The World Bank warned us about this in 2020 but we weren’t quite listening
By Precious Ebere-Chinonso Obi
Most of what I read about EdTech these days starts with AI: Adaptive tutors, AI lesson planners, chatbots that grade essays overnight. It’s easy to assume the core questions facing education technology are new questions, born sometime after ChatGPT entered the conversation.
So it caught me off guard, going back through a World Bank approach paper published in 2020, years before generative AI became a household term to find an argument that reads like it was written for exactly this moment. The paper is called Reimagining Human Connections: Technology and Innovation in Education at the World Bank, and its central claim is almost defiantly simple: education is, at its heart, about human connections, and any technology introduced into a classroom should be judged by whether it strengthens those connections or quietly erodes them.
The crisis the report was already describing
The numbers in the report are sobering on their own. Even before COVID-19 closed schools across the world, the World Bank was tracking what it called a global learning crisis: 53% of 10-year-olds in low- and middle-income countries couldn’t read and understand a simple story, a figure that rose past 80% in the poorest countries, with 258 million students out of school entirely. Then the pandemic hit, and the education of roughly 85% of children worldwide over 1.6 billion students was interrupted, accelerating a shift to remote and digital learning that had been building for years.
The report’s argument is that this acceleration was treated, almost everywhere, as a technology question, more devices, more connectivity, more platforms when it should have been treated as an education question first. Most EdTech investment in low and middle-income countries to that point had focused on access to devices and the internet, with far less attention paid to how that access was meant to improve actual teaching and learning. The result, the report notes plainly, is that EdTech’s impact on student performance has been mixed at best.
That line should sting more than it usually does. We’ve spent a decade treating access as the finish line, when the report was already telling us it’s barely the starting point.
The paper organizes its argument around five principles, and what strikes me reading them now is how directly they anticipate the AI-specific anxieties we’re having today.
Ask why comes first, and it’s the one I keep returning to: before adopting any tool, the question isn’t what the technology can do, but what specific educational problem it’s solving. It’s the same instinct I’ve written about in this column under a different name, technology should be a crane that extends what a teacher or student can do, not a shortcut that quietly does the thinking for them.
Design and act at scale, for all is a direct warning against a pattern we’re watching repeat with AI tools right now: technology rolled out to the best-resourced schools first, widening exactly the gap it claimed it would close.
Empower teachers rejects the idea that technology’s job is to work around teachers. The report is explicit that the strongest EdTech makes teachers more central to learning, not less, a useful corrective for anyone imagining AI as a way to need fewer, or less skilled, teachers in a classroom.
Engage the ecosystem and be data-driven round things out: no ministry of education can do this alone, and no decision should be made on hope rather than evidence about what’s actually working in a given classroom and context.
What makes this report worth returning to isn’t that it predicted generative AI specifically but rather, it correctly identified the underlying failure mode of EdTech investment in general chasing access while neglecting purpose and that failure mode is exactly what’s at risk of repeating itself as AI tools get layered onto education systems that, in many places, are still working through the access problem from the last technology wave.
The report ends on a single word: start. Not start anywhere, and not start with whatever tool is newest. Start with the question the whole paper is built around, what change in a child’s learning are we actually trying to make happen and let that question, not the technology, set the agenda.
Five years on, in the middle of an AI moment the report’s authors couldn’t have fully foreseen, that still feels like the right place to begin.
- Precious Ebere-Chinonso Obi, CEO of Do Take Action, is an independent consultant on edtech, climate change, public policy, and women’s procurement empowerment
