Member-only story
A new philosophy for when to curve exams?
Especially for large lecture classes, it can be common for examinations to be graded on some sort of curve. The goal of this practice is to assure that the overall exam scores remain consistent between semesters, in particular statistics such as the mean and the standard deviation are kept in place. The problem with this is procedure that it is almost robotic in nature. Even if things are kept uniform in some sense, it only takes the scores of the students into account. If there is one thing I learned about designing exams in my mathematical reasoning course this semester, it’s that the scores are not the only thing that should be taken into account when evaluating how well students understand the material. Instead, the responses of the students on the exam — in particular the type and frequency of mistakes that are made — can also shed a lot of light on when it makes sense to curve exams. I’m going to highlight two of the exams I gave in my course this semester as illustrations of my thought-process to answer this important (and sometimes underappreciated) question.
The course I teach has three midterms and a final exam. Each midterm covers one of three thematic units, and the final exam covers all three of those units (with roughly equal weight). I’m going to jump to the third midterm for sake of this discussion, which covered probability, expected value, voting theory…