How can visual interactions support deep learning in geometry?

Self-explanation is a robust learning strategy, but automatic, scalable methods are needed to make it a practical strategy for large-scale implementation in classrooms. This study explored the effects of using visual interactions to engage students in self-explaining while they learned geometry using a computer-based intelligent tutoring system (ITS). The current study compared students who were asked to highlight diagram elements relevant to geometry principles during problem-solving against students who were not asked to highlight diagram elements. Verbal protocols generated during use of the ITS, as well as pre- and posttests targeting retention and transfer, were used to assess learning. Results showed that while the number of overall utterances did not differ across conditions, students who highlighted diagram elements produced a higher proportion of deep self-explanations that connected domain principles to problem diagrams and a lower proportion of shallow utterances that simply paraphrased diagram information (i.e., reading angles from the geometry diagrams). Shallow diagram utterances were negatively correlated with learning but deep diagram explanations were not correlated to learning. Thus, additional interactive elements may be needed to support successful self-explanation using visual interactions.