🧠 More Than Fast: Research Shows How Teacher-Guided AI Can Revolutionize Grading

Grading open-ended questions is essential—but exhausting. If you’ve ever left a stack of student worksheets untouched over the weekend, you’re not alone.

We all want deeper learning, but it comes at a cost: slow, inconsistent, and draining grading. Fortunately, new research shows that AI can dramatically reduce that cost—without removing the teacher’s judgment.

A new academic paper from Carnegie Mellon University, titled "Avalon: A Human-in-the-Loop LLM Grading System with Instructor Calibration and Student Self-assessment", confirms a truth we’ve believed at Assessain from the beginning: the best grading systems keep the teacher in control.

Let’s unpack what this study says—and how it backs up the very core of Assessain’s mission.

👩‍🏫 The Grader's Dilemma

We know open-ended questions lead to better understanding and deeper thinking.

But they’re also:

Time-consuming to grade
Prone to inconsistent scoring
A cause of delayed feedback

AI offers a solution, but many teachers are understandably skeptical. What if the AI gets it wrong? What if it overrides your professional judgment?

The Avalon study gives us a clear answer: AI works best when it works with you—not instead of you.

🤖 Human-in-the-Loop: Why Teachers Are Irreplaceable

The study evaluated a system where teachers and AI grade together. Here’s how it works:

The AI makes grading suggestions
The teacher reviews, calibrates, and can override anything
Over time, the AI learns how that teacher grades

The result? 85% less time spent grading, and teachers kept full control.

This is exactly how Assessain works:

You print and distribute your worksheet as usual

Students complete it on paper

You scan and upload the answers

The AI suggests scores, based on your rubric

You can tweak, approve, or re-grade any answer

📏 Calibration Is Not a Bug—It’s a Feature

The paper introduced a key concept: "Rubric Calibration."

Before the AI takes over, the teacher grades a few examples. This teaches the AI what "correct" looks like—from your perspective.

Assessain follows this same best practice. As you review the first few graded submissions, you're helping the AI understand your style. This ensures:

More consistent scoring
Better alignment with your goals
Less rework later

⏱️ Up to 85% Time Saved

The Avalon study found that human-in-the-loop grading reduced active grading time by over 85%.

Imagine grading an entire assignment in under 30 minutes, with full control and insight. That’s what Assessain is built to offer.

🔍 Beyond Speed: A Window into Student Thinking

The study didn’t just save time—it helped uncover student misconceptions that might have gone unnoticed.

At Assessain, when an answer is ambiguous or unclear, it gets flagged for your review. That’s not a bug—it’s a chance to diagnose learning gaps.

This turns Assessain into more than a grading tool—it becomes a teaching assistant that shows you where students are struggling.

💬 Feedback That Actually Reaches Students

Avalon included a student self-assessment step to boost feedback engagement—because in traditional settings, most feedback is never even read.

While Assessain doesn’t yet include student self-assessment, our goal is the same:

Fast turnaround = more relevant feedback
Students get their work back while they still remember it
Teachers can respond while the topic is fresh

✅ Why This Matters for Assessain Users

The Avalon paper validates everything Assessain is built on:

🧠 Keep teachers in control
🛠️ Use AI as an assistant, not a replacement
⏳ Save hours of grading time
🔍 Use grading to drive better teaching

We’re proud that our approach isn’t just intuitive—it’s backed by rigorous academic research.

🚀 Try It Yourself

Want to experience teacher-guided AI grading, built on the latest research?

Try Assessain today and see how it can help you:

Save time
Stay in control
Deliver better feedback
Spot where students struggle

👉 Start now

📚 Citation:
Armfield, D. et al. (2025). Avalon: A Human-in-the-Loop LLM Grading System with Instructor Calibration and Student Self-assessment. In: Cristea, A.I., Walker, E., Lu, Y., Santos, O.C., Isotani, S. (eds) Artificial Intelligence in Education. AIED 2025. Springer, Cham. https://doi.org/10.1007/978-3-031-99267-4_14