WHW-COMMISSION-2026-08-25-FAMILY-PRODUCTS · ready-to-commission
Commission the human review that AI cannot replace
Recommended order. Everyday Hero scenario adjudication first, Foxy early-years and language review second, and Primary Power generator validation third. An observed caregiver-child pilot follows only after expert content review closes critical defects.
Download the machine-readable brief · See the internal red/purple review
Independence and acceptance boundary
AI-assisted internal review prepared and prioritised this work, but it is not a substitute for the named human reviewers. Reviewers must disclose conflicts and review the supplied content hashes.
- Approval applies only to the exact reviewed content hash; a material content change invalidates it.
- A reviewer may cover more than one lens only where their qualifications genuinely cover each lens.
- Child or caregiver testing requires informed adult consent, age-appropriate assent, data minimisation and a written safeguarding and stopping protocol.
- No engagement may be described as certification, efficacy evidence or institutional endorsement unless the commissioned method and reviewer explicitly support that claim.
Three workstreams
Priority 1 · everyday-hero-quest
Scenario-level safeguarding, digital-safety and primary-education adjudication
Review unit: Every scenario prompt, response option, explanation and caregiver framing, identified by stable ID and content hash.
First pass: All 31 paused and two tier-3 not-queued scenarios first, followed by a risk-based pass over 51 active scenarios.
Required: Singapore primary educator; child-safeguarding professional; digital-safety specialist; accessibility or neurodiversity-informed reviewer; caregiver usability reviewer
Gate: No paused or tier-3 item returns to a child-facing queue until every required lens for that item records a hash-bound outcome.
Priority 2 · foxys-learning-patch
Activity-level early-years, speech-language, multilingual and accessibility review
Review unit: Every guided activity, prompt, feedback string, audio or visual cue, caregiver note and off-screen mission.
First pass: Three Minutes Together and every Mandarin or Malay vocabulary and feedback path, then the full activity inventory.
Required: early-childhood educator with experience around ages 18 months to 3 years; speech-language professional; native Singapore Mandarin or Simplified-Chinese reviewer; native Singapore Malay reviewer; accessibility tester covering keyboard, screen reader, reduced motion, sound-off and touch use
Gate: Do not make developmental, language-learning or accessibility-conformance claims until the relevant specialist has reviewed the exact artifact.
Priority 3 · primary-power-quest-sg
Question-bank and generator validation for P1-P3 English, Mathematics, Chinese and Science
Review unit: A reproducible stratified sample plus all generator rules, answer keys, explanations and difficulty labels.
First pass: At least 30 reproducibly seeded items per level-subject cell, every template family, all edge cases and every claimed difficulty transition.
Required: Singapore P1-P3 English educator; Singapore P1-P3 Mathematics educator; Singapore P1-P3 Chinese educator; Singapore P1-P3 Science educator; accessibility and caregiver-usability reviewer
Gate: Do not claim curriculum alignment, assessment validity, readiness, mastery or calibrated difficulty from generator tests alone.
Shared deliverables
- Item- or activity-level decision: approve for bounded practice, revise, or withhold.
- Reason for every revise or withhold decision, with an authoritative source or professional-practice basis where applicable.
- Exact replacement wording for high-priority defects when feasible.
- Reviewer name, role, relevant qualification or lived or professional expertise, organisation if any, review date, conflict disclosure and reviewed artifact hash.
- A short residual-risk note explaining what the review does not establish.
Optional follow-on study
Observed caregiver-child usability pilot
Start only after: The three expert content-review workstreams have closed critical defects.
Purpose: Observe comprehension, stopping behaviour, caregiver burden and accessibility barriers; not measure learning efficacy.
Minimum method: Small moderated sessions with a written protocol, informed adult consent, child assent where appropriate, no sensitive profiling, predefined stopping rules and a de-identified issue log.