Knowledge tracing (KT) predicts learners’ evolving knowledge states by tracking their performance over time. While temporal dynamics have been the focus of most KT studies, spatial structures among knowledge components (KCs) remain underexplored, despite containing rich latent information. Prior work suggested that inter-KCs relationships can enhance KT performance, yet the impact of hierarchical spatial structures remains unclear. This study investigates how multilevel spatial relationships among KCs affect the performance of KT models. Using causal structure learning, we infer causal links among KCs and incorporate the resulting spatial structures into both deep learning and traditional machine learning KT models. Experimental results showed that incorporating second-order spatial structures yielded consistent performance gains. These findings underscore the value of spatial structural information in KT. Furthermore, interpretable feature analyses illustrated how spatial features shape diagnostic predictions, providing insight into factors underlying students’ learning challenges. This spatial perspective not only improves KT models’ performance but also has the potential to inform more targeted and effective instructional strategies.
Amid the rapid development of artificial intelligence (AI), this paper underscores the historical inevitability of exploring long-term development trends in education and emphasizes AI’s growing influence in this field. It is anticipated that AI will become the leading force in education in the future, and for this reason, human teachers may be required to seek a path of symbiosis with AI. Enlightened by Opinions on deepening the implementation of the “AI Plus” initiative issued by the State Council of the People’s Republic of China (2025), this paper explores the inevitability of human society’s continued development of AI and the synchronization between China’s digital education transformation and building China into a leading country in education. Mainly, the paper analyzes AI’s impact on the underlying logic of education: Driven by intellectual curiosity, students can complete the learning process from building knowledge from scratch to profound understanding through continuous interactions with AI, which in turn, leads to AI gradually replacing education’s knowledge imparting function. Consequently, this change may weaken the primacy of competency development in the long run, leading to substantive alterations in multiple dimensions of the traditional educational model. This paper clarifies the analytical logical framework covering four key dimensions—from the micro to the macro levels—proposed in White paper on China’s smart education (Ministry of Education of the People’s Republic of China, 2025), namely future teachers, future classrooms, future schools, and future learning centers, laying a theoretical foundation and ultimately providing an overarching framework to guide the subsequent research of the relevant field.
Large language models (LLMs) are increasingly used to grade written responses, yet large-scale benchmarks against human expert evaluation remain scarce, especially across languages with differing resource levels. This study evaluates ChatGPT-4o using a reranked retrieval-augmented generation framework to grade Finland’s national high-stakes matriculation examination based on 1,016 students’ open-ended responses. We examined GPT-4o’s agreement with official grades, its recognition of grading-relevant keywords, and the effect of translated responses from a low-resource language (Finnish) into a high-resource language (HRL) (English). Using descriptive statistics and correlation analyses, the results show that GPT-4o’s grades on a 0–15 scale closely matched human expert evaluations; 75.00
The integration of large language models (LLMs) into higher education is rapidly transforming learning and research processes. This study investigates the adoption of one such tool, DeepSeek, a Chinese LLM, in Bangladesh’s higher education sector. It introduces geopolitical concern as a novel construct within the extended unified theory of acceptance and use of technology (UTAUT) framework to assess how national security, strategic alignment, and technological dependency influence trust, privacy concerns, and the behavioural intention to adopt AI tools in an academic context. A quantitative survey of 202 Bangladeshi university students was conducted using the extended UTAUT model. Partial least squares structural equation modelling was employed to analyse the impact of constructs such as performance expectancy, effort expectancy, and geopolitical concerns on behavioural intention and use behaviour. Geopolitical concerns significantly shaped trust and privacy perceptions, indirectly reducing adoption intentions. Despite privacy risks, students viewed DeepSeek in a favourable light due to its free access, cultural alignment, and potential to reduce Western tech dependency. Performance expectancy and facilitating conditions were key drivers of adoption, while social influence and hedonic motivation had negligible effects. This study contributes to the integration of geopolitical factors into AI adoption frameworks, highlighting Bangladesh’s unique position amid a growing diversity of globally available AI models. It emphasises the need for culturally congruent, sovereignty-sensitive AI tools in the Global South. Policymakers and AI developers should address geopolitical sentiments and infrastructure gaps (e.g., internet reliability) to foster adoption. DeepSeek’s frugal innovation model provides a blueprint for emerging economies seeking affordable AI solutions.
Personalized learning resource recommendation aims to provide learners with appropriate learning resources to alleviate information overload caused by the explosive growth of data on online learning platforms. Current research predominantly utilizes student interaction data to enhance the quality of recommendations. However, this approach neglects the intricate dependency networks among learning resources, which directly influence the effectiveness of knowledge acquisition pathways, and lacks the capacity to model individual learning abilities and objectives. To address these limitations, this study introduces a unified learning resource recommendation method (ULRRM), which integrates multidimensional graph information and employs conceptual graphs as an intermediary framework to unify resource representations across varying levels of granularity. Specifically, a resource dependency graph is established to guide resource-dependent learning through conceptual dependency relationships, thereby encoding the topological constraints of resources. Then, a local–global dual view is constructed using session history to capture both short-term behavioral patterns and the evolution of long-term interests, thereby enabling the recommendation of learning resource sequences that incorporate multidimensional graph information. Extensive experiments conducted on real datasets demonstrate that the proposed ULRRM method surpasses baseline approaches across several widely recognized evaluation metrics.
