Reducing bias when grading exams13 Dec 2010
As a college instructor, I regularly think about how to grade exams more consistently and fairly. I don’t want to favor students who regularly come to office hours, or to inadvertently award different scores for roughly equal solutions. Everyone should be graded on exactly the same scale. Since I don’t consider the staircase method a viable option, here are a few things I do to try to facilitate this:
Blind evaluation. All my exams have a title page with no problems. The first thing I do is advance them all one page. My goal is to not attach a student’s name with their solution until the entire exam has been graded. I shuffle the exams because I don’t want to grade them in the order they were collected. The order of submission can play a subconscious role in grading – if an exam was turned in early, surely it was a good student, yes? Maybe, maybe not. Plus, I may remember who turned in their exam first or last. I don’t want to see a student’s name until I add up all points for a given exam and then turn back to cover page to record the final grade.
Grade each problem, one at a time, in a single sitting. I grade all the problem 1’s, then the problem 2’s, etc. Each problem is graded in one sitting, with absolutely zero interruptions if at all possible. For a class of 50 students, I can grade a problem in about 1.5 hours – this is not too hard for me to block out and complete in one sitting. This is the only way I have found to award points consistently. I can keep the majority of solutions in my memory so I can easily remember how many points each type of error costs. I find myself often going back to other exams to compare or make corrections to ensure that each is done to the same standard. If you grade half of the exams one day and half the next, you may be in an entirely different mood or psychology – this can result half the exams being graded according to one standard, and half according to another. A final added benefit is that if a student makes a really bad mistake on one problem, I’m not predisposed against them on the next problem because I’m unable to connect the two this way.
Some people may think my approach takes it too far, but I feel better knowing I’ve done everything I can to reduce bias and grade everything equally. I’d love to hear some thoughts from the professors in the audience!