[GCM Reporter Vivien · November 8 · New York] When John Dewey and Paul Monroe joined Hu Shih and Kuo Ping-wen in founding China Institute in 1926, they could hardly have imagined that a century later, educators from both countries would gather at the same university to wrestle with a technology none of them could have foreseen — and to ask what it will do to the way children learn.
That was the question hanging over the 2026 U.S.-China Education Summit, which brought nearly 100 scholars of education, psychology and technology from China, the United States and Canada to Columbia University this weekend. In keynotes and panels alike, the conversation kept circling back to two concerns: how schools should teach and learn now that artificial intelligence is in the classroom, and how to protect the mental health of adolescents growing up surrounded by screens.
Peter Walker, co-chair of China Institute, who delivered one of the opening keynotes, framed the differences between the two systems as a matter of lenses. American education, he said in an interview during the summit, asks first whether children are being raised to become good, healthy, happy citizens. China’s system asks how many science and engineering graduates it can produce each year. “It’s a totally different view of the world,” he said. China’s primary schools, he noted, are reinforced by parents and teachers alike and are turning out STEM students in record numbers with strong mathematics scores, while the decentralized American system makes that kind of goal-setting far harder.
What concerns him more is the imbalance in who crosses the Pacific. By his estimate, roughly 300,000 Chinese students come to the United States each year, compared with about 5,000 Americans who go to China — leaving very few Americans with direct experience of the country. Language compounds the gap: English is a common second language among educated Chinese, while Mandarin’s tones and characters deter most Americans. Still, he has noticed well-educated American families hiring Mandarin-speaking nannies so their children grow up with the language. Mandarin may be among the fastest-growing languages in the United States, he said, though it is “starting with a very low base.”
That imbalance helps explain one of the summit’s concrete outcomes: a new U.S.-China Youth Education Exchange Scholarship, organized by the Asian Education Beijing Forum with participating universities, offering 500 places for American high school students to attend winter and summer programs in China.
The tools have arrived; the teaching has not
On artificial intelligence, the refrain heard most often was that the technology has reached the classroom before pedagogy has caught up. Garud Iyengar, director of the Avanessians Data Science Institute at Columbia Engineering, used his keynote to describe how AI is changing the production and acquisition of knowledge. By the afternoon, a roundtable in the AI forum put the question bluntly: should education in the age of AI be human-centered or technology-driven?
Participants acknowledged that schools have spent enormous energy debating whether to allow AI and how to police it, and far less figuring out how to teach with it — how curricula should change, what homework and exams can still measure, and what role teachers should play. That led to a deeper question: where does the ability to use AI well actually come from? The answer from many in the room was surprisingly traditional. Asking good questions, telling a sound AI answer from a fabricated one and building a complex argument all depend on the logical foundation laid by years of reading, writing and mathematics. Foundations, in this view, are not what AI replaces; they are what makes using it well possible.
To turn those discussions into sustained research, Paul Pu, a senior research scientist at the University of Toronto’s Ontario Institute for Studies in Education, launched the International Association for AI and Education Innovation. Its advisers include Huang Ronghuai of Beijing Normal University, who holds the UNESCO Chair on Artificial Intelligence in Education; the comparative education scholar Ruth Hayhoe; and M. Jamal Deen, vice president of The World Academy of Sciences. The association will focus on teacher professional learning and on evaluating educational AI products — two gaps that schools in both countries share: teachers unsure how to teach with the tools, and products that have rarely been independently tested.
Learning to fail
The second thread was adolescent mental health. Robert Siegler, a member of the National Academy of Education and a leading scholar of cognitive development, opened the youth forum with remarks on how children develop in a changing world. Xiaodong Lin, the Cleveland E. Dodge Professor at Columbia and director of its Center for Education Practice, Innovation and Competence, steered the discussion toward failure. Failed exams and lost competitions are part of every childhood, she argued, yet few schools treat coping with setbacks as something that must be taught. She launched a U.S.-China Youth Wellbeing Initiative inviting schools and researchers in both countries to spend three years turning research on how students, scientists and athletes experience setbacks into classroom curricula that can be piloted and evaluated.
The two threads meet there. As learning tools grow ever more frictionless, students may have fewer chances to struggle with a hard problem and keep thinking through failure — the very experience that nurtures both resilience and deep learning.
The summit will now alternate annually between the two countries, with the next meeting set for China in 2027. AI will likely have reshaped many more classrooms by then. But the questions left at Columbia this week — how children learn to think, and how they learn to fail — seem likely to remain where the next conversation begins.
