Supporting elementary teachers’ pedagogical decision-making through machine learning–based behavioral diagnostics

Abstract

This study investigates how elementary school teachers interpret and use a machine–learning–based behavioral diagnostic system to support pedagogical decision-making regarding students’ non-cognitive competencies. Employing a descriptive-analytical approach grounded in the principles of human-centered artificial intelligence, data were gathered through questionnaires, semi-structured interviews, and document analysis. The findings indicate that teachers were able to clearly interpret the diagnostic outputs and deemed them relevant to their classroom practices. Additionally, there were high levels of perceived usefulness in identifying students' learning needs (M = 4.25) and supporting instructional planning (M = 4.22). Over 80% of teachers reported using the system to differentiate learning activities and adjust instructional strategies. Moreover, the diagnostic system served primarily as a decision-support tool, assisting teachers in validating their professional intuition, designing more responsive differentiation strategies, and engaging in deeper reflective practice regarding the development of students' non-cognitive skills. Importantly, teachers maintained full professional autonomy by selectively interpreting and contextualizing system recommendations based on their students' knowledge. These findings suggest that machine learning-based behavioral diagnostics can effectively enhance teachers' pedagogical reasoning when designed within a human-centered framework. Rather than replacing teacher judgment, the system improved teachers’ awareness of students’ motivation, engagement, self-regulation, and perseverance, thereby facilitating more informed and responsive instructional decision-making in elementary education.