.
The modelling step, the process of selecting the relevant attributes of the learning process and structuring them into a correct data representation, is not a trivial task. Learning is a complex and human process and its success depend on several endogenous (e.g. psychological states) and exogenous factors (e.g. learning contexts). The data-driven approach, however, works best whenever the data collection becomes continuous and unobtrusive by mean of sensors or trackers, which limits the scope of investigation only to measurable indicators, i.e. those attributes whose values are easy to measure over time.
Wong et al. (2012) assert that seamless learning takes place in ten different dimensions including Physical/Digital, Personal/Social, Formal/Informal, across time, across locations, with multiple devices or multiple learning methods.
In this post we focus on the first two dimensions, which in the context of Predictive Learning Analytics are rather critical:
- Physical/Digital – as modern learning takes place across physical and digital environments;
- Personal/Social – as learning is the result of both individual and a collective effort.
- Physical/Personal – segment of the physical activities, includes physiological responses which can be measured with biosensors. An example study is Learning Pulse which uses heart rate and step count to predict learning success. Ref Di Mitri (2016).
- Physical/Social – segment of the physical interactions, includes tangible interactions among individuals which can be measured by physiological and by contextual sensors. An example is .the one of Pijeira-Díaz (2016) which investigates the physiological responses of learners in collaborative learning environments.
- Digital/Individual – segment of the digital activity, includes all the relevant actions that individual learners do in a digital environment. There are many example studies about this as most of the time this segment is the only one observed in Learning Analytics applications.
- Social/Digital – segment of the digital interactions, includes all the digital social interactions (e.g. use of social media, collaboration tools). An example study for this is the Inquiry Based Learning application developed by Suarez.
If attributes are selected for each of those segment a fair representation of the learning process is achieved. Next step consists in defining a good set of success indicators which allow to go beyond-final-grades. We will talk about it in next episode.
Di Mitri, D., Scheffel, M., Boerner, D., Drachsler, H., Ternier, S., Specht, M. et al., 2016. Learning Pulse : using Wearable Biosensors and Learning Analytics to Investigate and Predict Learning Success in Self-Regulated Learning
Pijeira-Díaz, H.J., Drachsler, H. & Kirschner, P.A., 2016. Investigating collaborative learning success with physiological coupling indices based on electrodermal activity.
Wong, L.-H., 2012. A learner-centric view of mobile seamless learning. British Journal of Educational Technology, 43(1), pp.E19–E23.
Suarez, A., Ternier, S. Dojo-IBL https://dojo-ibl.appspot.com