Grading Exams: How Gradescope Revealed Deeper Insights into Our Teaching

Teacher has stack of papers to grade

Having clear insight into students’ thinking and where there might be gaps in their understanding of a topic is incredibly valuable. It allows a skilled instructor to adjust their teaching to help all students learn more effectively. But with large lecture classes, it can be hard for instructors to glean this kind of detailed insight. The mathematics department at Oregon State University has found a solution to this challenge.

OSU is a sizable institution with more than 30,000 students. At any one time, we have 600 to 800 students taking College Algebra. Our goal in teaching this many students effectively is to provide each student with as similar an experience as possible. In my role as a course coordinator, I lead a team of 11+ instructors and graduate teaching assistants who teach College Algebra each term.

In an effort to keep the many sections of College Algebra consistent, we, like many universities, give common exams. Where things get really difficult is in the grading of these common exams. In the past, each instructor would grade their students’ exams based on a rubric developed by the team, and we would cross our fingers that everyone interpreted and applied the rubric in the same way. Since we were grading by hand, and only grading our sections’ exams, we had a more difficult time knowing how all College Algebra students, as a whole, were performing. Each individual instructor could draw their own conclusions from the patterns they were noticing but that was the extent.

It wasn’t until we changed the process of grading our exams dramatically, so team members only graded two or three questions instead of an entire exam, that suddenly an entirely new world of data-informed instruction and consistency in grading opened up to us. This was made possible to us by using Gradescope from Turnitin, which allows us to grade handwritten (or electronically submitted) student work collaboratively online.

Grading by question instead of course section

When team members are each grading a batch of exams in their entirety for a section of students, they might be able to make some generalizations about where students’ misconceptions lie. Team members can come together after grading their subset of exams and compare notes to see if others had similar impressions or spotted the same trends.

But when team members, instead, take a deep dive into just a few questions each, they become experts on how students approach those particular problems, and they can share this wisdom with the rest of the team. When team members can see at a glance how all 800 students have answered a specific question, that’s where the real insight becomes possible.

Suppose students are asked to sketch a graph of  y = x2 + 3. Gradescope allows the grader to see individual thumbnails of all 800 students’ graphs and sort them based on the different ways each student answered, for the purpose of grading. This grouping mechanism provides the grader powerful data about the number of students that 1) drew the correct graph, or 2) shifted the parabola down instead of up, or 3) drew the wrong function.

For instance, if 70 percent of all students in the course drew the correct function with the correct translation; 20 percent knew it was a parabola, but got the translation incorrect; and 10 percent of responses were wrong for other reasons, instructors would know exactly what concepts students knew and which concepts to review again in class.

Concrete data to shape instruction

Gradescope provides us with very fine grain data for each question regardless of type, from multiple choice to open-response questions. This allows instructors to modify instruction, reconsider test questions, and easily compare the similarity between versions of exams.

Having an easy way to sort each question and the response into groups of similar answers allows the instructional team to see instantly how many different responses students gave and how popular each one was—insights that are invaluable when planning follow-up instruction. It also saves team members precious time when providing students feedback. They can apply the same comment with one single click to every student who answered a question the same way.

With the insights we get from this approach, we can see exactly what students are doing well and where they’re going wrong.  Whereas before, we were only guessing about the misconceptions that existed across all 800 students in the course, now we have concrete data to shape and inform our instruction.

At the time of writing this article, Katy Dumelle was the assistant director of integrated professional development for Oregon State University’s College of Science. She has also been a full-time math instructor and course coordinator for the college but has recently started working with Turnitin to help other faculty integrate Gradescope into their courses.