Evaluating an Intelligent Sketching Feedback Tool for Scalable Spatial Visualization Skill Training

Spatial visualization skills are essential and fundamental to studying STEM subjects. The increasing need for STEM education poses scalability challenges in spatial visualization skill training. Many researchers and practitioners face a major challenge of supporting sketching, an essential component in spatial visualization training, at scale. Because of the enormous error space, traditionally, a significant amount of human effort is required to grade and provide individualized feedback for students' technical drawings. Our team leveraged data mining and unsupervised learning techniques to build an intelligent sketching feedback tool. The tool not only allowed students to practice their sketching skills in a scalable manner but also graded students' sketches and provided customized and actionable feedback based on the error patterns in real-time. We deployed our tool in a university-level spatial visualization class with about 60 students. Students interacted with our tool for eight weeks. We performed an interview study to understand students' experience and attitudes towards using such an automated feedback tool for practicing spatial visualization skills. Through a grounded theory approach, we identified themes that informed our future improvement of the tool. We discuss the future design of sketching feedback tools in spatial visualization training in general.