Smart Adaptive Feedback Systems for Science Concept Mastery in Classrooms

Thesis Abstract

The rapid integration of digital technologies in science classrooms has intensified the demand for intelligent guidance that can dynamically tailor feedback to students’ evolving conceptual understandings, yet existing systems often provide generic support that fails to address individual misconceptions or align with curricular standards. This study addresses the gap by developing and evaluating a Smart Adaptive Feedback System (SAFS) designed to foster science concept mastery through real-time, data-driven feedback. The aim is to determine whether SAFS enhances conceptual understanding, reduces persistent misconceptions, and improves instructional efficiency compared with traditional formative feedback. Specific objectives are (1) to design a computational feedback engine that identifies student misconceptions in physics and biology concepts from interactive assessments; (2) to implement adaptive feedback policies grounded in cognitive-load theory and constructivist principles; (3) to evaluate SAFS performance across secondary school science classrooms in terms of learning gains, misconception resolution, and time-to-mastery; (4) to assess teacher perceptions of SAFS usability, integration with existing learning management systems, and impact on instructional planning; and (5) to refine the theoretical model of adaptive feedback via iterative cycles informed by empirical results. Methodologically, the study adopts a mixed-methods design comprising a quasi-experimental study and an embedded explanatory sequential component. The population comprises 40 science teachers and approximately 1,200 students from 20 secondary schools within a metropolitan district. A stratified random sampling approach selects 12 schools for the intervention and 8 as controls, ensuring representation across urban, suburban, and rural contexts. In the intervention condition, SAFS will be deployed over a full academic term with 6 distinct modules on core science topics (energy, motion, ecosystems, cell biology, chemical reactions, and genetics). Data collection instruments include (a) standardized concept inventories administered at pre-, mid-, and post-test points; (b) system-generated interaction logs capturing response accuracy, time-on-task, hint usage, and feedback iterations; (c) validated attitudes and motivation scales; (d) classroom observation protocols and teacher interviews; and (e) an end-of-term usability questionnaire. Reliability and validity are established through Cronbach’s alpha for inventories (>0.80), test-retest reliability checks, and content validity review by subject-matter experts. Analytical methods comprise (i) multilevel linear modeling (MLM) to assess effects of SAFS on post-test conceptual scores, controlling for baseline achievement and school-level factors; (ii) logistic regression for mastery of identified misconceptions, based on misclassification analyses from item-level responses; (iii) time-to-mastery analyses using survival analysis to determine how quickly learners reach proficiency thresholds; (iv) thematic analysis of interview and observation data to triangulate quantitative findings and illuminate contextual factors influencing adoption; (v) a process evaluation employing regression discontinuity-inspired checks around measured prompts to examine the impact of feedback timing and granularity. Theoretical frameworks informing the study include Vygotsky’s zone of proximal development, cognitive-load theory, and a constructivist view of feedback, with SAFS embedding models of adaptive messaging, such as mastery-oriented prompts and error-specific hints, aligned to Bloom’s taxonomy of cognitive processes. The expected technical contributions include a validated feedback policy library, a modular architecture for integration with common learning management systems, and a data-informed model linking feedback characteristics to misconceptions and learning trajectories. Key expected findings anticipate statistically significant improvements in post-test conceptual scores (effect size d > 0.40) for the SAFS group, higher mastery rates of targeted misconceptions, and shorter time-to-mastery compared with controls. Qualitative data are expected to reveal enhanced student engagement, perceived personalization benefits, and constructive teacher experiences with the system, yet potential challenges related to system interpretability and alignment with diverse curricular standards. The study contributes to knowledge by advancing practical, scalable ICT-driven approaches for adaptive feedback in science education, delineating how real-time diagnostic insights can inform both student learning paths and instructional planning. It also offers a theoretically grounded model of adaptive feedback that synthesizes cognitive-load and constructivist perspectives with empirical validation. Recommendations include extending SAFS to additional science domains, refining the feedback ontology for cultural and linguistic diversity, and exploring long-term retention effects through multi-term longitudinal studies. The conclusion emphasizes that carefully designed adaptive feedback systems can meaningfully enhance concept mastery while supporting teachers in delivering timely, individualized guidance.