AI Feedback Assistant: MyEssayFeedback
Submitter: Marc Watkins, U of Mississippi
——————————————————
The experiment:
Feedback is a cornerstone of any writing course; it also manages to be one of the most labor-intensive activities in education. Many faculty regularly struggle to provide students with formative feedback in a timely way because of increasing teaching loads and class sizes. Students who do receive feedback often don’t get personalized responses because of the above reasons.
In the Spring of 2023, I began working with Eric Kean and Anna Mills on piloting MyEssayFeedback.ai, an app where educators built specific assignments using their own prompts and students then submitted writing and received real-time feedback from an LLM. I do not feel comfortable allowing students to engage in any LLM without providing them with basic AI literacy about what these systems are and how they function, so using any type of automated feedback should be grounded in an ethical and pragmatic framework. It should also not be used as a tool to eliminate human interaction, but one to augment it. Allowing students to receive formative feedback on writing to learn activities from a generative system poses risks, yes, but also potentially allows faculty to spend more time conferencing with students one-on-one and freeing up time for them to develop more personalized commentary on student essays.
Results:
In a fully online course, students completed low-stakes writing assignments and received AI, peer, and instructor feedback. Many appreciated the AI’s rapid response time compared to waiting for peers or the instructor. The volume of AI feedback also exceeded limited human commentary. Most found the generative feedback useful for revisions, though some disliked receiving critique from a “soulless” machine.
Given the range of responses, I think there may be room for both human and generative feedback in education. Balance is key—faculty need to be in the driver’s seat when deciding what aspects of their labor they want to automate, and students also deserve a voice in whether the feedback they receive is human, generative, or a mix of both. There are clear gains a learner can receive from timely, personalized feedback, but finding a middle ground beyond speed, efficiency, and vivid human feedback means navigating a labyrinth of personal, political, and cultural preferences that are converging around generative AI. Faculty and students find themselves in the middle of this storm, and we need more discourse and careful experimentation to understand how generative feedback impacts education.
Relevant resources: https://docs.google.com/document/d/179H_kLwEWImHA95slX3gYzd2xrBo-FDesj6bwm3cf7Q/edit?usp=sharing
Contact: mwatkins[AT]olemiss[DOT]edu
At our institution we have been told it would violate FERPA for an instructor to upload a student essay into ChatGPT. In your assignment, the student themself upload their own student work into the AI tool you utilize; I suppose that makes it “OK” according to FERPA. This assignment gives me pause, should there be an ethical concern about requiring students to upload their student work in an AI tool for an assignment? Should the assignment require the ability for students to opt-out of gathering feedback from AI if they are not comfortable uploading their work into an AI tool? Should students be explicitly informed about the privacy and data collection practices of the AI tool, including log-in credentials, tracking, analytics, etc? Wondering your thoughts?