Evaluating the Impact of LLM Feedback through Self-Determination Theory

This paper presents an in-depth study evaluating the impact of Large Language Models (LLMs) on students’ situational motivation and perceived competence through personalized feedback. While generative models offer unprecedented scalability in education, their psychological impact on learners remains underexplored. Through an experimentation, students received personalized feedback generated by LLM or by human tutors, calibrating them on the students’ psychometric profiles, including personality traits, learning styles, and self-efficacy. Grounded in Self-Determination Theory (SDT), pre- and post-test analyses reveal that LLMgenerated feedback significantly influences the perceived competence, which serves as the primary mediator for situational motivation. The study highlights that the pedagogical efficacy of LLMs depends not only on linguistic quality but also on the strategic consideration of learners’ baseline self-efficacy, suggesting a shift toward AI-based tutoring systems that are sensitive to the psychological state of the learner.