Outcomes & evidence
How the JC SEN Portfolio measures a child's growth, what it reports and to whom, and one firm rule: every figure shows its as-of date and its source. No results are claimed on this page — findings are to be confirmed.
Draft content 6 item(s) marked "to be confirmed" by the programme office.
How growth is measured
The Platform follows each child. Growth is measured against the child's own past, never against other children. There are no scores, grades or rankings anywhere.
Progress is tracked across several areas — for example communication, social skills and self-management — not one academic mark. Each area follows the programme's framework. The final list of areas is to be confirmed.
The approach is strengths-based and non-diagnostic. Nothing in the record diagnoses, labels or scores a child. When data is missing, the Platform shows the gap instead of guessing.
What counts as evidence
Evidence is everyday, not a test. A teacher files small moments into the child's own folder. Over time these pages become the record that growth is measured from.
- Evidence item — A note, a photo or a short clip of an everyday moment (demo clips up to 20 seconds). It takes about fifteen seconds to capture. It can be tagged to a framework goal, and the teacher chooses whether the child sees it and whether it is shared with the family.
- Observation record — A teacher's dated professional note. If the AI suggested it, the teacher must accept it first; it then carries the label 'AI-assisted · teacher approved'. Rejected suggestions are discarded — nothing is saved until the teacher accepts.
- Goals and review cycles — Learning goals owned by people, with version history and approvals. The portfolio timeline is the evidence behind each review, including individual learning plan reviews.
- The portfolio — One timeline per child, mixing evidence, observations, insights and reports. It is the record the next teacher inherits.
- AI reads only the child's records — AI drafts read only the child's own records, the approved framework and approved reference material. Every summary links to the records it used. Missing data is shown, not filled.
What is reported, and to whom
Each person sees the same evidence in the way that is right for them. Reports are drafted from the folder, approved word by word by the teacher, and shared only with consent. Term reports are the main written output; goals are revisited on their review cycles. How often the programme reports, and to whom: to be confirmed.
How it is evaluated
There are two kinds of checking. First, every AI feature must pass a set of stage gates (G0–G7) before it is released, and the portfolio feature it relies on must be accepted first. Second, the programme's outcome evaluation — what is measured, with which tools, who runs it and when — is to be confirmed. Researchers will study only de-identified group figures, with small groups hidden.
- G0 ConceptThe feature is written down. The rule from the start: it suggests — it never decides.
- G1 Data readinessThe records it will read must exist and be trustworthy — AI is a layer over portfolio data, never the product.
- G2 Offline evaluationOutputs are checked away from any classroom. Real school data is never used for testing.
- G3 Human factorsTeachers can understand, correct, reject and trace every output.
- G4 Security and privacyThe Personal Data (Privacy) Ordinance is the floor. Access is by role and by school only.
- G5 PilotA limited trial before wider use.
- G6 Production with kill switchEach AI feature has its own off switch. The portfolio keeps working with AI off.
- G7 Quarterly reviewAI features are reviewed every quarter. Every AI call is in the audit trail.
Every figure shows date and source
This is the proof-strip rule. A number with no source is not allowed on any public page or report. Every figure carries an as-of date and a source line, like this:
*As of [date] UTC · Source: platform records (demonstration data)*
Anything AI-assisted or shared also shows where it came from: grounded on N evidence items · reviewed by the teacher · version. Each AI output records the model, prompt and version used.
Right now every figure in the demo is calculated live from a pre-loaded synthetic database. It shows how the Platform works, not results from any school.
Findings — to be confirmed
Placeholder — no findings yet. No outcome has been measured, and no result is claimed on this page. Nothing here says the Platform improves outcomes, saves time or changes workload — those things have not been measured.
When findings exist, each one will appear here in the proof-strip format: the finding · its value · as-of date · source · who reviewed it.
Until then: to be confirmed.
Common questions
Short answers to the questions people ask most about measurement.
Does the AI decide a child's support tier?
No. The evidence-and-goals record can help people decide a child's tier under the 3-Tier Support Model. People decide the tier. The AI never diagnoses, labels, scores or assigns a tier.
Are children compared with each other?
No. Progress is measured against the child's own past. There are no rankings and no scores of what a child cannot do.
Are the video 'insight' signals validated?
No — they are provisional. The clip signals in the demo are placeholder outputs marked 'model: spec pending' until the detection model is supplied by the programme's research partner. They are derived signals, not raw video, for the teacher's eyes only, and framed as a prompt for judgement — never a diagnosis.
Is this real school data?
No. Demonstration · synthetic data only — no real children. Every figure is calculated live from the pre-loaded demo database.
Who can see the numbers?
Families see only what a teacher has shared, under consent. NGO coordinators and researchers see de-identified group figures only, and small groups are hidden. Every share and every AI call is in the audit trail, by student id, not name.
When will findings be published?
To be confirmed. No dates are given on this page.