Many children begin to learn to code in a self-directed context, such as by creating an animation, game or phone app. Recent research has begun to investigate and evaluate the results of this process: children's projects. However, little is known about the different trajectories novices have during the long-term process of self-directed programming learning. Our aim is to identify the existing types of trajectories and be able to determine a specific child's trajectory. If that trajectory does not lead to significant progress or continued motivation, we might be able to nudge them toward a different trajectory.
We hypothesized that there may be several main types of trajectories. To explore this, we clustered children's programming progression in order to identify trends. We used a data set of Scratch programs for its large sample size and diverse population [1,3]. We used Dr. Scratch's scoring of computational thinking skills as our initial feature set [2]. Our clustering analysis identified two main groups of users. The first group, 55.8% of users, generally progressed upward over time in their computational thinking skill score. The second group, 44.2% of users, showed little to no improvement over time.
These preliminary results raise many questions about these two high-level groupings of trajectories of computational thinking skill demonstration in Scratch projects. In the future we want to further explore the trajectory clusters and what they could mean for supporting self-directed learning of programming.