A Multi-Method Approach for Exploring Programming Trajectories Through Log Data: Insights from Data Visualization Tasks

Interest in data science education is growing as data becomes more prevalent in our daily lives and plays a central role in making informed decisions and understanding the world. Due to the interdisciplinary nature and broad scope of the field, further research is essential to unravel how K-12 students can effectively interact with data through productive learning experiences. This is particularly true in data visualization activities, in which students must employ a variety of skills to effectively extract and communicate data insights. In this study, we describe key actions involved in creating data visualizations using a block-based programming environment (PlayData). Based on qualitative video analysis, we identified six core data visualization programming moves: program creation, selection of parameters, output inspection, data inspection, program rearrangement, and visual design. Then, using learning analytics techniques and Epistemic Network Analysis, we developed a method for automatically categorizing and characterizing those moves based on fine-grained log data collected from the environment, which allowed the identification of patterns in students’ trajectories. We found that students’ work is distributed across several micro-tasks, each involving distinct types of interaction with the environment and holding a unique value in the process of engaging in programming, data analysis, and visual design. As students progress, there is a transition among these moves, suggesting the need for activities that ensure comprehensive exposure to all of them. Our study presents two main contributions: a novel approach to automatically categorize and describe learning trajectories in open-ended programming tasks and insights into how K-12 students engage with those tasks in a data-related context, laying a foundation for better supporting learning and research in this emergent area.