Who can predict their performance more accurately? An investigation of undergraduate students’ self-assessment behavior in mathematics courses

Abstract

We collected data on students’ self-assessment behavior from four sections of a Calculus II course. Students were asked to write their expected scores on each of the weekly in-class quizzes and problems in the exams, immediately after they completed them. They were then asked to justify their expectation in writing. One-on-one interviews were conducted with a purposefully selected sample of students. During the interviews, they were asked to explain their perceived reasons for their self-assessment behaviors. While the results from quantitative analysis seemed to partially reinforce the findings of existing research that low performers generally overestimate, high performers underestimate their performance, and those in-between performers were (almost) accurate predictors, results from qualitative analysis provided additional insights into their self-assessment behaviors. After analyzing qualitative data, we identified five categories of student behavior: knowing about knowing (KK), not knowing about knowing (NKK), knowing about not knowing (KNK), knowing something is not known but not sure what (KBNKW), and not knowing about not knowing (NKNK). The quantitative analysis showed that students exhibited greater accuracy in assessing their performance when they belonged to the categories KK, KNK, and KBNKW, while their accuracy was lower when they fell into the categories NKNK and NKK. In other words, students who had greater awareness of their level of knowledge were more accurate in predicting their scores compared to their peers, irrespective of their actual performance levels. The logistic regression model revealed a substantial increase in the likelihood of underperforming students overestimating their performance compared to their high-performing counterparts.