If AI Can Provide Feedback, Why Does Human Feedback Still Matter?

Imagine showing a draft of an important email to an AI tool. It points out a sentence that might sound abrupt and suggests a warmer version. You can use the suggestion, change it, or decide the original wording is right.

That kind of quick feedback can be useful. It raises a question for education, though: If AI can give students immediate feedback, why does human feedback still matter?

Part of the answer is that feedback is more than a comment on what a student has already done. Good feedback helps a learner understand where they are, where they are trying to go, and what they might do next. In their influential review of feedback research, John Hattie and Helen Timperley described these as three central questions: “Where am I going?”, “How am I going?”, and “Where to next?” They also cautioned that feedback can have different effects depending on what it focuses on and how students receive it. John Hattie, Helen Timperley, 2007

AI can help answer some of those questions. It can flag a confusing sentence, explain a concept another way, or give a student another chance to practice. Because it is available on demand, students may be able to get guidance while they are still working instead of waiting until an assignment is returned. But feedback only helps when a learner can make sense of it and decide what to do with it.

A response can be quick, detailed, and still miss the mark. AI might suggest changes that make a student’s work sound more polished without helping the student understand the underlying idea. It might focus on grammar when the main issue is the argument. Or it might recommend a revision that the student cannot evaluate because they do not yet understand the subject well enough.

Research comparing AI-generated and human feedback offers a reason for care. In a study examining feedback on student essays, trained human evaluators generally provided higher-quality feedback than ChatGPT. The quality of AI feedback also varied with the quality of the essay, declining for stronger essays. The study focused on secondary students’ writing, so we should be cautious about generalizing its results to every higher education course. Still, it shows why fluent feedback should not automatically be treated as sound feedback. ScienceDirect

Human feedback can draw on knowledge of a student’s goals, previous work, course context, and the standards of a discipline. An instructor may recognize that a student’s confusing paragraph comes from a misconception discussed in class, or that a technically correct answer does not yet show the reasoning the assignment is meant to develop. That context can help the instructor decide which response is most useful now.

The point is not that human feedback is always better. It can be late, overly general, or difficult for students to act on. AI may help address some practical limits by offering students more opportunities to ask questions, test revisions, and receive suggestions. Research on AI-supported peer feedback in university language courses, for example, found that students using an AI-supported approach produced higher-quality peer feedback than students in the comparison group. That suggests AI can help students engage more thoughtfully with one another’s work when it is part of a structured learning process. ScienceDirect

The educational question, then, is not simply “Who can give more feedback?” It is “What kind of feedback helps this student learn to make better decisions?”

Sometimes AI can provide a useful first response. A student might ask it to identify unclear parts of a draft, explain why a solution did not work, or generate questions to consider before revising. The student still needs to judge whether the response is accurate and relevant. Faculty can help students develop that judgment by discussing what useful feedback looks like, modeling how to evaluate it, and asking students to explain the changes they make.

That shifts the student’s role from receiving comments to working with feedback. What did I understand from this suggestion? What do I agree or disagree with? What will I change, and why? What do I want to ask my instructor or a classmate?

Those questions build capabilities that matter well beyond a single assignment. Students learn to assess their own work, weigh advice, and take responsibility for revision. AI can contribute to the feedback process, but it cannot make those learning goals automatic.

Human feedback matters because it can connect a student’s work to the larger purpose of learning and because a person can respond to what the student needs to understand next. AI can expand the opportunities for feedback. Educators still have to help students learn how to interpret it, question it, and act on it.

If AI can respond to students individually and adapt to their progress, the next question is what happens when that personalization changes the experience of learning together. If AI can personalize learning, what is the value of a classroom?

Continuing the Conversation

Series 1: AI Is Exposing Existing Problems ✓ Completed
Series 2: What We Do About It ✓ Completed
Series 3: Cultivating Human Thinking in an AI World ✓ Completed
Series 4: Learning Alongside AI ✓ Completed
Series 5: Preparing Students for an AI World ✓ Completed
Series 6: The Purpose of Education in an AI World
Current Post (3 of 8): If AI Can Provide Feedback, Why Does Human Feedback Still Matter?
Next Up: If AI Can Personalize Learning, What Is the Value of a Classroom?