<![CDATA[This study examines the use of Artificial Intelligence (AI)-based formative feedback in the evaluation of science (IPA) learning among Primary School Teacher Education (PGSD) students. It addresses the limitations of conventional evaluation, particularly delayed feedback and the limited opportunities for conceptual correction in science learning. The study employed a mixed-methods approach with an explanatory sequential design; the quantitative component used a one-group pretest–posttest pre-experimental design without a control group, involving 40 students across two classes. The intervention used a web-based AI-driven digital evaluation application that generated generative and personalized feedback. Quantitative data were collected through pretest–posttest scores and a questionnaire, while qualitative data were obtained through interviews and system log analysis. The results show a statistically significant increase in learning-outcome scores after implementation, with a moderate improvement (6–7 points) and a medium effect size (Cohen's d = 0.65). The quality of AI-based feedback was categorized as good (mean = 3.13), particularly in terms of speed and usefulness, with a response time of less than five seconds. Qualitative findings indicate that immediate feedback supported conceptual understanding, helped identify misconceptions, and encouraged reflective learning behavior; however, limitations were found in feedback clarity and network stability. Because of the absence of a control group, this improvement cannot be fully attributed to AI feedback. The findings are preliminary and context-bound, yet indicate a positive contribution of AI-based formative feedback as an adaptive evaluation tool that needs to be tested with stronger designs.]]>
Copyrights © 2026
