Against the backdrop of digital transformation in Chinese higher education and the rapid diffusion of Knowledge Graph-supported teaching environments, this study addresses a question that is often noted in practice but insufficiently explained in rese...
Against the backdrop of digital transformation in Chinese higher education and the rapid diffusion of Knowledge Graph-supported teaching environments, this study addresses a question that is often noted in practice but insufficiently explained in research: why do students exposed to the same Knowledge Graph support sometimes experience clearer learning structures and more controllable progress, while others feel an increased burden and struggle to use the system effectively? Adopting a learner-centered perspective, the study conducts a systematic quantitative examination of how Knowledge Graph environments shape learning outcomes and under what conditions such effects are more likely to become salient.
Conceptually, the study treats learners’ perceived ‘structural visibility’ in a Knowledge Graph learning environment as the key independent construct. Structural visibility refers to learners’ capability to use a Knowledge Graph to apprehend the overall course structure, specific learning pathways, and their connections to assessment tasks. This construct is operationalized by two dimensions: the alignment of teaching, learning, and assessment, and the clarity of learning pathways. Integrating cognitive load theory, social cognitive theory, engagement research, and technology acceptance models, the study proposes a chain mechanism of ‘structural visibility→self-efficacy→learning engagement→learning outcomes,’ while hypothesizing that technology acceptance moderates the path from structural visibility to self-efficacy. These assumptions form the structural equation model to be tested.
Methodologically, the study employs a questionnaire-based quantitative design. The participants were undergraduates from eastern and central China who had experience using Knowledge Graph supported teaching environments. A total of 786 valid responses were collected via an online survey platform. The measurement scales were finalized after expert content validity review and pilot testing. SPSS was used for data screening and descriptive analysis; Mplus was employed to examine reliability, convergent validity, and discriminant validity of the measurement model. Structural paths were then estimated within a partial least squares structural equation modeling (PLS-SEM) framework. A baseline model and several competing models were constructed to test direct effects, single-mediator effects, chain mediation, and the proposed moderation.
The results indicate that structural visibility exerts a stable and statistically significant direct effect on learning outcomes with a relatively large effect size. The single mediating effect through learning engagement also makes a substantive contribution, suggesting that structural visibility not only enhances learning outcomes directly but also operates indirectly by activating engagement. In contrast, the chain mediation via ‘self-efficacy→learning engagement’ is statistically significant yet smaller in magnitude. Path estimates further show that structural visibility most strongly predicts learning engagement; the effect of self-efficacy on engagement is comparatively modest, whereas engagement remains a robust predictor of learning outcomes. Technology acceptance demonstrates a clear moderating effect on the ‘structural visibility→self-efficacy’ path: when acceptance is high, structural visibility more readily translates into enhanced self-efficacy; when learners perceive lower ease of use and usefulness of the Knowledge Graph system, this translation is noticeably weakened.
The study advances the literature in three ways. First, by linking structural visibility to the integration of teaching, learning, and assessment, it offers a more fine-grained conceptualization of how structured information is externalized in Knowledge Graph environments, enriching discussions on structural explicitness and cognitive load regulation. Second, through comparisons between full and partial mediation models, it revises the implicit assumption that self-efficacy must occupy the central position in the mediation chain, arguing instead that in highly structured digital learning contexts, engagement is more likely to function as the pivotal conduit connecting learning environments to outcomes, with self-efficacy playing a supportive and conditional role. Third, it extends technology acceptance models beyond explaining adoption and frequency of use to illuminate boundary conditions for instructional effectiveness, showing how differences in acceptance may silently amplify or attenuate the outcome gains associated with structural visibility.
Practically, the findings suggest that the design of Knowledge Graph teaching environments should prioritize structural visibility so that students can clearly discern what to learn, how to proceed along appropriate pathways, and which assessment tasks align with those pathways. During implementation, structured task design, timely feedback, and peer modeling should be jointly employed to convert structural information into sustained behavioral engagement and deeper cognitive processing. In terms of technological support, universities should systematically assess students’ levels of technology acceptance. For groups with relatively low acceptance, institutions ought to provide measures such as simplified interfaces, guided operational examples, and peer mentoring, thereby reducing technological barriers.
Keywords: Knowledge Graph; structural visibility; self-efficacy; learning engagement; learning outcomes; technology acceptance