Trace data from student solutions to genetics problems reveals variance in the processes related to different course outcomes

Problem solving, particularly in disciplines such as genetics, is an essential but difficult competency for students to master. Prior work indicated that trace data can be leveraged to measure the invisible cognitive processes that undergird learning activities such as problem solving. Building on prior work and given the importance and difficulties associated with genetics problem solving, we used unsupervised statistical methods (k-means clustering and feature selection) to characterize the patterns of processes students use during genetics problem solving and the relationship to proximal and distal outcomes. At the level of the individual problem, we found that conclusion processes, such as making claims and eliminating possible solutions, was an important interim step and associated with getting a particular problem correct. Surprisingly, we noted that a different set of processes was associated with course outcomes. Students who performed multiple metacognitive steps (e.g. monitoring, checking, planning) in a row or who engaged in execution steps (e.g. using information, drawing a picture, restating the process) as part of problem solving during the semester performed better on final assessments. We found a third set of practices, making consecutive conclusion processes, metacognitive processes preceding reasoning and reasoning preceding conclusions to be important for success at both the problem level and on final assessments. This suggests that different problem-solving processes are associated with success on different course benchmarks. This work raises provocative questions regarding best practices for teaching problem solving in genetics classrooms.