Apr 28, 2025
Michael Harbaugh
Introduction
We’ve been listening closely to teachers and partners—and one theme kept coming up: the need for assignments that are pre-trained and trained before they ever reach students.
Today, we’re excited to introduce a new set of features designed to do just that. With these updates, teachers, PLCs, curriculum and instructional coaches, and district leaders can now create assignments or sets of assignments that are:
Aligned to the right level of rigor, based on your grading rubrics
Pre-trained with exemplar student work, so your EnlightenAI assistant starts in the right place
Ready to use and share, with consistent standards across classrooms and schools
This is all about one goal: ensuring the feedback you create with EnlightenAI is the highest-quality feedback possible.
What’s New: Pre-training Tools Built Into Assignment Creation
Think of these calibration tools as doing a scoring calibration exercise with your EnlightenAI teaching assistant before you hand it a stack of student essays.
The setup process will feel familiar. You’ll still begin with the basics of creating an assignment; rubric, prompt, submission length, grade level, reference texts, and so on. But now, you’ll also see an optional "Train your AI with examples" section where you'll be able to give your EnlightenAI assistant up to five Anchor Examples of student work:
Here’s how it works
Click in to see where you can add Anchor Examples across different performance levels:
Upload up to 5 student writing samples to give your assistant data to train with:
Run a pre-training check where EnlightenAI grades and provides feedback on each example, where you adjust rubric scores, rigor, and feedback style align exactly with your expectations:
From then on, whether you’re grading yourself or sharing the assignment with colleagues, you’ll know your EnlightenAI assistant is already calibrated and ready to deliver consistent, rubric-aligned feedback.
Why Pre-training Matters
Teachers often spend hours in rubric calibration meetings, working to align grading standards and ensure consistency. Our goal is to bring that process directly into EnlightenAI so that rigor, exemplars, and rubrics are all baked into your AI assistant before you start grading.
The result is not just that teachers can save time when someone shares an assignment with them that is ready-to-use and implement. It can also lead to greater accuracy and consistency. This can be crucial if the assignments are, for example, state test prep.
We ran a study comparing different grading scenarios:
Scenario 1: Two seasoned teachers independently scored the same essays.
Scenario 2: Two teachers used EnlightenAI’s traditional grading tools.
Scenario 3: Two teachers who first calibrated their AI assistant, then graded.
Scenario 4: A few other AI feedback tools were tested for comparison.
The findings were clear:
Administrator-trained (i.e. pre-trained assignements) produced the highest consistency levels; better than EnlightenAI’s already industry-leading accuracy and feedback quality.
Both EnlightenAI flows outperformed human-only scoring, and outperformed other feedback tools
Other AI tools landed closer to human-only levels (around .55–.57 in consistency).
Simply put, calibration plus EnlightenAI yields the most reliable feedback we’ve ever seen.
Note: if you want to learn more about quadratic weighted kappa scoring, you can read more here!
Conclusion
Pre-calibrated assignments are more than just a new feature for us; they’re a step toward solving one of education’s toughest challenges: delivering consistent, high-quality feedback aligned to standards.
With EnlightenAI’s calibration tools, teachers gain:
Confidence that their AI assistant is aligned with their expectations
Consistency across classrooms, teams, and districts
Capacity to focus less on norming and more on teaching
We’re thrilled to bring these tools to you, and we can’t wait to see the impact they’ll have on student growth and teacher collaboration.
