HOW ACCURATE ARE STUDENTS IN PREDICTING THEIR EXAM PERFORMANCE?

We sought to investigate the metacognitive judgment accuracy of undergraduate students (mostly freshmen, sophomores) responding to multiple choice exam questions in an introductory physical geology course. Students made predictions of their performance in the form of confidence judgments for each individual exam question. Each question was followed by a 5-inch long horizontal line. The left side of the line was labeled “Not at all confident in my answer” and the right end was labeled “Very confident in my answer.” Students placed a mark on the line to indicate their confidence in their response to each question. We measured the position of the mark and normalized its location on a scale from 0 (not at all confident) to 100 (very confident). We then derived estimates of students’ calibration (the match between students’ perception of performance with actual performance) from their confidence judgments to determine individual levels of monitoring accuracy and over- and under-confidence.

We analyzed the data to determine how calibration changed over time, the extent to which calibration predicted performance and individual differences associated with student interaction with course resources. Higher confidence generally correlates with better performance, but the correlation between these variables improved from the first to second exam, perhaps as a result of students becoming more familiar with the nature of the course assessments. Monitoring accuracy measurements reveal that low performers overestimate their answers by a large margin while high performers typically underestimate their responses and are more accurate in their rating of their confidence. A consistent secondary pattern emerges from the data showing that for almost all letter grades, some students are much more accurate in their predictions than others. We will discuss how these confidence measures reflect the number of activities students completed during preparation for the exams.