Incorporating AI into the grading process in a carefully crafted, human-managed way can make the work more efficient and give teachers the gift of time to ultimately provide students with richer, more detailed feedback.
Prompt for ChatGPT: Create an abstract art illustration to accompany a blog post titled "Is AI the New Scantron? Automated Grading in the Age of Artificial Intelligence." The image should show a tranformative upward progression from manual, laborious grading to streamlined scores and justifications. This should be depicted within a soaring momentum. A female teacher thoughtfully oversees the process from start to finish. The teacher is positive and suggested through simplified forms rather than detailed cartoon features. Use muted colors and tones. The frame should be square.
By Dani Kachorsky, PhD
It is the end of another school year. Some schools, like mine, celebrated graduation this past weekend, while others have a few more weeks to go. As I reflect on my final weeks of the semester, I realized that I didn’t quite feel the same end-of-the-year crunch that I am accustomed to. And, I have AI to thank for that.
Typically, the end of the year is stressful. Final exams, essays, and projects pile up fast, making the grading process feel endless. Teachers have additional responsibilities, as graduation practice, make-up exams, and cleaning up the classroom in advance of summer maintenance and repairs begins to take up much more of their time. I have long been jealous of math teachers, many of whom seem to sit back and calmly wait for a computer to crunch the numbers on their multiple-choice assessments. They shake their heads at English teachers, saying things like, “I would hate to have to grade that many essays.”
Scoring automation isn’t a new idea—we’ve had it since the invention of the trusty Scantron sheets we remember from our own school days, filling in bubbles carefully with No. 2 pencils. However, scoring automation has never been possible for certain tasks, such as essay and presentation assessment. But now, the landscape of automated grading is evolving rapidly thanks to AI. Traditional multiple-choice scoring tools have been joined by newer platforms such as Flint AI and Khan Academy's Khanmigo, which attempt qualitative evaluations using rubrics more typically employed by human teachers. But this raises important questions: Can we trust AI to evaluate student work fairly and accurately, especially when assessments become subjective and nuanced?
Right now, reactions to AI grading are mixed. Many teachers, myself included, feel strongly that assessment remains fundamentally a teacher’s responsibility. After all, grading isn't purely objective—it reflects our knowledge of our students' lives and the context surrounding their work. Interestingly, student opinions vary too. Some students prefer AI grading, believing it eliminates potential teacher bias or personality clashes; others feel strongly that only a teacher, with deep familiarity and understanding, can truly evaluate their efforts fairly (The Learning Network, 2024).
I've experimented extensively with AI grading tools like Flint and more generalized platforms, like ChatGPT and Claude. In my experience, these tools can indeed streamline the grading process, but they require supervision. Sometimes these tools grade average student work a bit generously while being tougher on exceptional submissions. However, if carefully guided with rubrics and sample work, these platforms can provide detailed rationales for rubric-based scores—something I rarely have time to write extensively myself.
Still, AI isn’t perfect. It often struggles with consistency over large numbers of assignments, occasionally confusing student work, or skewing too harshly on grammar and punctuation errors. Also, LLMs historically struggle with accurate numerical conversions of rubric scores, so double-checking math remains essential.
There's an important distinction between using general AI chats and dedicated chatbots specifically designed for grading tasks. General AI chats can lose track of the parameters and context if used extensively, potentially mixing up student assignments or rubric criteria. In contrast, designated grading chatbots are built with set parameters and retain consistent criteria throughout their use, significantly reducing errors and increasing reliability.
The real benefit of AI grading, in my view, lies in efficiency. While efficiency has become a dirty word in education since the rise of AI, tools that teachers can use to be more efficient with their time and in their practice can and, in my view, should be seen as a form of restorative pedagogy (Winn, 2013). Having AI conduct an initial assessment can save significant time, allowing teachers to quickly confirm or adjust scores rather than agonize over every rubric detail. Crucially, this gives me more time to provide students with richer, more detailed feedback than I otherwise could manage—feedback they might even actually read!
That said, transparency and ownership remain critical. Teachers should always make students aware of how they are being assessed. Also, teachers cannot use AI as a scapegoat for grading decisions. We must review, confirm, and stand by every evaluation. AI is a tool to support—not replace—our professional judgment.
Try it Yourself
Ready to explore AI grading? Here are a few practical steps:
Choose Your Tool: Explore dedicated educational platforms like Flint AI or Khanmigo for qualitative evaluations, or experiment with general-purpose AI like ChatGPT for more flexibility.
Prepare a Clear Rubric: Provide AI with a detailed rubric and sample student work representing different achievement levels.
Supervise Closely: Initially, review AI-generated grades closely to calibrate its evaluations and correct discrepancies.
Combine Human and AI Strengths: Use AI-generated scores and feedback as a foundation, adding your personalized insights and adjustments.
Stay Transparent: Always clearly communicate to students how AI is being used and ensure they understand you are ultimately responsible for their grades.
Remember, when possible, dedicated grading chatbots are better suited for grading automation than standard AI chats because they maintain consistent parameters and won't forget critical details you've set.
By thoughtfully integrating AI into your grading process, you can reclaim valuable time at the end of the year—time better spent celebrating your students’ accomplishments and preparing them for their next academic adventures.
References
The Learning Network. “What Students Are Saying About Teachers Using A.I. to Grade.” The New York Times, The New York Times, 5 Dec. 2024, https://www.nytimes.com/2024/12/05/learning/what-students-are-saying-about-teachers-using-ai-to-grade.html.
Winn, M. T. (2013). Toward a restorative English education. Research in the Teaching of English, 48(1), 126–135.
AI Transparency Statement
This post was created using the AI-assisted workflow I describe in a previous essay. I began by audio recording my thoughts and experiences, then used AI to transcribe and synthesize my reflections while maintaining my voice. I added and revised material through a few additional prompts in the LLM interface before copying the content into a document, where I made further revisions. I read the sources that are linked and cited in this post in advance of my audio recording and verified the accuracy of information prior to posting the essay.
AI Prompts
ChatGPT prompt for blog post: Write a blog post about using AI to automate and streamline the grading and scoring process. Below is a transcript of an audio recording about this content. Use this content and this content only to create the blog post. At the beginning of the post, acknowledge that we are at the end of the school year and teachers are under the gun to get final exams, essays, and projects graded. At the end of the post, include a section titled "Try it Yourself" with recommendations for how teachers can use AI to assist them with grading at the end of the year. Hold on to my "voice" as much as possible: [transcript inserted here]
ChatGPT prompt for blog post revision: In the paragraph that begins "Still, AI isn't perfect" or in a new paragraph immediately after, explain the difference between using an AI chat and a designated ChatBot designed specifically for the purpose of grading a particular assignment. In the "Try it Yourself" section, add a reminder that if possible ChatBots are better suited to grading automation than standard AI chats because they won't forget the parameters you set. Conmigo is actually spelled Khanmigo, so revise that.