In 2007, then Australian Prime Minister Kevin Rudd announced a ‘digital education revolution.’ His government allocated A$2.4 billion (A$3.9 billion in today’s money) to the project. A large chunk of that went to providing a laptop to every senior secondary student.

Australia was not alone. There was an international rush to adopt digital technology in classrooms. Principals were afraid their schools would be left behind. Eyewatering sums were spent on replacing pens and paper with keyboards and screens.

Now, many jurisdictions are suffering from buyers’ remorse. Research increasingly shows that digitising classrooms does nothing to improve learning. Overreliance on technology actively harms it.

Analyses of international test data have shown negative educational effects when digital technology is used intensively in classrooms. Neuropsychological research shows that children taught handwriting tend to become better readers than those who rely on keyboards. These are just two of many examples.

Guided by findings like these, many policymakers now seek to limit technology use in classrooms. Some countries, including New Zealand, have banned cell phones in schools. Others are reducing the amount of time students spend on devices.

Now we stand at the threshold of a new technological revolution in education – generative AI. Once again, techno-utopians are excited. And once again, educators fear being left behind.

When Rudd announced his revolution, there was little research on the educational impact of digital technology. This time around, there will be no excuse. There is already enough research on the educational harm generative AI can cause to justify extreme caution before letting it loose in schools.

In one study, researchers at Massachusetts Institute of Technology showed that AI use can reduce students’ brain engagement with academic tasks. Experimental participants asked to use AI to write essays were compared with others who used only internet searches or no technology at all. Brain scans of the AI users showed lower neural connectivity in areas associated with memory and creativity than the other two groups.

This comes as no surprise. Nearly two years ago, a New Zealand Initiative report, Welcome to the Machine, warned that allowing students to use AI for tasks they had not yet learned to do for themselves would inhibit learning.

Policymakers would do well to heed the lessons of history. Like any technology, AI can be used for good and ill. Its adoption should be guided by evidence, not by fear of missing out.