Brookings Study Warns: Risks of Classroom AI Outweigh Benefits for Students

A comprehensive new report from the Brookings Institution's Center for Universal Education delivers a sobering assessment: the dangers of using generative artificial intelligence in K-12 education currently eclipse its potential benefits. After analyzing interviews with students, parents, and educators across 50 countries and reviewing hundreds of research articles, researchers concluded that unchecked AI adoption "undermines children's foundational development" with "daunting" but fixable consequences.

The Cognitive Cost of Convenience

The study identifies cognitive decline as the most urgent threat. When students use tools like ChatGPT to complete assignments, they initiate a "doom loop of dependence," leading to reduced critical thinking and problem-solving abilities. "When kids use generative AI that tells them what the answer is, they are not thinking for themselves," explains lead author Rebecca Winthrop. This cognitive off-loading—different from past technologies like calculators—is turbocharged by AI's ease. As one student bluntly admitted: "It's easy. You don't need to use your brain."

Evidence shows students relying on AI exhibit measurable declines in content knowledge retention, analytical reasoning, and creativity. These deficits could create adults unprepared for complex decision-making, with Winthrop noting: "They're not learning to parse truth from fiction or understand what makes a good argument."

The Double-Edged Sword of AI Assistance

While risks dominate, the report acknowledges specific benefits:

• Language & Writing Support: Teachers reported AI helps language learners through adjustable reading levels and private practice. For writing, it can spark ideas and assist with grammar—but only when supplementing human instruction, not replacing it.
• Teacher Efficiency: Educators saved nearly six weekly hours using AI for administrative tasks like creating quizzes, translating materials, and drafting lesson plans.
• Equity Potential: In Taliban-controlled Afghanistan, AI delivered WhatsApp lessons to girls barred from classrooms. The technology also aids students with learning disabilities like dyslexia.

However, Winthrop warns AI simultaneously "massively increases existing divides." Free AI tools accessible to underfunded schools are often less accurate than premium versions available to wealthier districts. "This is the first time in ed-tech history," she notes, "that schools must pay more for factual accuracy."

Emotional Development in Peril

Perhaps the most startling findings involve emotional health. Students increasingly form relationships with AI companions, with one survey showing 1 in 5 high schoolers have engaged in "romantic AI relationships." Chatbots designed to agree with users prevent essential social-emotional growth. Winthrop illustrates: "If a child complains about chores, a chatbot says, 'You're right.' A friend would say, 'Dude, that's normal.'" This echo chamber impedes resilience and empathy development. As one expert observed: "We learn empathy not when perfectly understood, but when we misunderstand and recover."

Pathways to Protection

The report urges immediate action:

1. Redefine Learning Goals: Shift classrooms from transactional tasks toward curiosity-driven exploration to reduce over-reliance on AI.
2. Rebuild AI Design: Tools should challenge students' preconceptions rather than reinforcing them.
3. Establish Co-Creation Hubs: Governments should broker collaborations between educators and developers, following the Netherlands' model.
4. Mandate AI Literacy: National standards for teacher and student AI competency, as implemented in Estonia.
5. Prioritize Equity: Ensure marginalized communities aren't left with inferior AI tools.
6. Government Regulation: Policymakers must safeguard student cognition, privacy, and well-being.

With generative AI barely three years old, researchers framed their analysis as a "premortem"—a proactive examination of potential failures. The time to implement safeguards, they argue, is now: "The remedies are evident, but the window for effective intervention is closing."