Facilitating Programming Success in Data Science Courses through Gamified Scaffolding and Learn2Mine

This paper discusses the learning strategies adopted in a publically available, cloud-based learning environment, Learn2Mine, which facilitates student-progress as they solve data science programming problems. The learning system has been evaluated over three consecutive terms. Learn2Mine was initially introduced in an introductory course and pilot-tested for usability and effectiveness in Fall 2013. Students reported positive opinions on usability and effectiveness of the system in their completion of programming assignments. In Spring 2014, Learn2Mine was evaluated in an upper-level data mining course by comparing student submission rates and amount of programming accomplished for a group with access to the tool versus one without access. The group with access to Learn2Mine had an average assignment submission rate of 84%, while the group without had an average submission rate of only 48% (difference significant at p < 0.01). In Fall 2014, a controlled experiment was conducted in an introductory data science course: one group of students worked on a multi-part programming task with support of scaffolding and gamification as implemented in Learn2Mine, while the other section did not. The group with access performed significantly better in overall task completion (p < 0.01).