探討科學認識信念及探究能力對於學生於生物探究數位學習環境中進行探究學習之學習歷程與學習表現的影響

This project aims to investigate the effects of science learning self-efficacy and inquiry abilities on students' learning processes and learning performances in a digital biological inquiry learning environment. In collaboration with expert teachers, a digital learning environment was developed using Unity, featuring two modules: "factors affecting the fermentation rate of yeast" and "factors affecting catalase activity". The inquiry process is defined by four core stages: "Identifying Problems," "Planning and Research," "Argumentation and Modeling," and "Expression and Sharing". Study 1 conducted an empirical analysis of students from gifted and regular classes, revealing that gifted students' scoring rates were significantly higher than those of regular students in most stages. However, no significant difference was found in the post-test performance of the "Planning and Research" stage, reflecting that the digital platform interface can provide suitable guidance, thereby bridging the planning ability gap between students of different aptitudes. Regarding science learning self-efficacy, correlation analysis found that "conceptual understanding" was significantly positively correlated with initial problem identification in the regular class and with later argumentation and modeling in the gifted class. Conversely, regular students' stage 2 post-test performance showed a significant negative correlation with conceptual understanding self-efficacy, suggesting that high self-confidence may not directly translate to accurate execution of digital experimental designs, indicating a potential inconsistency between self-assessment and actual performance. Study 2 focused on behavioral differences between science and social science track students, finding that while overall performance was consistent when controlling for pre-test scores, social science track students invested a significantly higher proportion of operation time in the first stage. Furthermore, Poisson regression analysis indicated that the risk of social science track students committing "experimental design errors" was 2.492 times that of science track students, reflecting the influence of academic background on familiarity with variable control methods. Hierarchical multiple regression analysis further confirmed that "restart counts" in stage 2 significantly negatively predicted performance—possibly reflecting disorganized planning strategies—while "operation time" in stage 4 positively predicted outcomes. Study 3 integrated eye-tracking technology to clarify the cognitive processes of inquiry; performance analysis showed that "Argumentation and Modeling" was the most challenging stage, with post-test scores significantly lower than pre-test scores. Regression models revealed that the Revisited Fixation Duration (RFD) on the "experimental introduction" zone in stage 1 was significantly negatively correlated with post-test scores, indicating that repetitive focus on basic background information reflects bottlenecks in information integration. Similarly, the Total Fixation Duration in Zone (TFDZ) and RFD on the "expected results" zone in stage 3 also negatively predicted performance, highlighting cognitive difficulties in students' inquiry learning. In summary, this research integrates multi-dimensional empirical data and clarifies eye-tracking indicators of the inquiry cognitive process, providing a predictive theoretical foundation and practical instructional strategy recommendations for differentiated inquiry teaching in digital environments.