This article from The Economist has been sitting in my open tabs for quite some time. Not because it is sensational, but because it captures something that has been in the air for a while. The topic of educational technology and learning outcomes becomes essential here. The title alone is provocative: Edtech is profitable. It is also mostly useless. In other words, educational technology generates substantial profit for companies, but little learning.

That sounds sharp. Perhaps too sharp. But it forces a question we have avoided for too long: what has all this technology in classrooms actually delivered? Not in dashboards, licenses, or marketing brochures, but in real learning gains.

The story that opens the article sets the tone. A school in Kansas tried a new adaptive math program. It sounded promising: personalised, efficient, modern. Teachers hoped for quick progress. Students were given laptops. But after several years, the results were minimal. The software felt repetitive, boring, and above all, a source of distraction. Eventually, the laptops were put away. Pencil and paper returned to the centre of classroom practice.

This is not an anecdote about technophobia. It is a story about overconfidence.

Fifty years after computers began entering schools, some classrooms now seem saturated with technology. In many countries, almost every student has a personal device. This did not happen because of strong evidence, but because of a powerful narrative: technology would make learning more personalised, more efficient, and more motivating. Those who doubted this were quickly labelled conservative or afraid of change. The pandemic pushed many countries, regions, and schools across the final line of hesitation.

Meanwhile, research evidence has continued to accumulate, and the picture is more nuanced than slogans suggest. It is also more nuanced than the article presents it. Large reviews show that educational technology can sometimes be effective, especially for practice in narrow domains such as spelling, reading, and basic arithmetic. The meta-analysis of digital reading interventions in primary education cited in the article itself even reports small to moderate learning gains.

But those gains are not automatic. They are strongest on tasks that closely resemble what is practised in the program. When we look at broader standardised tests or at transfer to other contexts, the effects become smaller or disappear altogether. Results also vary widely between tools. Some programs work reasonably well, others hardly at all. What makes the difference is not the technology itself, but the instructional design and the way it is used in the classroom.

This is not an argument against technology. It is an argument against magical thinking.

A pattern becomes visible when we look more closely. Technology tends to work best in narrowly defined areas with clear right and wrong answers. Drill and practice can help automate basic skills. Digital tools can support students with specific learning difficulties. But when learning becomes more complex, as in reading comprehension, problem-solving, or critical reasoning, that advantage quickly fades. Learning is no longer an individual optimisation exercise. It is a social and cognitive process that depends on explanation, feedback, misunderstandings, and dialogue.

This is where the tension shows itself.

Many edtech initiatives do not begin with pedagogy. They begin with technology. First comes the platform, and only later do developers attach a pedagogical story to it. The central question quietly shifts. Instead of asking, what are our learning goals and which tool best supports them? We start asking, we have this system, how do we adapt our teaching to it? In this reversed logic, technology starts to shape the lesson.

Teachers see the consequences every day. Students work behind screens that function both as learning environments and as entertainment machines. Distraction is not an accident. It is designed into the system. Teachers spend their time policing firewalls and filters instead of guiding learning. Feedback turns into scores, progress into graphs, and understanding into percentages. What once lived in relationships and conversations becomes measurable and manageable, not because it improves learning, but because software makes it easy.

Gamification pushes this even further. Points, badges, and levels promise motivation, but they often redirect attention from understanding to performance. Students learn how to navigate the system, not how to master the content. Engagement becomes visible. Thinking disappears from view.

These patterns echo earlier warnings from international organisations. During and after the pandemic, schools rapidly expanded technology use, often without a clear pedagogical strategy. The debate rushed toward devices and away from education itself. The real issue was never whether students should use smartphones, but who controlled learning: teachers or platforms.

This may be the heart of the problem today. Technology does not simply enter the classroom. It increasingly structures the lesson. What started as support slowly turns into direction.

Some schools now push back. Not out of nostalgia, but out of experience. They return to pencil and paper and discover that these tools are not enemies of modern education. They send no notifications, collect no data, and demand no business model. And, they create room for slowness, explanation, and conversation.

This does not mean we should return to technology-free schools. It does mean we must ask harder questions. What is this technology actually for? Which learning goals does it serve? Which students benefit? And at what cost?

One of the most confronting lines in the article does not come from a researcher, but from a parent: imagine if all that money had gone to teachers. To professional development, smaller classes, better feedback, and more time for instruction. The thought unsettles because it exposes the core assumption. We treated technology as a solution to problems that were always pedagogical and organisational.

More than ten years ago, Bill Gates predicted that we would need a decade to know whether edtech truly worked. That decade has passed. Hundreds of billions have been spent. The picture is clearer now. Technology can support learning, but it cannot replace good teaching. It does not automatically improve education. It often makes it more expensive, more complex, and more tempting to hand difficult pedagogical questions over to software.

The real lesson may not be that edtech fails, but that learning resists automation. Learning is not an app. It is a relational process between people, shaped by mistakes, explanations, time, and attention. Forget that, and we end up with beautiful dashboards and modest results.

Edtech can be useful. But only if it does not take over the lesson. That may be worth remembering in today’s debates about AI.