Grading Assistant
Assessment at scale, so every student is seen
An AI grading assistant, built in a real classroom on a real assessment framework.
What it is
SkillBridge – a working title – is an assessment assistant I built while teaching Language Acquisition at an IB Middle Years Programme school in Tokyo (2023-24). It ingests the MYP criteria and applies them, at scale, to real student work – written assignments (PDF, Word, handwritten scans) and spoken work transcribed from audio – producing a per-student report: transcript, summary, criterion-by-criterion scoring, and suggested questions for the face-to-face viva.
What it showed
Run across 120 students in five classes over a full year, it surfaced something uncomfortable. With a minute or two per assignment, human marking pattern-matches – rewarding confident, fluent, well-presented work and undervaluing less assured writing with real reasoning underneath. Guided by the full criteria, the system reads the whole thing and applies the same points to the first student and the last; the thousandth assessment is as thorough as the first.
The principle
This isn’t a criticism of teachers. The current system fails many students quietly, at scale – not through malice but through arithmetic: impossible marking loads, rationed feedback. Doing the first-pass assessment at scale changes what a teacher knows before they walk into the room, freeing their time for the human part – the conversation, the judgement, knowing the student. A collaboration, not a replacement.
Status
Built and run in a real school across the 2023-24 year, on work ranging from audio recordings to handwritten scripts.