Ungrading, AI, and Feedback as Scholarly Work

Along comes December, and the grading begins. Teachers at all levels stare ahead at a week or more of hunching over their desks, reading student work, grading exams, providing students with that final affirmation, those last course corrections, and the culminating appraisal of their work across the semester. It’s a tedious, difficult, and time-consuming part of the work of teaching. As necessary as it is routine.

I remember the end of the term very well. In all cases, I found myself looking at a high pile of papers, or a long list of handed-in digital assignments, most of them written reflections, proposed course syllabi, works of creative writing, video presentations on research, etc. None of them simple to evaluate. I did this to myself, of course. I encouraged diverse viewpoints and interpretations of final assessments; I asked students to break the mold of what they, or I, or any other teacher might expect for their work; I advocated for imagination, innovation, and imperfect experimentation over flawless execution. I refused any method of teaching writing or teaching pedagogy that involved quantifiable assessments; nor did I permit myself the luxury of marks in the form of letter grades or numerical scores. Each assignment—from those in a developmental composition class to those in a graduate course on digital pedagogy—required individual attention.

Teaching is, at its core, a profoundly relational act. The classroom—whether it’s in a lecture hall, online, or sprawled across emails and office hours—is a space where connections are built, tested, and nurtured. Feedback sits at the center of this work. It’s not just a transactional moment where information is corrected or clarified; it’s an intimate exchange where ideas meet and evolve, where vulnerability is handled with care, and where a student’s effort is honored with thoughtful attention. To offer meaningful feedback is to step into the life of another person’s learning, to engage with their thinking and their struggles, and to acknowledge the risk they take in putting their work out into the world.

And yes, it’s time-consuming. Anyone who has spent hours commenting on student essays knows the weight of it. But it’s also deeply rewarding. Giving feedback is how we engage not only with a student’s work, but with the ideas underlying the connection between that work and the topics of our class. 

When we take the time to write feedback, we connect our own thinking to a student’s thinking, thereby opening up the possibility for new knowledge and insight to develop. It’s a mistake to think that the feedback we write is only for the student’s enlightenment: it’s reflection upon reflection, all pointing back to the material we teach and study, and creating a new, more collaborative understanding of that material

Grading, in this way, is scholarly work, and can form a meaningful foundation for our relationship to students and to our own profession.

This is what makes feedback fundamentally human. It doesn’t adhere to a script or follow a predictable pattern. It’s messy, contextual, and often emotional. It emerges from each essay, each project, each assumption in a research paper. It requires presence.

And, it takes time. Sometimes too much time. I remember evenings on the dark weekends of December when I could not bear to lift another paper off the pile, to open one more email to respond to. Literally on the verge of tears, I wished for a better way to do this part of the work.

One of the prevailing idealisms around generative AI is that it can lift this burden of grading off a teacher’s shoulders. This is part of the premise of Anant Agarwal’s recent opinion piece in The New York Times, “How A.I. Can Revive a Love of Learning.” A rose-coloured essay on the virtues of an artificial teaching assistant by the founder of edX, the piece makes the claim that teachers will learn to re-enjoy teaching if they can turn over the grading to an AI that doesn’t mind the labour. Agarwal writes: 

“I’ve spent untold hours grading, reviewing assignments and preparing materials, which while useful, are not the parts of teaching that speak to and inspire students. These tasks can be automated.”

This is both a woe and a solution to that woe that I’ve heard again and again, both in other articles about AI as a teaching assistant, and from teachers with whom I’ve spoken through my work at Course Hero. Grading takes up too much time. It’s tedious, and unrewarding. If it could be automated, teachers would have more time to spend on students. As Agarwal notes (in echo of many others), AI lets us “instantly offer substantive feedback on assignments, freeing us from the endless grading cycle and giving us more time to connect with our students.”

It’s such a nice idea, but it’s tinny on the ear. For starters, let’s not quibble over whether “instant” feedback can be “substantive” feedback. It can’t be. Nothing instant is substantive. Not dialogue, not gratification, not even noodles. Instant isn’t nutritive.

To be clear, Agarwal is selling something. edX/2U was recently named to Fast Company's annual list of the World's Most Innovative Companies. The honor is in recognition of their advancements in using AI to enhance student experience in edX courses. According to Agarwal, "Through our AI advancements, we're not just shaping the future of learning, but empowering individuals worldwide to thrive in the age of AI." That humble-brag speaks volumes about Agarwal’s investment in a rhetoric around AI that’s sunny and bright, where the future sees all teachers getting help from artificial TAs who blithely go about grading and offering feedback without breaking a sweat.

But in reality, providing feedback requires attention that algorithms can’t simulate, no matter how sleekly they’re programmed. While AI might one day offer a quick critique or a polished suggestion, it can’t do the hard, relational work of seeing a student not just as a paper to be graded but as a person to be understood. Feedback, when done well, is as much about listening as it is about speaking. It’s about creating a space where students feel seen and heard, where their learning is affirmed as their own.

Interestingly, there’s an existing solution to Agarwal’s dilemma about grading and giving feedback taking valuable time from the work of teaching (never mind that those things are part of the work of teaching), and about his worry that we aren’t connected enough to students: and that’s ungrading. 

Now, I rarely dip my toes into discussions about ungrading. They seem charged, to me, and fraught. I’ve written one blog post, and the afterword to Jesse Stommel’s book about ungrading; and in 2022, I made the mistake of not carefully enough editing an article ghost-written for me for Times Higher Education—the publication of which caused a bit of upset in the ungrading community (including a Discord group that came together to write a takedown piece in response). People get possessive about ungrading, and I can’t blame them. In my experience, how we teach is as close to our hearts as how we write or how we parent. 

There are a myriad number of ways to address Agarwal’s concerns within the context of ungrading, but I’ll start with self-assessment. In many ungrading practices, assessment is handed over to the student to do for themselves. Jesse Stommel and others advocate for student self-assessment, offering that this approach creates a dialogue between students and their own work, and between students and their teachers. Rather than boil down to a number or letter, both of which have arbitrary and subjective meaning (but are objectively powerful), and both of which close off rather than begin conversations, self-assessment gives a student the ability to identify their own meaning in their work, conveys that reflection to their teachers, and opens up a space for dialogue.

And there are multiple other ways that ungrading can be applied to encourage students to engage in their learning, and to create connections between students and teachers. Grade-free zones, process letters, student-created rubrics, peer assessment—all of these shift the burden of grades, and their power over students’ lives, in favour of learning that’s assessed as it happens. In their NIH article, Jenny Inker and colleagues point out that “Ungrading has been shown to increase students’ intrinsic motivation to learn, their willingness to take intellectual risks, and the quality of their thinking.” Ungrading may not entirely take the grade off the table, but it does substantially change the way grades are perceived, doled out, and felt.

And to Agarwal’s point, ungrading also saves teachers’ time, while simultaneously creating the necessary conditions for deeper connections with students.

We don’t need AI to relieve us from the burden of grading, not when there are so many effective—and decidedly human—strategies available through ungrading. Investing in students and investing in learning does not need to look like letting teaching become automatable.