Pre-service teachers’ evaluative judgments and the "Human-in-the-Loop" paradigm in AI-assisted assessment design

The paradigm for education has changed from a traditional teacher-tool approach to a hybrid intelligence framework due to the quick rise of Generative Artificial Intelligence (GenAI). This study examines the assessments of pre-service teachers at a public university in Malaysia as they utilise GenAI to generate subjective assessment items, including essay questions and interpretative exercises. Even though GenAI provides previously unheard-of efficiency for structural scaffolding and brainstorming, research shows that raw AI outputs frequently fall back on Lower-Order Thinking Skills (LOTS) and lack contextual empathy for learning demands. This study examines the ways in which aspiring teachers evaluate AI-generated content through a qualitative reflective thematic analysis of 35 teacher-designed artefacts. The results show that although AI provides a good "starting point" for draughting, human interaction is necessary to guarantee cognitive calibration, cultural sensitivity, and marking reliability. In order to close the gap between generic AI outputs and regional educational realities, pre-service teachers served as "cultural mediators," manually incorporating ideals like Adl (Justice) and Sejahtera into evaluations. In order to guarantee that automated content creation promotes deep conceptual learning rather than unthinking dependency, this study indicates that the teacher must continue to serve as the "essential architect" and maintain the pedagogical compass