Navigating AI Essay Marking and Feedback

Summary: AI can do the heavy lifting on first-pass marking, but the feedback becomes far more powerful when students know it's been reviewed by a real teacher. The "stamp of approval" is what turns quick feedback into trusted feedback.

A practical solution to a practical problem

Incorporating AI into my teaching toolkit has been less about ushering in an education revolution and more about enhancing the existing educational framework.

When I first brought AI into my economics classroom, it was a practical (and intriguing) solution to a practical problem: too many essays to mark and not enough time.

The AI app I developed felt like a brilliant assistant, programmed to help me dig through the workload by producing detailed feedback on my students' essays. But I later realised the true value of this tool wasn't only the hours it saved — it was how well it meshed with my own teaching and feedback habits.

The "human in the loop" is the point

Here's the reality: the AI does the preliminary marking, but the feedback takes on extra value when I've had the chance to review and personalise it.

My students know this. They understand that behind the tech screen, there's still a teacher who is very much involved in their learning process.

That knowledge — that I'm overseeing and endorsing the AI's feedback — elevates its worth in their eyes.

This blend of AI efficiency and personal oversight hits a sweet spot:

• students get detailed feedback promptly
• but they also get feedback that's been refined by someone they trust

It turns out the teacher's stamp of approval matters. It signals that the comments aren't just automated responses (insightful as they may be) — they've been filtered through the same person who knows the student's work, habits, and potential.

What it looks like in practice

In practice, the process is straightforward.

With limitless energy, the AI churns through the initial assessment, highlighting areas for improvement based on careful prompting and exam-board alignment. Then I step in:

• tweak phrasing so it matches how I teach
• remove anything that feels generic or off-target
• add an extra nudge that fits that student's learning journey

It's not dramatic. No flashing lights. No orchestral music.

It's more like a quiet afternoon in the office, making sure what gets back to students will genuinely help them improve.

The impact on student engagement

The effect on engagement has been noticeable.

Knowing their work gets a final check from me seems to encourage students to take the AI feedback more seriously. They're more likely to:

• ask for a second attempt
• apply the critique
• improve iteratively rather than "do it once and move on"

It's gratifying to see a loop of continuous improvement taking shape, with the AI and me working in tandem to facilitate that cycle.

If teachers don't do this, there's a danger that AI marking gets seen as a clever gimmick (at least in 2024) — rather than a serious, trustworthy part of learning.

Final thought

Incorporating AI into my teaching toolkit has been less about replacing what teachers do and more about enhancing the framework that already works.

Students don't see the AI as a distant, impersonal judge. They see it as part of a collaborative effort — with their teacher ensuring the final advice is genuinely useful.

This would have felt like science fiction two years ago. But it's real, it's happening now, and it's exciting to watch students benefit from a smart blend of technology and personal teaching dedication.

Gary Roebuck is Head of Economics at Holy Cross New Malden, and creator of TeachEdge.ai.