Developing Computational Methods to Measure and Track Learner's Spatial Reasoning

Interactive learning environments can provide learners with opportunities to explore rich, real-world problem spaces, but it can be hard for educators and educational designers to (a) understand what students are “up to”, and to subsequently (b) provide guidance or feedback to help learners make progress. Educational Data Mining (EDM) offers the potential to help diagnose student activities within an interactive learning environment, but it has historically been applied to constrained and fairly well-understood problem spaces. This work represents part of a growing body of research that is applying EDM techniques to more open-ended problem spaces. The open-ended problem space under study here was an environmental science simulation, where learners were confronted with the real-world challenge of figuring out where it is effective to place green infrastructure in an urban neighborhood so as to reduce surface flooding. Learners could try out many different arrangements of green infrastructure and use the simulation to test each solution’s impact on flooding. The solutions proposed by the learners were logged during a series of experimental trials with different user interface designs for the simulation. Analyzing this data was difficult due to the large possible solution state space, and because there are many possible good solutions and paths to discover good solutions. This work proposes a procedure for reducing the state space while maintaining critical spatial properties of the solutions. Spatial reasoning problems are a class of problems not yet examined by EDM, so this work will set the stage for further research in this area. This work also demonstrates a procedure for discovering effective spatial strategies and solution paths, and demonstrates how this information can be used to give formative feedback to the designers of the interactive learning environment.