Beyond the Easy Button: AI Feedback Loops That Strengthen Student Writing and Judgment

Abstract

One of the central teaching challenges in the age of AI is that students can now use generative tools to bypass the struggle through which writing, reading comprehension, and critical thinking develop. This session shows how instructor-designed AI feedback can give students the repeated coaching they need to write more, revise more, and think more carefully, without turning AI into the author of their work.

In BUS 3200: Professional Communication with AI, students complete four progressively more challenging writing assignments using an AI-supported revision process: draft first without AI assistance, submit their own writing to a Feedback Machine, revise based on adaptive feedback, resubmit until satisfied, and reflect on their writing and process. Feedback Machines is a Canvas-integrated tool that helps instructors build assignment-specific AI feedback tutors aligned with their own rubrics and teaching priorities, with the feedback persona tunable from a kindly writing tutor to a no-nonsense professional editor. The tool provides feedback; it does not draft, rewrite, revise, or give a final grade to student work. It also provides a scoring mechanism that allows students to track their improvement across the categories of the instructor-defined rubric.

The session will share preliminary evidence from approximately 60 students across four assignments, including drafts per assignment, percentage of text revised between submissions, score improvement across the semester, and student reflections on their composing and revising processes. Across the semester, students wrote more, revised more substantively, and developed more precise language to describe their writing choices than a typical course workload would support.

The most useful findings, however, emerged from the friction. Students were often overwhelmed by feedback density, encountered contradictory or imperfect AI suggestions, plateaued in revision, and sometimes discovered they had misunderstood the assignment only after the Feedback Machine flagged the issue. Many initially treated the feedback transactionally before gradually learning to use it as instruction. Some also reflected pointedly on moments when they had used AI as a writing shortcut elsewhere and what that cost them as learners.

Rather than presenting AI as a solution to teaching, this session frames instructors as designers of learning experiences. Participants will see examples of assignment design, feedback prompts, student reflections, and revision evidence. They will leave with a transferable model for using AI as an editor, coach, and mirror—one that helps students develop their own writing and thinking rather than outsourcing those abilities to AI.

Presenters