Longitudinal methods to analyse networked learning

Abstract

We learn from others by example, through observation, or simply by combining known concepts to yield new concepts. While doing so, we inherently connect to one another (connectivism). Learners are interconnected by learning relationships (‘X learns from Y'), but also by shared interests, similar actions, or shared resources. When these connections are aggregated, they form a learning network, and the act of participating in that network is called networked learning.

Networked learning can be analysed using social network analysis (SNA). SNA can detect structural characteristics of the network, communities or clusters, but also underlying characteristics of network actors (learners). In networked learning research, SNA is used in four ways: visualisation, analysis, simulation and intervention. However, the majority of approaches focuses merely on the visualisation and analysis of the network, rather than simulation and intervention, which can be of great value to networked learning research. Intervention has already taken off in the form of learning analytics (dashboards), and the actions that result from them. Simulation, however, may reveal the underlying mechanism that should be the main driver for intervention.

Learning network simulation can be used to predict networking behaviour by modelling the influence independent variables (e.g. actor characteristics) have on the dependent variable (e.g. network size). In such a case, one way to analyse a learning network is to use existing longitudinal network data to estimate a model that explains that influential behaviour. Simulation parameters vary along a range to create several combinations of input parameters, and are subsequently simulated numerous times (also known as Monte Carlo simulation) to yield a model that explains the behaviour of the dependent variable in terms of the explanatory variables. Other approaches use multilevel or regression analyses to create a model that explains the dynamic nature of the network.

The current paper shows the ways in which longitudinal network analysis can be used. That is, we provide examples of research questions, and how they can be addressed by longitudinal network analysis. Also, we supply practical guidelines to collecting data for analysis in an off-the-shelf program like RSiena. We include five data types that can be used as explanatory variables: constant actor variables, dynamic actor variables, constant dyadic variables, dynamic dyadic variables, and composition change indicators.