Automated Generation of Informatics Educational Tasks Based on Generative Artificial Intelligence Models

Automatic generation of educational tasks using artificial intelligence represents a promising direction for personalized learning. However, the quantitative effectiveness of generative models (GPT, Claude, and Gemini) in the context of secondary school computer science education remains insufficiently explored, particularly regarding the alignment of generated content with different difficulty levels and its readiness for direct classroom use. This study presents a theoretical analysis of educational task generation approaches, including rule-based, template-based, machine learning, and large language model (LLM)-based methods. A prototype task generator was implemented in Python using the OpenAI API, supporting three difficulty levels (easy, medium, and hard) and exporting tasks to PDF and DOCX formats. Experimental validation was conducted with a pilot group of 15 students, involving 45 AI-generated tasks evaluated against five criteria: topical relevance, difficulty-level appropriateness, grammatical correctness, logical consistency, and readiness for use without further revision. The results indicate a topical relevance of 93.3%, difficulty-level appropriateness of 86.7%, grammatical correctness of 100%, and 80.0% of the generated tasks being suitable for immediate classroom use without modification. The average task generation time was 2.1 seconds per task. Based on these findings, practical recommendations are proposed for computer science teachers: fully automated use is appropriate for generating multiple-choice and test questions, selective teacher review is recommended for programming tasks, and mandatory expert verification is advised for complex algorithmic problems. The findings should be considered preliminary due to the pilot nature of the study (n = 15).

artificial intelligence, generative models, large language models (LLMs), natural language processing (NLP), machine learning, educational task generation, computer science education, personalized learning, Python, OpenAI API, PDF/DOCX export.

Automated Generation of Informatics Educational Tasks Based on Generative Artificial Intelligence Models