Confidence Assessment, Calibration, and Gender

Abstract This study considers the utility of repeated confidence assessments for improving students’ calibration and investigates whether there are any associations between calibration and gender. Calibration measures how well a student’s beliefs about their mathematical performance align with their actual performance. Confidence assessments require students to express their confidence in their answers and are commonly used to assess and improve calibration through providing confidence-weighted scores. We analysed data from N = 117 online participants, who gave demographic information before answering nine mathematics questions, grouped into blocks of three, stating their confidence alongside each of their responses. After each block of three questions, automated feedback was provided on the participant’s accuracy, and they were given their confidence-weighted score. We found no improvement in calibration across repeated confidence assessments. However, we found strong correlations between calibration and gender, age, location of education, and field of study. Male calibration was on average 25 marks (out of 90) higher than female calibration. Females lost a mean of 13.3 marks through under-confidence in a correct answer, compared to a mean loss of 3.7 marks for males. This strong association between calibration and gender suggests that efforts to improve calibration may need to be adjusted.