Original reading
An educational system that recommends practice questions can help a learner review a weak area, but its suggestion is not a diagnosis of ability. In our fictional classroom, a programme repeatedly offered Nam easy vocabulary because he had made several mistakes while the network was failing. The teacher checked his work and discovered that some answers had not been saved. She corrected the record and asked Nam to explain his reasoning. The example shows why educational AI needs human oversight and usable evidence. A system may detect a pattern in its data without understanding every cause. It should not receive unnecessary private information, and students should know how to question a result. Used carefully, automated feedback can support practice and free time for discussion. Used uncritically, it can reinforce a mistaken label. The goal is to assist learning, not replace judgement with an unexplained score.