Minnesota researchers warn AI can blunt learning if used too soon
A University of Minnesota report finds experts broadly agree that early, unguided AI use can disrupt the thinking skills kids are still building - especially in elementary school.
What happened: University of Minnesota learning scientists, led by educational psychology professor Panayiota Kendeou, surveyed experts on how generative AI affects learning and published a new report on the findings. The team found surprisingly little high-quality research on the subject overall. But where evidence did exist, experts broadly agreed that careless, general-purpose AI use could cause what the report calls 'severe and one-sided' disruption to core learning processes, with elementary-age students flagged as most at risk.
Why it matters: The report pushes back on treating 'learning' as one simple thing AI either helps or hurts. Experts identified several specific areas at risk: committing new ideas to memory, higher-order thinking, collaboration, social-emotional development, and human connection. Kendeou put the stakes bluntly: get this wrong, and an entire generation's capacity for independent thought could be shortchanged, since education is society's main tool for preparing citizens to think for themselves.
How it works, plainly: The concern centers on sequencing. Skills like writing require students to first struggle through putting their own thoughts into words, and that struggle is what builds memory and reasoning ability. If a tool does that work before the skill exists, the student skips the step where learning actually happens. Kendeou said she would not give young students open AI access if the goal is building those skills from scratch, but would allow it once students already write and think critically on their own. At that point, AI can help stress-test arguments, expand ideas, or surface new resources.
The rollout: The report does not call for banning AI in schools, but for matching access to a student's existing skill level: independent practice first, AI tools later, once reasoning and writing foundations are solid. Kendeou argues decisions should be built around what's best for kids and teachers rather than around the technology itself. Because rigorous research is still thin, schools and districts are largely left to apply this sequencing logic on their own, without detailed, age-specific guidance to follow.