An exploratory process mining on students’ complex problem-solving behavior: The distinct patterns and related factors

Abstract

This study showcases how process mining techniques can be used to uncover students' problem-solving processes. The data comprised event logs recoding the execution sequences of 2066 students who engaged in an example “Climate Control” task from PISA 2012. By analyzing students’ trace variants, eight distinct problem-solving patterns were identified, reflecting their varied approaches to key issues of problem exploration and representation. Then, the First-Order Markov Model and analysis of variance were used to highlight similarities and differences among these eight patterns. A multinomial regression analysis further examined their crucial predictors. The results revealed notable differences in the frequency of strategies employed, transitions between strategies, and the timing of representations across eight patterns. While some patterns achieved success by systematical or mixed use of the varying-one-thing-at-a-time strategy, others relied on guessing. Unsuccessful problem solvers might struggle due to an inability to control variables, flawed reasoning from trials with alternative strategies, or unfamiliarity with the diagramming interface. The regression results suggested that prior knowledge, problem-solving attitudes (openness, perseverance), and the availability of information and communication technology in school play significant, albeit different, roles at various problem-solving stages. The uncovered key issues and behavior patterns provide empirical evidence for elaborating theories about problem-solving processes. The identified patterns of students’ successes and failures have implications for developing tailored interventions to enhance students’ problem-solving competencies.