Keyword: «artificial intelligence in education»

The article is devoted to the problem of the transformation of the role of generative neural networks in mathematics lessons: from a tool of hidden cheating to a didactic partner. Based on an empirical study (76 students in grades 6-10, Glazov, MBOU "Secondary School №. 3"), it is shown that the vast majority of schoolchildren already use neural networks to complete tasks. However, they do not blindly copy the answers, but check or redo them. A clear age trend has been revealed: among sixth graders, the proportion of "cheating" students exceeds the proportion of those who study with the help of AI, while by the 10th grade the situation is reversed. A methodological system of four strategies has been proposed and tested: "Critical reviewer", "Noisy task", "AI tutor" and "Methodical duo". It has been experimentally proven that the implementation of this system makes it possible to significantly increase the proportion of students who constructively use the neural network, compared with the control group. The methodological transition from prohibitive tactics to a partnership strategy is substantiated. The results can be directly used by math teachers at school.

The article analyzes the psychological consequences of the active introduction of generative neural networks (ChatGPT, DeepSeek, Alice AI) into the process of teaching English in secondary schools. The authors introduce and substantiate the working construct of “cognitive dependence”, describing the pathological delegation by a student of cognitive operations (translation, generation of grammatical structures, spell-checking) to an external algorithm. Changes in the structure of educational motivation are examined: the displacement of intrinsic interest by extrinsic reinforcement (points, levels) and a decrease in frustration tolerance. Based on theoretical analysis, three key symptom complexes are identified: the illusion of competence, atrophy of language intuition, and speech anxiety in “live” dialogue. Criteria for distinguishing between adaptive use of AI and clinically significant forms of dependence are proposed.