The Education Endowment Foundation (EEF) is launching a £2.5m research fund to measure how generative AI affects student learning. This initiative addresses an evidence gap as schools integrate tools like ChatGPT faster than researchers can track their educational impact.
Cognitive offloading concerns
The fund will investigate cognitive offloading, which occurs when students delegate tasks like recall, planning, reasoning, or drafting to AI chatbots. Researchers will commission studies on common classroom applications, including content summarization, essay planning, and feedback generation. These studies will measure outcomes on knowledge retention, working memory, problem-solving, motivation, and resilience.
The urgency of early evidence
A recent National Literacy Trust report indicates that more than two-thirds of 13- to 18-year-olds currently use AI for literacy and learning. This adoption outpaces any formal evaluation of its educational value. "The collection of robust evidence on their actual impact on learning has barely begun - especially for learners under age 16," said EEF chief executive Becky Francis. As educators evaluate these tools, structured guidance on the AI Learning Path for Teachers can help manage classroom integration.
Timeline and implementation gap
Research teams have until 30 June to register their interest, with studies beginning this year. The first findings will not be published until 2027. This timeline creates a practical challenge. Institutions must currently write and enforce AI policies without the longitudinal data the EEF is only now beginning to collect.
Why this matters for education professionals
School leaders and teachers cannot wait until 2027 to address AI in the classroom. You must base current policy decisions on the best available data regarding cognitive offloading and student engagement. Focus initial guidelines on transparent AI use and critical thinking verification, while preparing to update those frameworks as the EEF's findings emerge.
Your membership also unlocks:
