Abstract:
Problem solving is a fundamental skill, essential not only in academic contexts but also in everyday life. In STEM disciplines such as mathematics and physics, problem solving plays a crucial role: it is both a key learning outcome and a means of deepening students’ conceptual understanding. Despite its importance, the assessment of problem solving is typically limited to evaluating final answers or intermediate results, offering little insight into the reasoning processes and procedural strategies students employ. As a result, important aspects of students’ problem-solving competence often remain invisible. Previous research in physics education has explored problem solving using methods such as think-aloud protocols and interviews, which provide rich insights into students’ reasoning but are difficult to implement in authentic classroom or examination settings. This creates a need for alternative approaches that enable the study of problem-solving processes using naturally occurring data. This study employs methods from process mining to analyze undergraduate students’ written physics exam solutions. Treating written solutions as traces of problem-solving activity and as sequences of ordered steps, the study examines process models generated through the analysis of coded written solutions. The results indicate differences in the structural organization of correct and incorrect solutions, with correct solutions generally exhibiting more coherent and sequential patterns of problem-solving activity, while incorrect solutions tend to display less structured and more fragmented processes. These findings suggest that process mining can provide meaningful insights into students’ problem-solving behavior in examination contexts. Such insights may support instructors in providing more targeted feedback and in designing assessment practices that more accurately capture students’ problem-solving skills and conceptual understanding.
