While the current deployment of learning analytics has its justification from the viewpoint of present stakeholders, it can be argued that it does not truly support personalised learning. Gašević, Dawson and Siemens (2015), therefore, suggest a deployment that acknowledges the following three widely accepted ‘truths’ of contemporary research in educational psychology postulated by Philip Winne:
- Learners construct knowledge
- Learners are agents
- Data include lots of randomness (2006, p. 1).
2.1 – Towards a student-controlled deployment of learning analytics
Building on these three axioms (Winne 2006) I argue that in higher education, data mining and analytics that emphasise personalised learning should:
- Improve the availability and approachability of resources that prompt students to question their perspectives on the world and their knowledge sources to support the development of understanding, critical thinking and creativity in complex systems (Knight, 2001).
- Adapt to each learner’s context without turning learning into an entirely self-sufficient, student-centred process that ignores the social, supportive and shared endeavour of education (Selwyn, 2016).
- Regard students as the main stakeholders of data-driven analysis (Slade and Prinsloo, 2013; Khalil, M; Ebner, 2015), acknowledge their sense of responsibility (Roberts et al., 2016) and put students in control of their data.
- Acknowledge that data deviations matter in education (Winne, 2006). The average does not represent the individual, and personalised recommendations should be based on individual learner-generated data, content and context.
- Protect students’ anonymity (Bachmann, Knecht and Wittel, 2017) and prevent stakeholders from inadequately labelling learners as good or bad students (Greller and Drachsler, 2012; Khalil and Ebner, 2015).
2.2 – Playground: ideas for the future
Based on these principles, I suggest a client-side deployment of learning analytics as an app that can be installed on personal student devices. This approach requires further investigations regarding the technical implementation, but it could have the following advantages:
- Informed consent: Students can actively opt-in or opt-out of learning analytics (Roberts et al., 2016). They can choose not to install the application, or close the app if they do not want to be monitored.
- Student ownership of data: Only students have access to their sensitive data and de-identification techniques such as anonymisation, masking or blurring (Khalil and Ebner, 2016) preserve the data’s original format while ensuring that the student’s identity is not revealed (Raghunathan, 2013).
- Semantic and serendipitous augmentation: A client-side system can follow a student from the learning management system to the browser, a PDF reader, a mind mapping tool, a word processor, and a reference manager if the student gives consent to access these applications. Access to content metadata within these applications allows applying semantic technologies that go beyond the current state of quantitative analytics (Siemens, Dawson and Lynch, 2013). Students can be directed towards peripheral and interdisciplinary information showing patterns, analogies and exceptions that challenge their thinking and augment the learning experience with potentially useful, and eventually serendipitous connections (Bawden, 2011; Makri et al., 2014).
- Learning history: Over time, the personal learning history allows to go beyond the context of a single course or educational programme by pointing out unnoticed connections to previous content, activities and assignments.
- Course-specific instructional conditions: Course-specific metrics (Gašević et al., 2016) can be retrieved from a learning management system and inform students about cognitive strategies and learning activities that aline with the pedagogical context and the desired learning outcomes of a course.
- Connecting agents: The system can encourage students to seek assistance if they are in academic distress or a potentially risky situation. Such support can be provided by teachers or by peers, who are working on the same learning resource. However, the student chooses if and what data and information are shared with other stakeholders.
- Teacher interventions: De-identified and aggregated data can inform teaching adjustments and course design improvements. Teachers can improve their feed-forward interventions by monitoring students’ engagement with learning activities and resources and identify critical points where they struggle or fail to proceed. Furthermore, the system can inform teachers about search terms, resources, or discussions that get increased attention or seem extraordinary and might require clarification.
2.3 – Field trip
nStudy: Philip Winne and his colleagues at Simon Fraser University are currently developing a web-based application for personal learning that uses data and analytics based on an agentic model of self-regulated learning. nStudy tracks student interactions with information, for example, if they browse web pages or take annotations and allows them to tag text or link text to specific terms or notes. The data and analysis are used to help students to better self-regulate their learning and provide data for learning science.
2.4 – ACTIVITY
- Can you find other concepts or tools designed for personalised learning that take into account some of the most pressing issues and challenges in the field of data and analytics?
- Video: Data, Analytics and Learning. Voices from the field: Phil Winne
Watch this short interview with Prof. Phil Winne. He talks about self-regulated learning, the current issues with educational data mining and introduces nStudy. What does Winne advise to use as a basis for making individual recommendations to learners and why is nStudy different from current examples of learning analytics deployment?
2.5 – Useful resources
- Journal article, Technology, Knowledge and Learning
Winne, P. H., Nesbit, J. C. and Popowich, F. (2017) ‘nStudy: A System for Researching Information Problem Solving’, Technology, Knowledge and Learning, 22(3), pp. 369–376.
