Teaching UML Modeling in the Age of Generative AI: Combining Manual Modeling, Prompt Engineering and Model Evaluation

Abstract

Generative Artificial Intelligence is rapidly transforming software engineering practice, including the creation of Unified Modeling Language (UML) artifacts. While Large Language Models can generate UML diagrams from natural language descriptions, their implications for software modeling education remain insufficiently understood. This paper presents an educational framework for AI-supported UML modeling developed based on 25 years of experience teaching the Object-Oriented System Analysis course, analysis of student UML projects, and insights obtained by the author from projects investigating AI-assisted software artifact generation. The framework combines four complementary learning activities: manual UML modeling, (meta)prompt engineering, AI-assisted model generation, and systematic model evaluation. Rather than treating AI as a replacement for modeling activities, the framework positions AI-generated artifacts as objects of comparison, reflection, and critical assessment. The paper concludes by outlining directions for future classroom deployment and empirical evaluation of the framework in software engineering courses.