Cognitively adaptive chatbot for students requiring guided support in mathematics: Design and interaction protocol

This paper presents Guidy, a cognitively adaptive educational chatbot designed for students with average levels of cognitive abilities, aimed at supporting independent learning in mathematics. The chatbot assists students in solving mathematical problems by using guiding questions, concise explanations, and control of key solution steps, with the goal of fostering independent thinking and developing self-checking habits. Guidy is implemented on the Poe platform, utilizing the GPT-4o-mini base model and a domain-specific knowledge base focused on the topic Proportions for 6th grade students. An interaction protocol is defined, encompassing task structuring, selection of an appropriate solution strategy, and final verification of results. Technical testing demonstrates correct system behavior and indicates strong potential for effective support of students’ independent learning.