Thin Data, Thick Description: Modeling Socio-Environmental Problem-Solving Trajectories in Localized Land-Use Simulations.

Many learning technologies are now able to support both user-customization of the content and automated personalization of the experience based on user activities. However, there is a tradeoff between customization and personalization: the more control an educator or learner has over the parameters that define the experience, the more difficult it is to develop learning analytic models that can reliably assess learning and adapt the system accordingly. In this paper, we present a novel QE method for automatically generating a learning analytic model for the land-use planning simulation iPlan, which enables users to construct custom local simulations of socio-environmental issues. Specifically, this method employs data simulation and network analysis to construct a measurement space using nothing but log data. This space can be used to analyze users’ problem-solving processes in a context where normative measurement criteria cannot be specified in advance. In doing so, we argue that QE methods can be developed and employed even in the absence of rich qualitative data, facilitating thick(er) descriptions of complex processes based on relatively thin records of users’ activities in digital systems.