Analysis of AI and Student Feedback Characteristics and Satisfaction in Mathematics Classes Utilizing AI-based Formative Peer Assessment

Formative assessment supports learners’ thinking and self-regulation but faces challenges due to teachers’ workload and the difficulty of providing immediate feedback. This study introduces an AI-driven formative peer assessment system (AIFPA), which generates answers and feedback as a virtual peer and assists in classifying responses and assigning peer evaluation tasks. Applied in three lessons with 108 eleventh-grade students, the results showed that AI feedback was stronger in structural coherence, while peer feedback offered greater emotional support, complementing each other. Students tended to emphasize diagnosis and improvement when assessing AI answers, but used praise and suggestions when reviewing peers’ work. Post-surveys indicated generally positive perceptions of AI-supported lessons, though reliability of AI feedback remained a concern. AIFPA demonstrates the potential of combining AI and peer assessment to enhance the feasibility and depth of formative assessment in classrooms.