Deconvolution of Students' Perceptions of Examination Fairness

Moore, Michael
Journal Title
Journal ISSN
Volume Title
University of Guelph

The components of students’ perceptions of fairness for written examinations were deconvoluted and evaluated through qualitative and quantitative means. These perceptions are important for maintaining and developing student motivation, instructor/institutional reputation, and an internal locus of control for students. All methods of analysis showed the consistent result that the test’s representativeness of the course material was the most important factor in shaping fairness perceptions toward it: comments about representativeness comprised 48 % of free-form sentiments about tests (out of 8 possible categories) and ratings of tests’ representativeness showed the strongest correlation with ratings of fairness out of all factors tested (standardized odds ratio: 3.6 for chemistry students). The second greatest correlation with fairness ratings was assigned grades (standardized odds ratio: 2.9). After controlling for assigned grades, the correlations between fairness ratings and differences in grade expectations were inconsistent and weak. Little/no additional predictive power for fairness perceptions was gleaned by considering factors such as the difference between students’ performance estimates before and after writing their midterm nor the difference between estimates of test performance compared with actual test performance. Several recommendations come from this work. It is essential to communicate course priorities clearly. Communication of prioritization occurs implicitly (through how much time the instructor chooses to spend on certain areas) and explicitly (through direct instruction and grade weighting). The messages of both communication methods should be aligned. If this cannot occur, then there should be direct communication acknowledging and justifying this. Failure to achieve this alignment and communication correlated with lower fairness perceptions, feelings of test representativeness, and grades, all of which are linked with student motivation. If a test question (or several) were found to be flawed, perceptions of fairness can be salvaged through the removal of that question from the test. This removal should be made public and with concise justification. It is imperative that only one revision event occurs: multiple revisions (especially those that decrease grades) communicate negligence to students.

Fairness, Deconvolution, Regression, Mixed Methods