Three Approaches to Collaborating with AI to Provide Students with Feedback
Using AI for feedback can take different forms
There is more than one approach to integrating AI feedback, and approaches can vary greatly in the level of AI usage. Finding the right balance depends on the teacher, the assignment and the priorities of the day.
ChatGPT prompt for image: Create a minimalist, light hearted illustration of a human teacher with pink hair and a robot gesturing animatedly while looking at a pile of student work stacked on a desk. There should be an air of productivity and progress in the image that gives a sense that a lot is getting done.
By Dr. Dani Kachorsky, Ph.D
I've been thinking a lot lately about the delicate dances of balancing AI-generated feedback with the kind of human, nuanced support that only teachers can provide. I’ve recently written posts about how to enhance teacher feedback with AI tools and how to utilize AI-generated feedback in classroom contexts. (I’ll give you three guesses as to which chapter of the book I’m currently writing!)
Over the past few years, I’ve experimented with a few different models in my classroom, and I wanted to share some thoughts on what’s working, what’s challenging, and what I still wonder about as we move forward with these tools. As an English and AP Research teacher, my experimenting has been limited to English Language Arts and AP Research. However, I see these approaches as being useful to any content area or skill that requires iteration—from social studies and science to math, physical education, health, or fine arts.
The Rotational Approach
One method I’ve tried is what I call a rotational approach. Here’s how it works: students first write a draft under controlled conditions—say, a lockdown browser where AI isn’t available—so I know I’m getting their raw, unedited thoughts. I often have to admit, the early drafts can be rough—capitalization errors, punctuation all over the place, and sometimes ideas that just aren’t there yet. But that’s okay; it’s part of the learning process.
Once I have that draft, I have students ask AI to generate some preliminary feedback. In many cases, the AI picks up on common issues—the lack of a debatable thesis, disjointed paragraphs, or the need for better evidence-to-claim connections. Although the feedback tends to be generic, it serves as a solid first pass that helps students understand the basics of what needs improvement. Then, after the AI’s feedback has done its part and students have revised their initial drafts, I step in to offer more targeted, nuanced suggestions tailored specifically to each student’s unique challenges. It’s a bit like having an extra set of eyes that catches the obvious so I can focus on the details.
Rotational Approach – Other Subject Examples:
Social Studies: Students draft an analysis of a historical event in a controlled setting; AI identifies factual inaccuracies or missing context, and then the teacher makes more suggestions about refining the argument with deeper historical insight.
Science: After writing a lab report, students ask AI for feedback on issues in methodology clarity and data interpretation, and the teacher later offers detailed guidance on experimental design and scientific reasoning.
Drama: Students record a monologue under exam conditions, and ask AI to analyze the audio for factors like clarity, pacing, and tone, flagging areas that need improvement (for example, excessive pauses or unclear articulation). After addressing this general feedback, students perform for the teacher and receive in-person coaching to help refine each their delivery.
The Guided Model
Another approach that’s been eye-opening for me is the guided model. In this method, I begin by closely reading a student’s work and jotting down specific areas for improvement—maybe they’re struggling with contextualizing evidence, or their thesis doesn’t quite hit the mark. I then prompt the AI with those targeted issues along with the assignment description, the rubric, and even examples of past student work. This extra context allows the AI to generate feedback that is much more specific than a blanket “improve your thesis” comment.
I love this approach because it creates a sort of partnership between me and the AI. The teacher sets the stage with concrete insights, and the AI fills in the gaps, sometimes even suggesting actionable examples or steps the student could take to improve. Of course, there’s always that nagging concern about how much of the creative process gets handed over to a machine, but I see this as an opportunity for students to refine their voice rather than replace it.
Guided Model – Other Subject Examples:
Health: Students write reflections on their wellness, and after the teacher notes areas for improvement (like clarity in discussing nutrition or stress management), the AI offers detailed feedback and suggestions for strengthening their arguments.
Math: When students explain their reasoning for solving a complex problem, the teacher identifies lapses in logic or incomplete steps, and then prompts the AI to provide clearer explanations and suggest alternative problem-solving strategies.
Music: Students perform a live piece that’s recorded. The teacher takes notes on their performance and specific issues—perhaps their timing is off or their dynamics need more expression. Then, the teacher feed these targeted observations along with the performance criteria into the AI. The AI responds with focused suggestions (like improving phrasing or adjusting tempo), which the teacher uses as a springboard for one-on-one coaching sessions with students.
Experimenting with ChatBots
A third strategy I’ve been exploring is using AI chatbots—essentially, custom-built bots trained specifically for certain assignments or skills. The idea is to create an interactive tool where students can ask for guidance in real time. For example, if a student is writing an essay about the American Revolution, the chatbot can be programmed with the assignment guidelines, rubric details, and approved source material. It then serves as a sort of on-demand tutor, providing advice on areas like maintaining a professional tone or adhering to specific writing conventions.
Setting up these chatbots is surprisingly user-friendly, and you don’t need to be a tech expert to get it up and running. However, I always make sure to place firm guard rails around what the chatbot can do—after all, it’s there to support, not to do the work for the student. While I find these bots offer an engaging way for students to iterate on their drafts, I remain cautious. There’s always the risk that a chatbot might give overly generic advice if not properly trained or that students might lean too heavily on it, bypassing deeper engagement with the material.
ChatBot Approach – Other Subject Examples:
Social Studies: Develop a chatbot that answers student queries about historical events and primary sources, offering evidence-based explanations and guiding them through document analysis.
Science: In a science lab setting where students are engaged in hands-on experiments, create a custom AI chatbot to field real-time questions about lab procedures, equipment setup, or troubleshooting unexpected results. As students work through their experiments, the chatbot offers hints and safety reminders on demand. Afterward, the teacher can follow up with a discussion to deepen their understanding and address any lingering questions.
Math: Use a chatbot to lead students through complex word problems by prompting them with step-by-step questions, encouraging critical thinking rather than simply providing solutions.
Final Thoughts
Ultimately, the integration of AI into classroom feedback is a work in progress—a blend of technology and human intuition. I firmly believe that AI is not here to replace teachers but to augment our ability to help students become better creators, thinkers, and problem solvers. Whether through rotational feedback cycles, guided AI prompts, or interactive chatbots, the key is to keep the human element at the center.
I’m excited to see where this journey takes us and remain open to refining these methods as technology evolves. I’d love to hear your thoughts or experiences if you’re navigating a similar path. How are you balancing machine precision with human empathy in your teaching?
Author’s Note: 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.