How this platform implements the EDB “Guide on Applying AI in Teaching”

Last updated: September 2026

In 2026 the Education Bureau published Appendix 2, the “Guide on Applying Artificial Intelligence in Teaching in Primary and Secondary Schools” (the “Guide”), setting out principles for using AI across learning, teaching, and assessment. This page explains how the QTS Science platform implements them by design.

How the eight principles are implemented

Education first, technology as support
AI is only an assistant on this platform: values education is woven into reading design, with corresponding EDB values are recorded from the passage content and question design, and left untagged where nothing naturally corresponds, while nurturing work — including final grading — always stays with teachers.
Use technology wisely, guard against over-reliance
A built-in submit gate blocks answers copied from the passage, careless responses, and submissions identical to classmates, reducing the chance of students outsourcing their thinking to AI or coasting; AI comments are provisional, encouraging genuine effort over machine dependence.
All teachers on board, driving pedagogical change
Teachers prepare lessons with AI assistance (translation, question setting, rubric drafting, building interactive simulations), offloading repetitive work to AI and freeing time for teaching design and student interaction.
Manage risks, cultivate AI literacy
Before first using AI-marked features, every student reads and accepts a short agreement: AI can be wrong, AI feedback is provisional, and they are responsible for what they submit. AI-marked results are clearly labelled “provisional — teacher finalises” in the student interface.
Reflect, improve, and keep pace professionally
Usage and cost analytics support panel-level review; the AI grading prompts are version-managed and continuously revised based on real use.
Understand deeply, apply cautiously
Every piece of AI-generated teaching content (simulations, article translations, questions, marking suggestions) reaches students only after a teacher personally reviews it and presses “publish”. AI-generated Weekly Science content passes an independent fact-check, and items with insufficient evidence are dropped before students ever see them. Teachers can also trial-run AI grading in a student preview before assigning.
Comply with the law, learn safely
What is sent for AI marking is limited to the passage, the questions, and the student's answer text — no names, student IDs, or other identifying details are included. Sign-in is restricted to the school email domain with role-based permissions, sandboxed content execution, and audit logs. See the Privacy Policy for details.
Education equity, diversity and inclusion
Readings ship with Cantonese and English text-to-speech and adjustable font sizes for diverse learning needs; all AI features are provided centrally by the school — students never need to pay for any AI tool themselves.

How the curriculum implements values education

Reading passages use the twelve values of the EDB Values Education Curriculum Framework (perseverance, respect for others, sense of responsibility, national identity, commitment, integrity, benevolence, law-abidingness, empathy, diligence, unity, filial piety) as the curriculum tag system: values are recorded where they naturally correspond to the passage content and question design, and left untagged where nothing naturally corresponds; the teacher dashboard provides values coverage analytics for panel-level review.

Values are carried by real content, not preached: passages bring values out of authentic science contexts, questions assess understanding of the content and analysis of the values, and no item measures a student's stance or attitude — value tags exist for teacher planning and coverage analysis, never to judge whether a student holds or agrees with a value.

Kept separate from National Security Education: the NSE series carries no values tags, and the two curriculum lines stay clearly apart in tags, question design and assessment logic (see “How the curriculum implements National Security Education” under Related below).

Learning, teaching, and assessment

Learning

Students never touch generative AI tools directly; AI supports them through simulations, read-aloud audio, and instant feedback. The system does not allow submitting AI-generated content as one's own work and cross-checks classmates' answers for duplication.

Teaching

AI helps teachers produce and refresh course resources, run first-pass marking, and analyse class performance; teachers hold final authority over all AI content.

Assessment

AI provides instant provisional marking with bilingual feedback, which teachers review and finalise (a teacher's finalisation overrides every AI result); the anti-carelessness rules are a hard system gate that AI scores can never overturn.

Under planning (deliberation in progress)

The following directions — tied to the Guide's “Assessment as Learning” and AI-literacy goals — are under deliberation with the panel and will ship when ready: individual student learning reports (helping students reflect on their own strengths and weaknesses), and a school-based AI literacy learning unit.

How this platform implements the EDB “Guide on Applying AI in Teaching”

Last updated: September 2026

How the eight principles are implemented

Education first, technology as support

AI is only an assistant on this platform: values education is woven into reading design, with corresponding EDB values are recorded from the passage content and question design, and left untagged where nothing naturally corresponds, while nurturing work — including final grading — always stays with teachers.

Use technology wisely, guard against over-reliance

A built-in submit gate blocks answers copied from the passage, careless responses, and submissions identical to classmates, reducing the chance of students outsourcing their thinking to AI or coasting; AI comments are provisional, encouraging genuine effort over machine dependence.

All teachers on board, driving pedagogical change

Teachers prepare lessons with AI assistance (translation, question setting, rubric drafting, building interactive simulations), offloading repetitive work to AI and freeing time for teaching design and student interaction.

Manage risks, cultivate AI literacy

Before first using AI-marked features, every student reads and accepts a short agreement: AI can be wrong, AI feedback is provisional, and they are responsible for what they submit. AI-marked results are clearly labelled “provisional — teacher finalises” in the student interface.

Reflect, improve, and keep pace professionally

Usage and cost analytics support panel-level review; the AI grading prompts are version-managed and continuously revised based on real use.

Understand deeply, apply cautiously

Every piece of AI-generated teaching content (simulations, article translations, questions, marking suggestions) reaches students only after a teacher personally reviews it and presses “publish”. AI-generated Weekly Science content passes an independent fact-check, and items with insufficient evidence are dropped before students ever see them. Teachers can also trial-run AI grading in a student preview before assigning.

Comply with the law, learn safely

What is sent for AI marking is limited to the passage, the questions, and the student's answer text — no names, student IDs, or other identifying details are included. Sign-in is restricted to the school email domain with role-based permissions, sandboxed content execution, and audit logs. See the Privacy Policy for details.

Education equity, diversity and inclusion