Discovering Students' Complex Problem Solving Strategies in Educational Assessment

A huge amount of log data accumulates automatically during computer-based educational assessments that can be analyzed for diagnostic or educational purposes using data mining techniques. In this paper, we describe our work of mining students’ complex problem solving interactions when tackling previously unknown and dynamically changing situations. Based on log data analyses, we discovered several problem solving strategies and examined relationships of these strategies and test outcomes. We applied clustering algorithms to discriminate between students with different levels of proficiency in problem solving. We identified four groups of students: two clusters represent successful problem solvers who differ in their level of efficiency, one group of inefficient students might need further practice to be able to solve these kinds of tasks, and finally we found a mixed-strategy group of students. Students in this last group were dynamically developing their problem solving strategy and in parallel, the ratio of correct responses increased from task to task during assessment. In sum, our findings help to advance research on cognitive processes; we support educational researchers in better understanding complex problem solving behavior and identify levels of problem solving proficiency.