Since the beginnings of ChatGPT and with the public’s growing familiarity with generative artificial intelligence (AI), the education community has debated its promise and perils. Rather than wait a decade to conduct a post-mortem analysis of AI’s failures and opportunities, the Brookings Institution’s Center for Universal Education embarked on a year-long global study – a pre-mortem – to understand the potential negative risks that generative AI poses to students, and what we can do now to prevent these risks, while maximizing the potential benefits of AI.
After interviews, focus groups, and consultations with more than 500 students, teachers, parents, education leaders, and technologists in 50 countries, a careful review of more than 400 studies, and a Delphi panel, we find that at this point in its trajectory, the risks of using generative AI in children’s education eclipse its benefits. This is largely because the risks of AI differ in nature from its benefits (i.e., these risks compromise children’s fundamental development) and may prevent the benefits from being realized.
It’s not too late to bend the arc of AI implementation
We find that AI can potentially benefit or harm students, depending on how it is used. We all have the agency, capacity, and imperative to help AI enrich, not diminish, student learning and development.
- AI-enriched learning. Well-designed AI tools and platforms can provide students with a number of learning benefits if deployed as part of a comprehensive, educationally sound approach.
- Learning diminished by AI. Over-reliance on AI tools and platforms can endanger the fundamental learning ability of children and young people. These risks can impact students’ ability to learn, their social and emotional well-being, their trusting relationships with teachers and peers, and their safety and privacy.
To this end, we propose three pillars of action: Thrive, Prepare and Protect. Under each pillar, we present concrete recommendations for governments, technology companies, education leaders, families and everyone else who touches this issue. We urge all relevant stakeholders to identify at least one recommendation to implement over the next three years.