CATEGORY-BASED PROMPT CONSTRUCTION FOR ARTIFICIAL INTELLIGENCE IN EDUCATION: A PEDAGOGICAL PROMPT ENGINEERING MODEL AND WEB TOOL

Abstract

This article proposes a pedagogical model, and a web tool built on it, for constructing prompts by educational material category so that teachers and researchers can work effectively with artificial intelligence (large language models). The relevance of the study stems from the fact that the quality of a generative model's output depends directly on the quality of the instruction, while many educators lack prompt-writing skills. The aim was to develop a solution that produces ready-to-use, structured prompts for eleven types of educational material: lesson plans, research projects, scientific articles, 3D models, multimedia, animation, virtual materials, assessment, didactic games, graphics and software development. A design-based approach was applied, combining prompt engineering principles with the TPACK framework. As a result, a five-part prompt structure (role, task, context, requirements, output format), a set of category-specific parameters, a mechanism for adapting prompts to the target type of AI tool, and a model-assisted prompt refinement step were developed. The tool requires no installation and runs via a link on computers and mobile devices. The article discusses the didactic potential and limitations of the model and outlines a protocol for its empirical validation.