Semi-Automatically Mining Students’ Common Scratch Programming Behaviors
Minji Kong, Lori Pollock · 2020
In this paper, we introduce a semi-automatic approach to enable scalable data collection and analysis of students’ programming processes in Scratch. We describe a logging tool that records students’ interactions with the Scratch environment as they code. We analyze these programming interaction logs using a combination of automated and manual techniques to uncover patterns that provide insights into common programming behaviors among users during the coding process. We demonstrate this semi-automatic logging and mining approach with a case study, in which students performed the same open-ended coding task in two college courses introducing programming (37 Education and Human Development majors taking an Ed Tech course and 44 non-CS majors taking a general introductory CS course). Our findings demonstrate that we can uncover patterns that help identify which constructs students commonly used for a given task, various types of tinkering behavior within a given task, and some indications of students’ levels of understanding of the blocks. This kind of information is a valuable starting point for where and how to direct teaching efforts.