Exploring student anxiety and experience in performance-based assessments using AIvaluate: an LLM-augmented emotionally intelligent pedagogical AI conversational agent

Abstract Performance-based assessments such as oral presentations and viva voce exams are valued for their pedagogical benefits but can also be associated with heightened student anxiety, which may affect a range of learners and hinder student motivation and overall assessment experience. This study explores the potential of AIvaluate; an emotionally intelligent, LLM-augmented conversational agent, to provide a more supportive assessment environment for such learners. Using a counterbalanced quasi-experimental, within-subjects design with 35 pre-university students, we compared experiences of traditional face-to-face assessments with AIvaluate-mediated sessions. Emotional state reports, usability ratings and qualitative feedback were analysed to evaluate anxiety, system usability and learner perceptions. Results indicated that students experienced significantly lower anxiety during AIvaluate sessions, and usability scores for the tool were rated in the “good” range. Thematic analysis further revealed perceived advantages such as reduced social pressure, flexible pacing and ease of use, alongside limitations including technical challenges and a lack of dynamic human interaction. Importantly, while these findings suggest that AI-mediated PBAs may reduce student anxiety, which may be beneficial for learners who experience heightened levels of anxiety that exceeds the individual zone of optimal functioning, further investigation is necessary to ascertain whether this is ultimately beneficial for learning and whether AI-mediated assessments can improve student attainment. Future research could explore the value of hybrid models that combine AI conversational agent-based and face-to-face assessment formats, and examine the long-term effects of AI-led assessments on student performance and well-being.