Several MPs (Charlene Chen, Kenneth Tiong, David Hoe and others) jointly questioned MOE on the safeguards and roadmap for introducing AI from primary school. Education Minister Desmond Lee answered four questions together, setting out MOE's "Four Learns" framework — learn about AI, learn to use AI, learn with AI and, most importantly, learn beyond AI. The calibrated roadmap: Primary 1–3 covers AI literacy only (awareness of AI's presence) with no work requiring direct AI use; from Primary 4, once pupils have foundational literacy, numeracy and executive-functioning skills, they may use purpose-built educational AI tools with built-in guardrails under teacher supervision (e.g. the writing assistant LEA and Maths LEA in the Student Learning Space), which are designed not to spoon-feed answers and to redirect off-task pupils "Socratically". A mandatory 10-hour "Code for Fun" programme (coding, computational thinking, AI basics) starts from Primary 4, with optional five-hour "AI for Fun" modules on generative AI and computer vision. Pupil data is anonymised and not used to train external models; commercial off-the-shelf tools require checks that inputs contain no personally identifiable information. On research, A*STAR's SG-LEADS longitudinal study (data collection from 2027) will track how children's AI use affects learning and well-being, alongside short-term school-based studies. Kenneth Tiong pressed MOE using Sweden's Karolinska Institute conclusion that "digital tools impair rather than enhance student learning" and Sweden's 2023 reversal of digitalisation (over €200m to reintroduce physical textbooks); Desmond Lee replied that Sweden had gone all-digital from age five and then fully reverted to analog, whereas Singapore takes a blended approach — keeping physical textbooks and teacher-centric teaching, treating AI as a tool, and crucially distinguishing general-purpose AI from purpose-built educational AI, since failing to do so would risk the wrong policy of not using AI at all. On parental opt-out: SLS classroom tools that are part of teaching cannot be opted out of, but externally-brought-in tools requiring consent will not be used without it. Eileen Chong raised the "equity paradox" — that more disadvantaged children with less adult supervision at home may lean on AI more, eroding the very cognitive development it is meant to support; the Minister called this an "evergreen" concern, to be met through internalised AI literacy and home-school-community partnership.
Key Points
- • MOE "Four Learns": learn about, use, learn with, learn beyond AI
- • AI literacy from P1; tool use from P4, supervised, education-only tools
- • SLS tools refuse spoon-feeding, prompt Socratically against offloading
- • On Sweden reversal: blended path, general-purpose vs educational AI split
Government Position
A calibrated, blended, teacher-supervised AI-literacy path that strictly separates general-purpose from educational AI.
Questioning Position
Cites Sweden's reversal to question the evidence for early AI introduction, warning of cognitive offloading and an equity paradox for disadvantaged pupils.
"It is far better to start AI literacy young and start getting our kids to use AI for learning in a highly scaffolded teacher supervised and of course, parents supervised way."
Original transcript excerpt
Dr Charlene Chen (Tampines), Mr Kenneth Tiong Boon Kiat (Aljunied), Mr David Hoe (Jurong East-Bukit Batok) and others asked the Minister for Education about safeguards, prerequisite skills, the Primary 4 entry point, the implementation roadmap, data transparency, procedural fairness and teacher support for AI use by primary-school pupils. Minister Desmond Lee answered the questions together. He set out the "Four Learns" — learn about AI, learn to use AI, learn with AI, learn beyond AI — and a calibrated, learning-sciences-informed roadmap: AI literacy from Primary 1 with no direct-use work in P1–3; from Primary 4, supervised use of purpose-built educational AI tools with built-in guardrails (the writing assistant LEA and Maths LEA in SLS), which refuse to spoon-feed and redirect off-task pupils Socratically; a mandatory 10-hour Code for Fun programme and optional AI for Fun modules. Pupil data is anonymised and not used to train external models. A*STAR's SG-LEADS longitudinal study will collect data from 2027 to track AI's effect on children's learning and well-being. Dr Chen asked about safeguarding independent thinking, vulnerable and SEN pupils; Mr Tiong pressed on productive struggle, metacognition and Sweden's reversal of digitalisation; Mr Hoe raised teachers asking pupils to use public tools like ChatGPT, age-restriction consent and letting parents try the tools; Ms Eileen Chong Pei Shan (NCMP) raised the equity paradox; Ms Cassandra Lee and Ms Elysa Chen asked about age-based safeguards, the AI "nutrition label" and IP/creator rights. Mr Lee replied that Singapore takes a blended approach — retaining physical textbooks and teacher-centric teaching, treating AI as a tool and strictly distinguishing general-purpose AI from educational AI — and that AI literacy, the home-school partnership and continued calibration are the core safeguards.