The Difference Between Correcting and Connecting

At its core, feedback is a relational act. When a student submits a piece of work, they are not simply handing over a text for error correction. They are offering up their intellectual effort, and in return, they are looking for validation that this effort has been seen, understood, and valued by an educator they respect.

I first came across this dynamic when reading Corbin and colleagues’ (2025) framework for making sense of generative AI feedback in education. They introduce a helpful distinction by separating feedback into two categories. Traditional, human-led feedback is described as “recognitive“—meaning it is grounded in a shared human connection, built on mutual respect, trust, and an understanding of how vulnerable it feels to learn something new. In contrast, they label AI feedback as “extra-recognitive“, which is simply a transactional transfer of information that is not relational.

AI can come across as excellent at the technical side of marking, like spotting errors and suggesting fixes, but it cannot provide the social spark that motivates a student to care, nor should it do so. When prompted and steered correctly, large language models are indeed able to provide useful feedback. Yet, algorithms are forever absent from relations, caring and context. Since encountering that framework, further evidence and insights have continued to emerge that align perfectly with this foundational idea.

Why High-Quality Feedback Isn’t Always Enough

We now have some empirical evidence of this divide in practice. A recent study compared human teacher feedback to generative AI that was specifically programmed to “think” through problems step-by-step (Farrokhnia et al., 2026). The researchers found that the algorithm actually produced structurally superior, more comprehensive feedback than the educators.

Yet, remarkably, this higher technical quality did not lead to better student revisions. The AI simply scaled its critique based on the surface quality of the initial draft, whereas the human teachers adaptively tailored their guidance to support the specific developmental needs of each learner. The quality of the human feedback drove actual student improvement; the technical perfection of the AI feedback did not.

This mirrors earlier findings from the MIT Media Lab, which demonstrated what happens when we separate the words of feedback from the human delivering them (Morris & Maes, 2026). Researchers provided students with identical, AI-generated feedback on a creative coding task. Half the students were told a human teaching assistant had written the comments, while the other half were informed the feedback was generated by an algorithm.

The students who believed a real person was engaged with their work spent significantly more time on the task, iterated on their work more frequently, and produced richer outputs. The only difference was the perception of human presence, yet that perception was the absolute catalyst for student effort.

Reclaiming the Human Element

These findings reinforce a much broader philosophical shift currently happening in the sector. As the recently published Manifesto for Feedback in the Age of Generative AI declares, we must protect feedback as a deeply relational, messy and ethical practice (Winstone et al., 2025).

We cannot reduce educational encounters to mere transactional efficiency. As the manifesto rightly points out, dumping more detailed comments onto a student does not automatically equate to better learning. True feedback engagement requires time, care and a collaborative dialogue between educator and learner that an algorithm simply cannot replicate.

The relevance of generative AI, therefore, depends entirely on how we draw the line between marking and teaching. AI can serve brilliantly as a low-stakes testing ground: a space where students can refine the mechanics of their work without fear of human judgment. It can handle the heavy lifting of proofreading and structuring. But the final mile of feedback must remain fiercely human. Generative AI can correct academic work with unprecedented efficiency, but it requires a human educator to actually connect with the student.

References & Further Reading

  • Corbin, T., Tai, J., & Flenady, G. (2025). Understanding the place and value of GenAI feedback: A recognition-based framework. Assessment & Evaluation in Higher Education, 1–14. https://doi.org/10.1080/02602938.2025.2459641
  • Farrokhnia, M., Latifi, S., Papadopoulos, P. M., Hogenkamp, L., Gijlers, H., Khosravi, H., & Noroozi, O. (2026). Generative AI offers more, but students revise less: Comparing the effects of teacher and AI feedback on student essay revisions. International Journal of Educational Technology in Higher Education, 23(1), 6. https://doi.org/10.1186/s41239-026-00579-9
  • Morris, C., & Maes, P. (2026). Same Feedback, Different Source: How AI vs. Human Feedback Shapes Learner Engagement (arXiv:2602.11311). arXiv. https://doi.org/10.48550/arXiv.2602.11311
  • Winstone, N., Gravett, K., Noble, C., Nicola-Richmond, K., Bearman, M., Jensen, L. X., Jones, A., Corbin, T., de Kleijn, R., Gabelica, C., Kainth, R., Poobalan, A., & Reedy, G. (2025). Manifesto for feedback in the age of generative artificial intelligence (p. 845268 Bytes). figshare. https://doi.org/10.6084/M9.FIGSHARE.30195568