UNC School of Education faculty member Jeffrey A. Greene, Ph.D., McMichael Professor and the School’s associate dean for research and faculty development, was one of three lead coordinating authors of a new consensus report on ways Generative AI can support and threaten learning in K-20 U.S. education.
Greene joined leading coordinating authors Panayiota Kendeou, Ph.D., of the University of Minnesota College of Education and Human Development and Nia Nixon, Ph.D., of the UC Irvine School of Education to convene 30 university-based researchers with expertise in learning, technology, and education. The group came to consensus on what aspects of learning could be enhanced by GenAI, which aspects were vulnerable to disruption by GenAI, and what these experts recommended to promote effective learning with GenAI.
The report is intended to provide evidence-based guidance on AI and learning for educators, parents, policymakers, and researchers.
“Educators, students, and families are struggling to determine whether and how Generative AI can support learning and education,” Greene said. “We created this report to give people an expert consensus on the best advice available.”
Matthew L. Bernacki, Ph.D., an associate professor at the School, was one of the 30 contributing experts.
The experts identified 65 distinct learning processes that could be affected by GenAI and then came to consensus on the vulnerability and enhancement potential for each. According to the report, their “process-level granularity reveals precisely where GenAI can enhance learning and where it can disrupt it.”
The group agreed on five key findings:
- Within the “acquire – practice – apply” cycle, GenAI works best in the middle.
True learning takes time and effort — two things GenAI tools are designed to reduce. GenAI shows most promise when used for practice to strengthen and consolidate understanding, not making initial learning acquisition or later application easier. - Experts recommend “using GenAI well” and “thinking well with GenAI.”
Using GenAI well includes AI literacy efforts focused on teaching students about how AI works, as well as integrations to enhance existing classroom routines. Thinking well with GenAI includes learning how to use GenAI to promote one’s thinking (rather than offloading thinking) and knowing when not to use it. - GenAI’s promise and pitfalls are most precarious in elementary grades.
Students must be able to learn independently before using GenAI. Developmentally, elementary students are in the earliest stages of building social-emotional, collaboration, and self-regulation skills, and have not yet acquired robust content knowledge to evaluate GenAI outputs for accuracy. - Human-like conversations can’t replace real human connection.
GenAI can produce human-like text, but it lacks personal experience, empathy, and perspective-taking, all of which underpin the authentic human relationships that education and life require. GenAI can augment what teachers do, but it cannot replace them. - Beware of simple summaries and hyperbolic headlines — strong evidence is still lacking.
Much of the existing research on GenAI in education does not meet high scientific evidence standards, making it hard to reliably evaluate effects. Reliable evidence requires long-term research with diverse student populations and thoughtful comparisons against existing, evidence-based practices.
Read the full report at z.umn.edu/gen-ai-report.