No dedicated national R&D programme on AI pedagogy and child-AI interaction

Government itself describes evidence for AI tutoring as limited. EEF runs individual trials without a dedicated AI programme, UKRI has no directed programme on learning with AI, and the strongest classroom RCT evidence, such as the LearnLM UK trial of 165 pupils, comes from US labs studying their own products. Nobody in the UK funds longitudinal research on what sustained AI use does to children's learning, cognition, motivation or safeguarding; the DfE safety expectations even ask products to report cognitive offloading rates, a construct with little research behind it.

The state is deploying AI tutors to 450,000 disadvantaged children ahead of the evidence. Without independent UK research infrastructure, the knowledge base will be owned by vendors evaluating themselves, and long-run harms or null effects will surface only after national rollout.

A UKRI and DfE directed programme, an EEF for AI: longitudinal cohort studies with National Pupil Database linkage, standing rapid-trial infrastructure in volunteer school networks, and open evaluation of the government's own tutoring tools.

// Build together: Counterparty: volunteer school networks plus DfE NPD-linkage approval; philanthropic funder can seed before UKRI-DfE programme.

The state deploys AI tutors to 450,000 children ahead of evidence, and no UK body funds longitudinal research, leaving the knowledge base to self-evaluating vendors.