Evaluation of AI generated illustrations for learning. A Practical Framework for BA Theses

Generative Artificial Intelligence (GenAI) increasingly supports the creation of visual learning materials, yet standardized evaluation methods for AI-generated educational illustrations remain limited. This working paper presents a practical quality framework developed specifically for bachelor's thesis projects supervised by the authors. We introduce four core mandatory criteria—Conceptual Representation, Visual Correctness, Legibility & Labeling, and Overall Educational Usability—each rated on a three-point scale (poor, moderate, excellent). Optional supplementary criteria allow researchers to adapt the framework to their specific domain and research focus. The document further outlines procedures for defining target audiences and learning objectives, offers guidance on effective prompt engineering, and specifies a standardised documentation protocol using structured tables with mandatory metadata fields to ensure reproducibility and cross-project comparability. Recommendations for open licensing and publication are included.