The Rise of Generative AI Calls for New Approaches to Grading
The Rise of Generative AI Calls for New Approaches to Grading
A piece co-authored by my students and me
In a recent post, I wrote about a “student partnership project” I was undertaking in my Spring 2023 course. Students in my class had the opportunity to complete a traditional final paper on their own or to co-write a piece with me about the teaching and learning experiments we undertook in the course. Two students volunteered for the co-authored project. We decided to write for an audience of higher ed instructors about our experiences with ungrading and the urgency of adopting new grading approaches in the age of AI. I’m really proud of the work these students did and am sharing what we wrote below. Special thanks to my colleague Josh Eyler for offering us feedback on an early draft.
While this piece is being published in September of 2023, it was written in April and May of 2023.
The Rise of Generative AI Calls for New Approaches to Grading
By Emily Pitts Donahoe, Trey Johnson, & M. Abigail Turner
At the close of our first full semester in a world with ChatGPT, college and university educators are still navigating an assessment minefield. Should instructors incorporate these new technologies into their assignments or exile them from the classroom? Is the arrival of generative AI a new crisis for higher ed, or does it simply magnify a longstanding problem with academic dishonesty?
As two students (Abi and Trey) and an instructor (Emily) in a spring 2023 writing course, we’ve seen first-hand both the positive and negative impacts generative AI can have on student work. AI tools can be powerful learning aids for first-year writing. They can also tempt students to short circuit the learning process by offering to do the work of thinking and writing for them.
Policing student cheating, however, is not a sustainable solution to this problem. AI detection tools are not currently reliable and not likely to be so in the long term. Alternatives like asking students to write their assignments in class by hand or instituting other kinds of surveillance can create inequities and damage teacher-student relationships. And banning ChatGPT outright misses an opportunity to teach students how to navigate a world in which generative AI exists around every corner.
More importantly, such approaches fail to address the deeper concerns that animate academic dishonesty and motivate students to misuse AI.
Students take shortcuts on their work for a number of reasons. But one important factor that contributes to cheating is an overemphasis on grades: students are more likely to engage in academic dishonesty when their focus, or the perceived focus of the class, is on grading.
If students are turning to ChatGPT to do their work for them, it’s because they see little value in that work beyond the grade they receive on it. It’s because attaining a certain grade, rather than learning, has become the purpose of their education.
Both student co-authors of this piece have experience with this mindset. Abi’s discouraging early experiences with writing-focused classes decreased confidence in her writing ability and increased her anxiety about grades. Writing became an exercise in finding out what the teacher wanted and turning in whatever material would achieve the highest grade—not in learning something new.
Trey had a similar experience. His approach to previous writing courses was to conform his writing to the rubric, with the goal of getting the highest number of points rather than growing or improving his skills. A competitive, results-based grading system led to a fear of failure, inhibiting his potential as a writer.
It’s not uncommon for students to approach their studies in this way—and that’s hardly students’ fault. They have grown up in a system that uses grades and standardized test scores to “track” them into ability groups, to determine their admission to college and eligibility for financial aid, to mete out postsecondary opportunities—even, disturbingly, to predict their future criminality.
Grades have serious consequences. For students juggling a variety of responsibilities and personal pressures, generative AI can offer an efficient route to a much-needed A–one that may otherwise feel impossible to attain. If getting good grades is the most important motivator for students, why wouldn’t they turn to ChatGPT to complete their assignments?
While addressing this problem requires many different solutions, we suggest a simple one: eliminating, or at least minimizing, the emphasis on grades and refocusing instead on learning and improvement.
Our class last semester took an “ungrading” approach to assessment. In our case, that meant that students didn’t receive points or letter grades on the work they submitted. Instead, they received extensive written feedback with the chance to revise and resubmit their work multiple times. Final grades were determined at the end of the semester in individual conferences between Emily and each student. Students suggested a grade for themselves drawing on evidence from their work in the course; Emily offered holistic feedback about their strengths and areas for future development.
This approach not only helped refocus the course on learning but also allowed students to engage in fruitful conversations about generative AI, considering its affordances and limitations as well as how it might be most productively used in the writing process. When students used AI inappropriately, it became an opportunity to discuss how that misuse was impeding their learning rather than how it was affecting their grade.
Additionally, the class prioritized process over product, allowing students to concentrate on continuously improving their writing. Deemphasizing grades helped students tap into their intrinsic motivation, inspiring them to do their own work rather than having someone (or something) else do it for them. Both Abi and Trey found that ungrading helped them build self-confidence, lower fear of failure, and grow as writers—without feeling the need to use ChatGPT.
Ungrading looks different for everyone. While you may not be able to employ the same practices we used in our course, there are a few things any teacher can do start minimizing grades and making ChatGPT irrelevant:
Allow students to reattempt or revise their work. Most instructors believe that mistakes are opportunities for learning, but our grading systems rarely send the same message. Instead of incentivizing growth, they penalize failure. Allowing students to revise and resubmit not only reduces stress but creates feedback loops that lead to better learning. When Abi and Trey received comments on their work and were able to keep improving it, using ChatGPT to generate assignments became less appealing.
Refocus on formative feedback to improve rather than summative feedback to evaluate. Not all feedback benefits students in the same way. Giving quick remarks to justify a grade is not as useful as giving feedback to students on how to improve. Providing formative, rather than summative, feedback—and providing it without a grade attached—can help students refocus on learning rather than simply generating a product to achieve a specific grade.
Incorporate self-assessment. In our class, students were asked to discuss the specific choices they made in their writing and reflect regularly on their own progress toward the learning goals—tasks in which generative AI is marginally useful at best. While this may be difficult to implement for all courses, self-assessment is still achievable. For smaller classes, teachers might hold brief in person meetings with students 2-3 times a semester to hear from them about their progress in the course. For larger classes, teachers might assign self-assessment surveys to help students reflect on and evaluate their learning.
New technologies expose the need for new teaching approaches. And changes to our grading system, which originated in the 19th century, are long overdue. While generative AI presents a number of challenges, it also presents one very important opportunity: the chance for all of us—teachers and students—to stop worrying about grades and refocus on learning.
Nice piece and I appreciate your three recommendations at the end. Simple steps to help better align assessments with student learning.
At a meta level I’m curious about the experience of writing a piece together with your students. I’m considering doing something similar with a few of mine this semester and would be interested to hear what worked best in your experience and anything you would try differently if you were to do it again.