Learning from Others’ Codes in Game-based Robot Programming: Behavioral Patterns and Outcomes

Abstract

In programming education, the cultivation of algorithmic thinking (AT) through trial and error has gained importance. This study introduces a game-based programming environment designed to enhance learners' scores by encouraging them to iteratively refine their code. To promote ongoing engagement in the trial-and-error process, a new feature was introduced that allows learners to view “Score-based Gradual Worked Examples”—specifically, the code of peers who have achieved slightly higher scores. A classroom implementation involving 49 university students was conducted, and behavioral log data were analyzed to examine how learners used the system. K-means clustering identified seven distinct behavioral patterns. Among them, learners who frequently referenced others’ codes and made incremental code changes tended to achieve greater learning gains. These findings suggest that not only the frequency but also the approach to referencing—such as making small, meaningful modifications—may be associated with more effective development of algorithmic thinking. Furthermore, preliminary results from expert evaluations indicated that presenting code based on algorithmic similarity to learners’ own work may be more beneficial than a simple score-based approach. These insights enhance the design of adaptive support mechanisms in programming education.