An upcoming online event with Dr. Seiji Isotani, a leading scholar at the intersection of artificial intelligence, learning sciences, and educational innovation. Dr. Isotani is Professor of Computer Science and Learning Technology at the University of São Paulo, and serves as Faculty Director of the Learning Analytics and Artificial Intelligence Program at the University of Pennsylvania.

The session will focus on “AI in Education Unplugged”—a practical and equity-oriented approach to designing AI tools and interventions for contexts where digital infrastructure, connectivity, and skills are limited. In his contribution to the OECD Digital Education Outlook 2026, Dr. Isotani traces why “minimum infrastructure” assumptions have long constrained the field, and reframes the core question: how can the benefits of AI reach learners and communities where infrastructure is close to non-existent?

A central insight in this discussion is that the technology strategy must start from real-world constraints. Dr. Isotani notes that in 2022, mobile phone access was widespread globally, including in low-income contexts, while stable internet access remained far more limited—only 15% in low-income communities in the data he cites—implying that equal connectivity could take roughly a century at current trajectories. This is why “AI Unplugged” emphasizes designs that can function with basic mobile devices and intermittent connectivity, rather than assuming always-on broadband or one device per student.

We will also explore concrete implementation pathways. Dr. Isotani describes how, after the COVID-19 disruption, his team supported large-scale writing recovery in Brazil by enabling teachers to photograph students’ handwritten essays, upload them whenever internet access was available (often via a single connected location at school), and receive analysis plus recommendations through teacher-facing dashboards. He reports that outcomes were analyzed at substantial scale—around half a million students across 1,500 municipalities—with measurable benefits, and that similar approaches are being extended to mathematics and other subjects.

Finally, we will discuss quality, safety, and pedagogy: how recommendations can be grounded in evidence-based instructional strategies, why teacher/parent/mentor “proxy users” may be essential in low-device settings, and how generative AI changes (and sometimes complicates) assessment when preserving students’ original mistakes matters.


AI in Education Unplugged - Dr. Seiji Isotani