Problem
Feedback workflows need consistency and speed, but the system cannot substitute model output for teacher judgment or expose private instructional records.
Project Brief
Instructional AI Workflows
A synthetic Precalculus free-response demo converts teacher-defined rubrics and observations into structured evaluations, a reviewer packet, student-facing feedback, and remediation plans without grading raw work automatically.
Overview
Teacher-controlled AI assistance
The teacher-controlled rubric-to-feedback workflow makes evaluation evidence, review status, and remediation outputs inspectable. Its offline, deterministic human-in-the-loop design requires no live LMS or AI service.
The public demo uses synthetic inputs and pre-authored teacher observations; it does not automatically grade raw student work.
Current Artifacts
- Precalculus FRQ synthetic rubric demo
- Synthetic learner inputs
- Pre-authored teacher observations
- Structured evaluation JSON
- Teacher reviewer packet
- Reviewed student-facing feedback
- Remediation planning output
- Public safety rules for synthetic submissions and rubrics
System Story
From rubric evidence to reviewed language
Approach
JSON rubrics, synthetic responses, and pre-authored teacher observations feed deterministic workflow logic that separates evidence capture, draft generation, teacher review, release language, and remediation planning.
Result
The Precalculus demo produces structured evaluation artifacts, review-required cases, a reviewer packet, approved feedback, and a rubric-organized remediation plan from synthetic inputs and pre-authored teacher observations.
Lesson
The strongest use of AI here is a controlled review pipeline, not automatic scoring. Evidence and teacher approval remain explicit at every release boundary.
Demo Evidence
What the demo shows
Synthetic FRQ input
The source repository includes fake learner responses and a teacher-defined rubric for a Precalculus linear model task.
Structured evaluation
The demo organizes pre-authored teacher observations into rubric-level evidence, scores, private teacher notes, and review-required flags for synthetic learner examples.
Reviewed output
Only approved release notes are shown as student-facing feedback; teacher-facing review details remain separate.
Remediation planning
The demo groups rubric evidence into teacher-facing next steps for model setup, solving process, and context interpretation.
Projects Qualification
Supporting project · prototype claims · qualified use
Clear supporting example of governed AI-assisted instructional workflow design; less technically deep than the top three campaign projects.
Safety Boundary
No private instructional records
Public demos must use synthetic submissions, synthetic rubrics, fake course identifiers, and public-safe sample outputs. Do not publish real student submissions, feedback, grades, comments, Canvas IDs, assignment IDs, course IDs, API tokens, or teacher-only review artifacts.