Special schools are eligible too

The AI for Empowering Learning and Teaching Funding Programme is open to government, aided, caput and Direct Subsidy Scheme schools, and the aided category clearly covers special schools. In other words, a school that admits students with special educational needs can apply for the same HK$500,000 per school, drawn from the Quality Education Fund and available until 31 August 2028.

Rather than waiting for a perfect 'SEN-only' tool to appear, start from the students' actual learning needs, choose a general tool that can be adjusted, and let teachers reshape how it is used. Whether a tool fits depends on how much room there is to adjust it and on the teacher's design, not on whether it carries a special education label.

AI uses that fit special educational needs

The uses below sit closer to the day-to-day goals of special education:

  • Individual pace: AI can adjust the difficulty of questions and the speed they appear at in real time based on how a student responds, letting a class with a wide range of ability each work at its own pace. The teacher sets the range, and the AI fine-tunes within it.
  • Multi-sensory input and output: present the same content as text, images and speech, or let students answer by voice or by choosing a picture instead of writing, lowering the barrier that a single channel creates.
  • Communication support: for students with difficulties in language or expression, AI can act as a go-between for communication, for example expanding a short input into a full sentence or turning text into speech, with the teacher checking the content.
  • A teacher-led way of working: a special education classroom is especially no place to leave a student alone with an open-ended chatbot. The safer setup is for the teacher to operate it while the whole class or a small group joins in the judgement, so the AI's output passes through the teacher as a filter.

Student data and consent: extra care

Students with special educational needs are mostly minors, and some cases involve sensitive assessment or health data, so personal data must be handled more carefully than in an ordinary classroom. Before signing, get the supplier to confirm four things in writing: where the data is stored, how long it is kept, whether it will be used to train models, and whether it complies with the Personal Data (Privacy) Ordinance.

On consent, it is best to explain the tool's purpose, the categories of data it processes and how to opt out to parents clearly in writing, and to keep a record of consent. For any registration tied to a student's identity, avoid having students sign up to public AI services under their real names, and use school-managed accounts or education tools that come with a preset framework instead.

Write goals that are concrete and measurable

A special school still has to meet the same set of commitments: introduce AI-assisted teaching in at least three subjects across two levels each, develop at least six teaching demonstrations, and hold open lessons, sharing sessions and student AI literacy activities. The bar for the commitments is not lowered to suit the school's circumstances, but the form the demonstrations take can fit the reality of special education.

When you set goals, avoid vague ones like 'raising interest in learning'. Use descriptions you can observe and record instead, such as the change in how often a particular student can complete a three-step word problem independently with a voice prompt, and keep three kinds of evidence: student work, lesson records and teacher reflections. This is what the report needs, and it gives your internal review something solid to go on.

The programme requires a separate account and records kept for at least seven years, so a special school should keep especially complete copies of its procurement and consent documents, so that a later review does not catch it off guard.

Steady the classroom routine before turning to AI

The grouping, switching between tools and multi-sensory activities of an AI lesson demand more of classroom routines than an ordinary lesson does. A special education class may be small, but once the transitions turn chaotic, the individualised benefits AI brings are cancelled out. In order, steady the routines first and add AI second, so the gains hold up.