Analysis of AI and Teacher Feedback in Grading Students’ Activities

Abstract

Artificial Intelligence has recently emerged as an important topic in the educational landscape and how it affects teachers’ professional knowledge. We investigated the alignment of Artificial Intelligence (AI) specifically ChatGPT in providing feedback on student output compared to teachers in the science classroom. Using an explanatory sequential mixed-methods design, we analyzed 47 student submissions assessed by both teachers and ChatGPT, guided by pre-established grading rubrics on two different student outputs of a certain university in Mandaluyong City, Philippines. Using Spearman’s Rho and Cohen’s Kappa we quantitatively explored the correlations of the evaluations of both AI and Teachers. It revealed weak correlations and low inter-rater agreement, indicating limited alignment between AI-generated and teacher evaluations, especially in subjective and interpretive components. Qualitatively, we interviewed science teachers, the result highlighted that, while AI feedback was efficient, consistent, and structured, it often lacked contextual depth, emotional tone, and pedagogical insight. Teachers valued AI’s ability to support routine tasks but emphasized the irreplaceable role of human judgment in assessing higher-order thinking and student-specific needs. We concluded that AI is a valuable supplementary tool rather than a replacement for educators. Our findings contribute to the ongoing discourse on ethical and pedagogically sound integration of AI in classroom assessment practices.