Under The Hood
In Your Hands
The Big Picture
The Solutions
The Methods
The Foundations
The Case
The Science
The Proof
The Arc
The People
The Promise
Written Reflections

Structured Prompts,
ML-Analyzed

Short, structured written prompts give every student a clear way in, and the same rubric-aligned machine learning pipeline that scores spoken responses, reads what they write. You get the answer, plus the thinking.

Dual Measurement

Two Kinds
Of Evidence.

The prompt below is a real NGSS science task. But a student can't reason through it well without perspective-taking: hearing a classmate's idea, weighing it, and proposing a way to test it. The science and the human skill live in the same response, so we can measure both from one authentic moment.

Playground · Perspective Reflection Step 2 of 3
Scenario · Science

A motor moves the model of the Moon along a track, and the models of the Moon and Earth are designed so they can spin on their axes.

What addition to the whole model is needed to show how solar eclipses occur?

Model of Earth Moon

Your group is figuring out why the Moon sometimes turns dark during an eclipse. A classmate says it's because the Earth's shadow covers the Moon.

01What do you think your classmate got right about the eclipse?
02Share your own idea about what's happening.
03What could your group try to find out who's right?
● Record
✎ Type
At first I thought my idea was better, but when Maya explained the shadow part I realized…
Response Feeds ML Scoring Submit →
Interface Recreation · Perspective-Reflection Prompt
The Scoring, Annotated

What The Models See In
A Written Reflection

Every reflection is scored the way a trained human rater would score it; against observable rubric levels, not sentiment. Here's a simulated response with the evidence the models pick up.

Written Reflection · Grade 7 · Perspective Prompt Simulated Example
At first I thought my idea was better, but when Maya explained the shadow part I realized she was thinking about where the Earth is, and I was only thinking about the Moon. Last month I would have just argued with her, but this time I asked her to draw it. Now I want to test it with the flashlight and the globe so we can see whose idea matches what actually happens.”
Perspective-Taking L4 Collaborative Reasoning L3 Curiosity in Action L3
Perspective Attribution

The student explicitly reconstructs a peer's reasoning ("she was thinking about where the Earth is") and contrasts it with their own frame — the distinguishing move between rubric levels 3 and 4.

Temporal Sequencing

"Last month I would have… but this time" marks self-referenced growth. The models flag this learning-progression framing as evidence of metacognitive monitoring, not just recall.

Counterfactual Proposal

The reflection closes with a testable next step tied to evidence ("see whose idea matches what actually happens") — forward-oriented reasoning scored under curiosity and collaboration rubrics.

Under the Hood

One Pipeline Behind
Every Response

Written reflections aren't a side channel. They run through the same measurement infrastructure as spoken responses: rubric-first, versioned, and benchmarked against human raters.

01

Rubric-Aligned Ordinal Models

Transformer-based models score open-ended student voice against observable competency levels — trained to score the way a trained human rater would, at scale.

02

Benchmarked Against Human Raters

Automated scoring is routinely compared to human raters using standard agreement metrics, with promotion thresholds and drift and subgroup-fairness monitoring.

03

Versioned and Audited

Model versions evolve over time and every score is traceable to the version that produced it. Reproducible scoring you can stand behind in a board room.

Next Step

See Written Reflections
In Your Framework

Through our Crosswalk methodology, every dashboard reads in your district's language: Portrait of a Graduate, CTE employability skills, community schools, or wellness.