AI-Powered Inquiry-Based Learning for Science Education Bridging Theory and Practice

Thesis Abstract

The integration of artificial intelligence (AI) tools within science education presents a timely opportunity to enhance inquiry-based learning (IBL) by enabling adaptive guidance, real-time feedback, and scalable classroom experiments, yet practical deployment often lacks empirical validation across diverse contexts. This study investigates how AI-enabled inquiry-based learning (AI-IBL) platforms bridge theoretical IBL constructs with classroom practice, aiming to determine their effects on student conceptual understanding, scientific reasoning, and engagement. The specific objectives are (1) to evaluate the impact of AI-IBL on students’ mastery of core science concepts in secondary-education settings; (2) to examine changes in scientific inquiry skills and metacognitive awareness; (3) to assess teacher facilitation practices and perceived practicality of AI-IBL integration; (4) to analyze how AI-IBL mediates student motivation and curiosity; and (5) to identify challenges and enablers of scalable implementation in resource-constrained schools. The study adopts a mixed-methods, quasi-experimental design grounded in constructivist and information-processing theories, notably Bandura’s social cognitive theory for self-regulated learning and Vygotsky’s sociocultural perspective on mediated action. A multi-site sampling frame comprises six secondary schools with a total student population of approximately 2,400, from which 24 classes (12 AI-IBL intervention, 12 traditional IBL control) totaling about 1,200 students are randomly assigned at the class level. Data collection instruments include (a) a validated conceptual diagnostic test administered as pre-, post-, and follow-up assessments (n?1,200), (b) the Science Inquiry Skills Rubric and the Metacognition in Science Inventory administered at three time points (n?1,200), (c) students’ engagement questionnaires (Likert-scale, n?1,200), (d) classroom observation protocols and teacher reflective journals, and (e) semi-structured interviews with 24 teachers and 60 students selected via purposive sampling. Validity and reliability are ensured through expert panel validation, pilot testing, Cronbach’s alpha analyses (? > 0.80 for all scales), and inter-rater reliability (? > 0.70) for observational coding. Data analysis proceeds with (i) ANCOVA and hierarchical linear modeling to test differences in achievement and inquiry skills while controlling pre-test scores and demographic covariates; (ii) structural equation modeling to explore mediation effects of self-regulated learning and motivational constructs on achievement outcomes; (iii) thematic analysis of interview and reflective journal data to identify affordances, constraints, and teacher-student interaction patterns, with triangulation across sources. The analysis plan also includes a multigroup SEM to compare effects across gender and prior achievement strata. It is anticipated that AI-IBL will yield statistically significant gains in conceptual understanding (effect size d ? 0.35–0.50), higher-order inquiry skills, and metacognitive awareness relative to traditional IBL, with more pronounced benefits for students of lower prior achievement. Expected findings include enhanced formative assessment through AI-generated prompts, dynamic hypothesis testing capabilities, and personalized scaffolding that align with situated cognition principles. Potential challenges anticipated are digital equity, teacher workload, algorithmic transparency, and fidelity of implementation, which will be examined through process data and alignment checks. The study contributes to knowledge by providing robust, multicenter evidence on the efficacy and mechanisms of AI-augmented IBL in science education, offering a validated framework for scalable adoption that integrates theoretical guidance with practical workflow designs. Implications for policy and practice include recommendations for inclusive teacher professional development, curriculum alignment, data governance, and sustainable procurement of AI-IBL tools. The main conclusion is that AI-powered, inquiry-centered interventions can meaningfully enhance science learning and inquiry competencies when designed with explicit alignment to inquiry pedagogy, equitable access, and teacher-supportive ecosystems. Recommendations emphasize iterative design-research cycles, ongoing teacher professional development focusing on AI literacy and inquiry facilitation, investment in equitable technology access, and further longitudinal studies to determine long-term retention of conceptual knowledge and transfer of inquiry skills to novel domains.