How to Grade Like a Physicist: Using Bayesian Statistics to Assign Grades

Among the difficulties involved in assessing students is the problem of grading: how to fairly combine a variety of assessments to produce a single grade. The legion of problems with the arbitrary 90-80-70% breakdown of letter grades is well documented, but it is less clear how best to replace it. Here I present an approach to grading based on Bayesian statistics and inspired by statistical techniques in particle physics measurements, which results not just in a single grade but in a probability distribution of grades from which uncertainties and confidence limits can be extracted. I discuss its implementation in a high school class and some of the ups and downs of the new system, as well as student responses.