Did ChatGPT Grade Your Paper?

AI is already in the classroom—and the moral dilemma is the point. Walk into almost any conversation about education right now and you’ll notice the same tension: AI isn’t a future possibility, it’s a present reality. Professors are experimenting with it, researchers are testing it, and students are already using it. The question is no longer whether AI can be used in education, but whether it should—and under what conditions. One of the most common and controversial uses is grading, where AI promises to “automate the boring stuff” and save educators countless hours. But beneath that promise sits a much deeper ethical challenge. Cameron Blevins, a history professor, put this to the test by comparing his own grading with a Custom GPT designed specifically for evaluating essays. Using three anonymized student papers and his existing rubric, the AI ranked the essays in the same order he did and produced extensive, rubric-based feedback across categories like writing mechanics, structure, and analysis. In some cases, the feedback was more detailed than what students typically receive. However, it also missed crucial historical inaccuracies and disciplinary nuance—exactly the kinds of moments Blevins sees as central to teaching rather than mere evaluation (Blevins, 2024). A second source from the C.A.R.E.S. Lab at the Harvard Graduate School of Education reinforces this pattern at scale. In their walkthrough of grading essays with the ChatGPT API, researchers demonstrated that AI can score and classify large batches of essays in seconds, with a moderate positive correlation (0.62) between AI-generated scores and human scores, and classification accuracy reaching approximately 84 percent with substantial agreement beyond chance (Kim, 2024). From a technical standpoint, the message is clear: AI can grade writing efficiently and with reasonable consistency. But the moral dilemma begins where efficiency ends. One ethical concern is accountability. If an AI assigns a score, who is ultimately responsible for that judgment? If the model is biased, inaccurate, or blind to deeper misunderstandings, it is still the student who suffers the consequences. The C.A.R.E.S. Lab explicitly cautions against relying solely on large language models for final grades, emphasizing the need for human oversight to prevent unfair outcomes (Kim, 2024). Another concern is the devaluation of academic labor. Blevins estimates that AI could reduce grading time per essay from 15–20 minutes to just a few minutes, a tempting efficiency gain. Yet he also warns that institutions might use such tools to justify larger class sizes, fewer instructors, and heavier workloads, transforming AI from a support tool into a mechanism for further eroding already strained teaching labor (Blevins, 2024). Perhaps the most unsettling ethical issue, however, is the loss of human connection. Blevins describes a dystopian but plausible loop in which students use ChatGPT to write essays and professors use AI to grade them, resulting in a closed circuit of automated writing and automated feedback. When he tested this scenario, the grading AI praised the AI-written essay and awarded it an A—an outcome that feels technically correct but educationally empty (Blevins, 2024). Education, after all, is not just about producing text and assigning scores; it is about mentorship, context, and understanding students as individuals. Transparency further complicates the issue. Blevins argues that if AI is used in grading, students deserve to know how and why, though he suspects many would feel uneasy about being evaluated by a machine. That discomfort matters, because trust is a core ingredient of learning, and once students begin to see education as a transactional system optimized for algorithms, intellectual risk-taking and genuine engagement tend to disappear (Blevins, 2024). Taken together, these sources suggest that the ethical path forward is not an outright rejection of AI, nor blind adoption, but a careful balance. AI can assist with first-pass grading, pattern recognition, and draft feedback, but human educators must retain responsibility for final judgments and meaningful interaction. The moral dilemma of AI in the classroom is not about whether it works—it often does—but about what kind of education we want to preserve. If AI frees teachers to spend more time mentoring, contextualizing, and connecting with students, it can be a powerful ally. If it replaces those human elements with automated transactions, it risks hollowing out the very purpose of education. Sources: Blevins, C. (2024). “Professor vs. ChatGPT: The Grading Showdown.” Kim, Y. (2024). “How to Grade Essays with ChatGPT,” C.A.R.E.S. Lab, Harvard Graduate School of Education.