Understanding students' performance in a computer-based assessment of complex problem solving. An analysis of behavioral data from computer-generated log files.

Computer-based assessments of complex problem solving (CPS) that have been used in international large-scale surveys require students to engage in an in-depth interaction with the problem environment. In this, they evoke manifest sequences of overt behavior that are stored in computer-generated logfiles. In the present study, we explored the relation between several overt behaviors, which N=1476 Finnish ninth-grade students (mean age=15.23,SD=.47 years) exhibited when exploring a CPS environment, and their CPS performance. We used the MicroDYN approach to measure CPS and inspected students' behaviors through log-file analyses. Results indicated that students who occasionally observed the problem environment in a noninterfering way in addition to actively exploring it (noninterfering observation) showed better CPS performance, whereas students who showed a high frequency of (potentially unplanned) interventions (intervention frequency) exhibited worse CPS performance. Additionally, both too much and too little time spent on a CPS task (time on task) was associated with poor CPS performance. The observed effects held after controlling for students' use of an exploration strategy that required a sequence of multiple interventions (VOTAT strategy) indicating that these behaviors exhibited incremental effects on CPS performance beyond the use of VOTAT.