A system for generating chatbots to support learning in the field of exact sciences using generative artificial intelligence models

The integration of generative artificial intelligence (AI) into education, especially for teaching exact sciences, represents an innovative opportunity to enhance student engagement and understanding. Chatbots such as ChatGPT can promote interactive learning, allowing students to explore complex scientific concepts through personalized support and real-time feedback. This approach not only transforms traditional pedagogical methods but also fosters deep curiosity and understanding among students. This study examines the task of increasing the degree of automation in creating AI-powered chatbots and their integration into the learning process for exact sciences, particularly mathematics, for school students in online educational settings. The relevance of the research is driven by the need to improve the success rate of educational activities under wartime conditions, eliminate knowledge gaps, bridge disparities in knowledge and skills among school students in exact sciences when preparing for further studies in higher education institutions, enhance the efficiency of independent learning in online educational settings, and stabilize the socio-emotional state of children. The purpose of this study was to develop a web platform for generating various types of chatbots using artificial intelligence models to improve the quality of school students’ preparation in exact sciences in online educational settings. The objectives include: analyzing the challenges and peculiarities of creating chatbots and preparing high-quality datasets; developing the structure and describing the functionality of the chatbot generation system; providing examples of creating various types of chatbots; conducting experiments to determine the effectiveness and cost of chatbot generation. The following results were obtained: A method to optimize the formation of datasets for chatbots was developed. A web platform for generating various types of chatbots was created, including assistant bots, bots for generating variants of mathematical problems and step-by-step explanations of their solutions, and testing bots using artificial intelligence models to improve the quality of school students’ preparation in exact sciences in online educational settings. Conclusions. The scientific novelty of the research is associated with improving the method of creating structured chatbot datasets while maintaining the thematic integrity of text and context, ensuring a more accurate selection of relevant information by chatbots for responding to user queries. The effectiveness of the proposed approach is illustrated through examples of creating an assistant bot, a bot for generating variants of mathematical problems and step-by-step explanations of their solutions, and testing bots, which demonstrated cost optimization and more efficient resource utilization. A significant advantage is the convenience of tools for creating and configuring chatbots, as well as their use through a showcase of ready-made chatbots in messengers.