Towards a Standardized Learning-Risk Framework (L-RISK) for Annotating Educational Hallucinations in Generative AI: A Methodological Approach

Abstract

We introduce a standardized methodological framework (L-RISK) for annotating educational hallucinations grounded in Conceptual Change Theory and Cognitive Load Theory. The framework operationalizes hallucinations through a pedagogically informed error taxonomy, an ordinal learning-risk severity scale, and task- and prompt-aware evaluation procedures. We demonstrate feasibility through a pilot application in supply chain education.