Statistical report on the pedagogical experiment on the use of IIS GPT during mathematics and physics lessons (2019-2024)

This report presents the results of a five-year pedagogical experiment conducted by the International Innovative School on the integration of artificial intelligence (AI) tools in the teaching of mathematics and physics in secondary education. A total of 2,500 students participated in the experiment, which aimed to assess the long-term impact of AI-supported learning on student outcomes, engagement and satisfaction. The study assessed twelve core parameters across three domains: A: feedback quality, including response timeliness (A1), readability (A2), amount of additional explanation (A3), visual clarity (A4), and inaccuracies (A5); B: cognitive and performance impact, such as motivation (B1), error correction behaviour (B2), and test performance (B3); C: engagement and personalisation, including classroom participation (C1), satisfaction (C2), and degree of feedback individualisation (C3). Over the course of the experiment, AI-based tools - including adaptive learning systems, chatbots, and visual explanation platforms - were systematically integrated into classroom delivery and homework support. The results show consistent improvements across almost all parameters. In particular, 82% of students reported timely and helpful responses (A1), while test performance (B3) increased by an average of 21%. Engagement (C1) and satisfaction (C2) metrics increased significantly, especially when feedback was tailored (C3). Teachers observed fewer misunderstandings (A5) and more targeted clarifications (A3). The findings suggest that with proper pedagogical integration, AI technologies can meaningfully enhance STEM education by supporting personalisation, motivation and clarity in learning.