AI, productive struggle, and the variable that decides everything

In Forbes last month, Ben Gomes, Google’s head of learning, said AI cannot solve education’s real problem: student motivation.

“Why you learn is a very human thing.”

My friend Johan S Roos makes the same case in his new book Human Magic, reaching back to Aristotle: “All humans by nature desire to know.” Curiosity, he argues, is the first of five distinctly human capabilities AI cannot replicate — and the one most quietly eroded when every question is met with an instant answer.

He calls it cognitive atrophy.

The World Bank’s education team put hard evidence behind that worry. A student can ace every assignment with AI and learn almost nothing — because the brain only forms lasting expertise through what cognitive scientists call productive struggle. A Turkish study found unguided AI access dropped student performance 17 percent.

In contrast, a Harvard study, with a well-designed tutor and prepared teachers, more than doubled learning. A Nigerian programme delivered in six weeks what normally takes 18 months.

Same technology, opposite results.

The variable is the teacher.

That is the fork in the road. Use AI to short-circuit the struggle and you hollow out the very thing learning depends on. Use it to give teachers back the hours they lose to admin, and you free them to do the work no machine can do — lighting the spark in a child sitting in front of them.

We built Efekta for the second path. AI backbone, human spark.

As a result, everywhere we have rolled out we have seen at least a 25% improvement in learning outcomes across millions of students. This is based on state run exams, where students have no AI support.

The question for governments is not whether to bring AI into classrooms. They have no choice. But they do need to make sure the solution they choose protects the productive struggle.

---

Originally posted on LinkedIn.