Using gameplay data to investigate students' problem-solving behaviors in Zoombinis

One of the most critical intellectual abilities in today’s K-12 education is problem-solving skills. In game-based learning environments, it is essential to analyze students’ problem-solving behaviors so that instructors and researchers can provide timely feedback and relevant interventions. However, it is difficult to comprehend how students solve problems in games because game-based learning is always implicit and complicated. This study aims to investigate students’ problem-solving behaviors in a specific puzzle-based game context. Specifically, we used the Continuous Hidden Markov Model (CHMM) to identify students’ problem-solving stages, as well as sequence analysis to examine their problem-solving strategies. According to the findings, CHMM performed well to determine students’ problem-solving stages, and sequence analysis efficiently identified five problem-solving strategies. By applying the integrated techniques, we automatically categorized students’ real-time problem-solving behaviors and examined frequent strategies they applied, which can be used to better understand students’ problem-solving processes and provide real-time interventions for future studies.