Most of the time, I study human interactions in online settings using network methods. Often that implies being asked about the relationship between learning and social interactions because ‘I can also learn alone!’. Over years I grew high intolerance to this question, but having been asked it at a job interview lately (I opted out of the job) I pondered around the question of ‘What outcomes of learning result from social interactions?’ Knowledge construction and idea generation is an obvious one, but I was interested in the others as well. As the result of this exercise, I am working on a paper that explains why interpersonal relations are a powerful drive for learning conceived more broadly. The idea is to elaborate what can be called ‘a learning capital’, or the potential to learn with and from others, with learning (outcome) being a variety of constructs, others being a variety of subjects and artefacts, and the mechanisms of ‘learning’ ranging from mimicry to deliberate adoption.
In my view, when it comes to human interactions specifically, research of this ‘learning capital’ boils down to the following research problems:
- Why do learner networks form?
- What does an effective learning network look like?
- What drives a formation of an effective learning network?
- What peer effects on learning can we observe?
Each of these three is a huge area of work on its own, and each (in my vision of where this goes) should result into interventions with network structure, selection based on individual (node) attributes, instructional activities, and technology that allows for all of this to be seamlessly manipulated and affords interaction to start out with.
Besides being a vast and deep research question, each of these require rather sophisticated methodologies tapping into causal claims and statistical methods that allow for inferences from interrelated non-random observations. I am happy that all the conversations I had with learning analytics researchers who work with networks have been about how ‘we can not do this with this model’. To me it is a sign that we are thinking as problem-driven, not method-driven.
The choice of methods for me is down to a set of econometrics models and network modelling. I also see agent-based modelling as an essential part of the methods pipeline, prior to the actual experiments in the classroom to manipulate the variables that interact towards the emergence of learning in networks.
To start asking better questions and apply better methods, some folks at LAK have talked about an open call for a special issues around networked learning and learning analytics. We are still refining the focus but it is definitely coming.