Anne Duchene

Anyone who teaches a large class knows the pain of grading exams. In my principles of microeconomics (Econ 001) course, there are 800+ students each year. Until a couple of years ago, the exam grading process was a difficult experience. Here are the main challenges I encountered throughout the years:

• Managing massive stacks of exams was messy and time consuming. I typically have one or two TAs grade the same question across all exams. But that meant they had to coordinate on how and when to swap exams—there were always delays and fears of losing an exam in the process. We also experienced delays and frustration when we changed the grading rubric, and all TAs had to go back and regrade everything.
• Because of these time concerns, I was often hesitant to use essay or short answer as opposed to multiple choice questions­—which were easier and quicker to grade. While it is possible to write multiple choice questions that are not obvious and require thinking, their nature makes them unfair assessments: students who have a correct reasoning but made a small mistake don’t get credit, while students who answered randomly can potentially get full credit.
• Time constraints also meant students didn’t get meaningful individualized feedback on their exams—in a large class, it is just impossible to comment each mistake in each question of each exam. At best, we could write some general feedback at the top of the exams, which wasn’t very constructive.
• While I asked graders about the mistakes they ran into repeatedly, I often got inconsistent or vague responses. Because I couldn’t get a real sense of the common misconceptions among students, I also didn’t review them in class afterwards, nor clarify them in subsequent semesters.
• Then someone told me about Gradescope, and it became a game changer. Gradescope is a web app that allows grading PDF-submissions (tests or homework) online. Students work on paper either remotely or in person, then their exams are scanned and uploaded to Gradescope. If the exam is taken remotely, students scan their work into a PDF using a cell phone app. This semester I gave my exams in person, and scanned them myself, using a high-speed, high-capacity scanner. Scanning exam papers is relatively painless, and it is something I was already doing before, to discourage students from making regrade requests on modified answers. This spring, scanning 200 eight-page (four double-sided sheets) final exams took me and my TAs about 10 minutes. Removing staples with a paper cutter was the most tedious part, and took about 20 minutes. Once that part is done, exams are immediately available to all graders wherever they are, and it gives us the freedom to grade at any time, without filling up our backpacks with papers. 

What strikes me the most is how much of a timesaver Gradescope is. My exams are always a mix of multiple-choice, fill-in the blank and open-ended questions. Gradescope’s AI-assisted grading allows me to group answers for multiple choice and fill-in-the-blank questions. After Gradescope has clustered similar answers together, I can categorize them and then grade each group at once. This spring, I graded 12 multiple choice questions for approximately 200 students in 15 minutes, while it probably would have taken a couple of hours to grade by hand.

Open ended questions typically take more time to grade. But with Gradescope, we can grade one question at a time across all assignments, without flipping through stacks of paper. When taking the exam, students write their answers to each open-ended question in a designated box in the exam handout. Once exams are uploaded, I tell Gradescope where each question’s box is, and it will display only that box by default on our screen when we are grading that question. The downside to grading the same box across all submissions is that graders tend to ignore earlier boxes, from which a mistake might have carried over. So, while grading correct answers becomes easier, grading incorrect ones requires more time and investigation. However, I find that overall it helps us grade students’ work more quickly and more consistently, as it often requires a simple click to select the relevant criteria in the grading rubric. 

In fact, my approach to the grading rubric itself has been transformed by Gradescope. I used to grade a few students’ work to get a feel for the quality distribution, and then write a rubric that ensured similar mistakes would get the same score. Modifying the rubric along the way was painful, if not impossible, as graders needed to find dozens of affected exams in their pile. But I can now have a dynamic rubric that evolves as we grade: we can expand it or change it, and the score changes are automatically applied to already graded exams. Knowing that I can adjust the rubric at any time has made my grading experience much less frustrating than it used to be. 

Once exams are graded, students can review them online and download a graded copy. As a result, my TAs no longer spend valuable class time passing back exams in recitation. Besides, Gradescope tells me whether a student has or hasn’t reviewed their graded exam—something that I used to learn too late, when my TAs handed me the unpicked-up exams at the end of each semester. In particular, it helps me keep track of students who did not do so well: are they engaged and trying to improve, or have they disconnected from the course? Having students review their exams online means that I can write comments in the grading rubric that will be visible to all students, or to those who made a specific mistake.  I find myself writing longer and more detailed feedback (as I write each one only once), which helps make grading more transparent and create a sense of fairness. And I can use my and my TAs’ energy in more productive ways than rewriting the same feedback over and over, or adding up scores by hand—a task that was tedious and prone to errors.

I appreciate how Gradescope has made grading easier and more transparent, but I view it as more than just a grading tool: It also provides very useful statistics. While exams are being graded, it shows how grading is progressing, in particular who graded each question and in what proportion, which can be very helpful when working with several TAs who split up the grading load. More importantly, Gradescope provides a comprehensive set of statistics informing me how students have performed. During or after grading, I can look up and see exactly how many (and which) students made a particular mistake on a question. Did those students lose points because they made a small math error in their calculation, or because they don’t understand a fundamental concept? When I realize there is a common misconception among my students, it influences me in the short-term (I review the concept in class following the exam) but also in the long-term (I spend more time on that concept the following semester). And if I think exam grades are too low, I don’t have to curve them. Instead, I can use statistics on students’ performance to adjust the point scale, by assigning less weight to questions where the average performance is relatively low—Gradescope applies the changes automatically to all students and re-computes exam scores accordingly.

There are many features of Gradescope I have yet to explore, but even sticking to its simplest tools has significantly lessened the amount of time it takes me and my TAs to grade exams, while ensuring grading fairness and consistency.

Anne Duchene is the director of microeconomic principles and a senior lecturer in the department of economics.

This essay continues the series that began in the fall of 1994 as the joint creation of the College of Arts and Sciences, the Center for Teaching and Learning and the Lindback Society for Distinguished Teaching. 

See https://almanac.upenn.edu/talk-about-teaching-and-learning-archive for previous essays.