AIEd Knowledge BaseAIEd KB
On this page

Research Article

From AI literacy to AI-integrated inquiry-based science teaching: the serial mediating roles of AI-TPACK and science teaching self-efficacy among Chinese pre-service science teachers

added 2026-09-16 · updated 2026-09-19

Synthesis: A survey of 548 Chinese pre-service science teachers tested how AI literacy becomes the intention to teach science through inquiry with AI. The study found a serial chain in which AI literacy strengthens AI-TPACK, AI-TPACK in turn builds science teaching self-efficacy, and that confidence predicts the intention to integrate AI — suggesting that general AI literacy is a necessary but not sufficient foundation for AI-integrated inquiry-based science teaching.

Key Findings

  • AI literacy is a strong foundation for AI-TPACK. Pre-service teachers who could understand, evaluate and use AI tools also saw themselves as more able to combine AI with science content and inquiry pedagogy; this was the strongest single link in the model.
  • AI-TPACK builds science teaching confidence. Participants who felt able to align AI with content and pedagogy reported higher science teaching self-efficacy, so AI-related pedagogical knowledge works as a source of confidence rather than only as a technical skill.
  • Confidence, not AI knowledge, is the strongest predictor of intent to integrate AI. Of the constructs tested, science teaching self-efficacy had the largest direct association with the intention to use AI in future inquiry-based science lessons.
  • AI literacy acts mostly through professional knowledge and confidence. The serial pathway from AI literacy through AI-TPACK and self-efficacy to integration intention was significant, alongside separate indirect routes through each mediator, with a smaller direct link also remaining.
  • Background differences did not change the picture. Adding gender, year of study, major, university type, AI use frequency, AI training and teaching practicum experience as controls left every focal path essentially unchanged, and none of the controls directly predicted integration intention.
  • Teachers felt capable but hesitant. Science teaching self-efficacy had the highest average score of the four constructs, while intention to integrate AI into inquiry-based science teaching had the lowest — capability beliefs alone do not turn into a commitment to use AI in the classroom.

Study Design & Method

A total of 548 Chinese pre-service science teachers, mostly second- to fourth-year undergraduates at normal and comprehensive universities, completed an online questionnaire distributed through the Wenjuanxing platform by convenience sampling. Their majors covered science education, physics education, chemistry education, biology education and geography education, and most had already taken educational technology, science teaching methods or subject-specific pedagogy courses. Four constructs were measured on seven-point Likert scales: AI literacy, AI-TPACK, science teaching self-efficacy (adapted from the STEBI-B personal science teaching efficacy subscale) and intention to integrate AI into inquiry-based science teaching. The original English scales were translated and back-translated into Chinese, reviewed by three experts for content validity, and piloted with 30 students before the main survey.

The data were analyzed with partial least squares structural equation modeling, chosen to explain variance in the endogenous constructs and to test multiple direct and serial indirect paths; the model accounted for roughly 35% of the variance in integration intention. Measurement quality was established first, with loadings, reliability and convergent and discriminant validity meeting conventional thresholds, and with common-method and collinearity diagnostics showing no serious threat. Background variables were then entered as controls in a robustness step to confirm that the findings were not driven by prior AI or teaching experience.

What this means for practice

  • Faculty developers. Embed AI training inside science pedagogy and inquiry design rather than treating general AI literacy as sufficient preparation — literacy reached integration intention only through AI-TPACK and confidence — so that pre-service teachers frame inquiry questions with AI, evaluate AI-generated scientific models, and judge the accuracy and bias of AI content.
  • Faculty developers. Build AI-TPACK through subject-specific modules on concept visualization, experimental simulation, data interpretation, and explanation construction, rather than addressing AI only in generic educational technology courses.
  • Instructors. Convert reported confidence into practice: science teaching Self-Efficacy scored highest of the four constructs while intention to integrate AI scored lowest, so add scaffolded AI-integrated teaching experience to teacher preparation.
  • Administrators. Extend AI-enabled reform beyond general AI or digital literacy: curricula and assessment should test whether pre-service teachers can apply AI within scientific inquiry, model-based reasoning, and ethical classroom practice.

Limitations

  • The cross-sectional design precludes strong causal inference: the serial chain is a theoretically grounded association, not a demonstrated developmental sequence.
  • Data come from self-report questionnaires, so social desirability and subjective perception may shape responses; no classroom observation or design-task evidence was collected.
  • The sample comprised only Chinese pre-service science teachers, limiting generalizability to other countries, regions or types of teacher education institution; generalizability across science disciplines (physics, chemistry, biology, earth science) was not tested; future multigroup comparisons are needed.
  • AI-TPACK was measured as a first-order construct of perceived AI-integrated pedagogical knowledge; it did not capture the sociocultural dimension of digitality emphasized by DPACK, nor DiKoLAN AI's subject-specific competencies in data processing, simulation and modeling, and scientific information evaluation.

Connected Concepts

  • AI Literacy — the independent variable; shown to be a necessary but insufficient foundation for AI integration intention.
  • Technological Pedagogical Content Knowledge (TPACK) — AI-TPACK, an extension of TPACK, serves as the central mediating professional-knowledge construct.
  • Inquiry-Based Learning — the target teaching context into which pre-service teachers intend to integrate AI.
  • Self-Efficacy — science teaching self-efficacy is the key psychological mediator of integration intention.
  • Science Education — the subject-specific domain in which AI-TPACK and efficacy are situated.
  • Teacher AI Competency — the broader competence framework the study's knowledge-and-confidence mechanism feeds into.
  • Professional Development — the context and site for the proposed pedagogical and curricular interventions.

Connected Articles

Citation

Zou, J., Li, N., Wang, X., & Du, J. (2026). From AI literacy to AI-integrated inquiry-based science teaching: the serial mediating roles of AI-TPACK and science teaching self-efficacy among Chinese pre-service science teachers. Frontiers in Psychology, 17, 1911909.

Metadata

Study designStructural Equation Modeling
DisciplineScience Education
LevelHigher Ed, Teacher Education
Intended audienceResearchers, Faculty Developers, Instructors, Policymakers
Learning and instructionInquiry-Based Learning, Pedagogies and Teaching Strategies, Self-Efficacy
Foundations of AI in educationAI Literacy, Teaching, Technological Pedagogical Content Knowledge (TPACK), Teacher AI Competency
SourceHistoryEditChat

Embed this page

Copy the code below to embed a chromeless version of this page in a learning management system or other website. The embedded view hides the site header, navigation, and footer.