Pedagogy first, technology second: Cross-level relationships between teacher professional knowledge and student

Synthesis: Shen and colleagues examine how teacher professional knowledge shapes student learning in K-12 AI education, drawing on social cognitive theory and the TPACK framework. Using a multilevel structural equation model, they analyzed data from 46 secondary school teachers and 2,832 students to test cross-level effects of two knowledge components: teacher AI knowledge (TAIK, the technological/technical side) and teacher pedagogical AI knowledge (TPAIK, the pedagogical side).

Key Findings

• Technical AI knowledge alone was insufficient — without pedagogical AI knowledge, TAIK could even slightly diminish students' perceptions of AI for social good.
• Pedagogical AI knowledge (TPAIK) played the fundamental role, fostering both students' perceptions of AI for social good and their behavioral intention to learn AI.
• Neither TAIK nor TPAIK was directly associated with students' AI knowledge gains — the teacher-knowledge effects operated through perceptions and intention rather than direct knowledge transfer.
• Findings support the "pedagogy first, technology second" guideline: pedagogical AI knowledge should be prioritized over technical AI knowledge for excellence in K-12 AI education.

What this means for practice

• Teacher educators. Foreground pedagogical AI knowledge (TPAIK) over technical AI knowledge in Teacher Education coursework, since TAIK alone was insufficient and could slightly reduce students' perceptions of AI for social good.
• Faculty developers. Build AI professional learning around subject-specific teaching methods rather than tool training, because TPAIK carried the positive associations with students' perceptions and intentions.
• Administrators. Fund pedagogical capacity before equipment and technical certifications when building K-12 AI education.
• Researchers. Model teacher-to-student pathways as indirect and cross-level rather than assuming AI knowledge transfers directly, because neither TAIK nor TPAIK predicted students' AI knowledge.

Limitations

• Only 46 teachers were available for multilevel analysis; the authors state the study was plagued by teachers' limited responses, though 46 teacher-level units were deemed acceptable.
• Those 46 teachers taught 2,832 students in 144 classes across 29 schools, so teacher-level estimates rest on a small number of higher-level units.
• All constructs — teacher knowledge and students' SAIK, SAISG, and SBIAI — were measured with self-report Likert questionnaires, with no observed classroom or achievement data.
• Data cover a single academic year (September 2023 to May 2024); the authors note longitudinal dynamics were not captured and that students' informal AI exposure may introduce unmeasured mediators.

Citation

Shen, W., Chai, C.-S., Chiu, T. K. F., Yau, K. W., Meng, H., King, I., Wong, S., & Yam, Y. (2026). Pedagogy first, technology second: Cross-level relationships between teacher professional knowledge and student learning in artificial intelligence (AI) education. Computers and Education: Artificial Intelligence, 10, 100564.