In a Singaporean school, an algorithm took just three weeks to reveal that 18 out of 27 fifth-graders were confusing the order of operations in equations, despite all of them maintaining above-average grades.
An article in the journal Frontiers in Education synthesizes findings from a five-year study on the implementation of adaptive platforms across 47 schools in Europe and Asia. The authors tracked how artificial intelligence and learning analytics influence the daily routines of teachers and the academic outcomes of their students.
The system logs every error and response time, then automatically selects the next task to build on mastered material while simultaneously addressing specific knowledge gaps. Instead of a simple list of grades, the teacher sees a map of the class’s most common hurdles and receives a ready-made set of 10- to 12-minute micro-lessons. This approach mirrors how a physician uses monitoring data to fine-tune a treatment plan rather than simply prescribing a generic medication.
A randomized study across 12 schools showed a performance gain of 0.3 standard deviations over the course of a semester. The impact was more significant in mathematics and the sciences than in humanities subjects. However, the improvement was halved in classes with more than 30 students, and the platform frequently disconnected in schools with unstable internet. Long-term data on whether these effects are sustained after a year is not yet available.
While automating routine diagnostics reduces the time teachers spend grading assignments, it shifts their focus toward data interpretation and lesson design. At the same time, underfunded schools are being left behind: without stable connections and hardware, these algorithms simply do not work, widening the educational gap between different regions.
What will happen to the teaching profession when a significant portion of diagnostic work is handed over to algorithms?
