[{"data":1,"prerenderedAt":126},["ShallowReactive",2],{"issue-detail-1-30":3},["Reactive",4],{"volume_id":5,"issue_number":6,"title":7,"cover_image":8,"publish_date":9,"id":10,"volume_number":6,"is_current":11,"published":11,"created_at":12,"updated_at":13,"articles_count":14,"articles":15},6,1,"","\u002Fimages\u002Fuploads\u002Fissues\u002F30\u002F1f286088a0b24004a9e30ce512a717fe_793322e057af85236e9ebe1e6871a262.png","2026-06-09",30,true,"2026-05-21T02:56:41.143229Z","2026-06-09T06:44:48.863602Z",0,[16,26,39,48,65,74,82,93,106,114],{"id":17,"title":18,"authors":19,"abstract":21,"keywords":-1,"type":22,"page_start":6,"page_end":23,"order_in_issue":14,"published_date":9,"view_count":24,"download_count":25},102,"The Future Education: K21CASE = iSTREAMS + iCDIOs",[20],"Fei-Yue Wang","Recently, with the rapid and powerful development of artificial intelligence (AI) technology, a major paradigm shift of educational models —parallel education reform supported by intelligent technology has begun to spread across the world, which commands a comprehensive overview and reflection of current education and teaching from the perspective of parallel intelligence and parallel education. For this purpose, this paper firstly discussed the origin and goal of education and proposed a new prospect to consider and analyze the relationship between parallel intelligence and parallel education. Then, the definition, basic framework and application process of parallel intelligence and parallel education was introduced, which highlighted pointed the fact and significance that parallel intelligence provided effective intelligent technology support for parallel education. After that, we focused on the parallel education reform enabled by intelligent education, and held that this reform not only included the reforms of teaching content and connotation, subject structure, and assessment method, but also involved the innovation of education form. Finally, looking forward to the future, it was emphasized that the concepts of parallel teachers and students, parallel schools and related technologies should be deployed to ensure the success of education reform, so as to create a new direction of education improvement and seek sustainable human welfare.","regular",9,84,11,{"id":27,"title":28,"authors":29,"abstract":34,"keywords":-1,"type":22,"page_start":35,"page_end":36,"order_in_issue":14,"published_date":9,"view_count":37,"download_count":38},92,"K21 as a Finite Educational Backbone in the Age of AI: Toward an AI-Integrated Lifelong Learning Architecture",[30,31,32,33],"Sixiu Liu","Yiteng Xu","Mingxuan Li","Hui Zhang","As artificial intelligence increasingly reshapes how people learn, research, create, and work, the central challenge of educational reform is no longer merely how to optimize separate phases of schooling. It is how to reconnect formal education with the much longer arc of post-school learning, capability renewal, and human–AI collaboration. This article argues that K21 should not be treated as synonymous with lifelong learning. Rather, K21 is better understood as a finite educational backbone that extends educational continuity from early schooling through graduate education and research training, while remaining bounded by institutional timeframes, credential structures, and school-based governance. On that basis, the paper makes three contributions. First, it clarifies the conceptual distinction between K21 and lifelong learning by differentiating their time horizon, institutional scope, organizing logic, and developmental function. Second, it proposes an AI-integrated lifelong learning architecture oriented to the long-term development of what this paper calls the AI-native generation, namely learners whose growth unfolds in environments where AI is routinely embedded in study, inquiry, and work. Third, it examines how K21 and post-K21 lifelong learning can be connected through foundational capabilities, longitudinal learning records, AI-supported pathway planning, and governance mechanisms. The aim is not to replace K21, but to reposition it: K21 remains important as the formal educational backbone, yet it cannot by itself bear the full burden of lifelong development in the age of AI.",10,15,169,14,{"id":40,"title":41,"authors":42,"abstract":44,"keywords":-1,"type":22,"page_start":45,"page_end":46,"order_in_issue":14,"published_date":9,"view_count":47,"download_count":25},93,"Teaching through Making: AI, 3D Printing, and Social Manufacturing for Education",[43],"Jingwei Ge","AI is rapidly reshaping educational practice, both on the consumption side (large-language-model tutors, AI-augmented classrooms) and on the production side (text-to-CAD generators, parameter recommenders, vision-based print monitors, automated documentation). Maker education, the strand of education that asks learners to build artifacts that physically work, is most directly affected when AI absorbs the operational layer of making. The key difficulty this raises is organizational: the educationally substantive layer of making is a residue of judgment at points of tension (with demand, with the physical world, with materials and process, with time and resources), and these tension sources lie outside the learner’s own control, so no single classroom can supply them at the rate the judgment layer requires. The difficulty decomposes into four organizational shortfalls (failure visibility, demand diversity, evaluation multiplicity, iteration continuity), each of which a single makerspace under-delivers. We propose Teaching through Making: a framework that rewrites the social-manufacturing objective function for an educational context, derives the four organizational conditions, and maps each to a combination of ACP-based mechanisms (artificial systems, computational experiments, parallel execution) over a five-layer resource architecture (infrastructure, platform, content, intelligence, governance). A crossed-factor computational experiment establishes that the toolkit raises the four indicators above single-classroom baseline in a controlled setting, with positive evidence on the depth indicators (evaluation tension, chain depth) and qualified evidence on the breadth indicators (failure exposure, demand diversity); empirical validation in real classrooms is identified as the natural next step.",16,25,178,{"id":49,"title":50,"authors":51,"abstract":60,"keywords":-1,"type":22,"page_start":61,"page_end":62,"order_in_issue":14,"published_date":9,"view_count":63,"download_count":64},95,"Teaching Reinforcement Learning in Large-Scale Interdisciplinary Classrooms: A Course Design and Practice",[52,53,43,54,55,56,57,58,59,20],"Lili Fan","Kaige Chen","Xuan Li","Yuchen Li","Lingxi Li","Xiaoxiang Na","Hui Yu","Levente Kovacs","Reinforcement learning (RL) has become a foundational paradigm in modern artificial intelligence, yet its technical complexity poses significant challenges for large-scale interdisciplinary education. Students from diverse academic backgrounds often face difficulties in understanding abstract concepts, maintaining engagement, and applying RL methods in their own domains. This paper presents the design and implementation of a reinforcement learning course tailored for large-scale interdisciplinary classrooms. The course integrates structured content progression, real-world analogies, and interactive learning activities to bridge disciplinary gaps and improve learning accessibility. Key design elements include cognitive scaffolding through everyday scenarios, level-based task progression from fundamental concepts to advanced algorithms, and an interactive feedback system based on QR-code quizzes that promotes continuous participation. The course was implemented in a large cohort of students from multiple disciplines, and preliminary results indicate improved engagement, participation consistency, and conceptual understanding. The findings suggest that carefully designed instructional strategies can effectively support reinforcement learning education in large, diverse classroom settings. This work provides practical insights and a scalable approach for teaching complex AI topics in interdisciplinary contexts.",26,32,164,27,{"id":66,"title":67,"authors":68,"abstract":70,"keywords":-1,"type":22,"page_start":71,"page_end":72,"order_in_issue":14,"published_date":9,"view_count":73,"download_count":36},88,"Building CASE by Generative AI Making Cases Out of Control AutomationScience and Engineering",[69],"Yonglin Tian","Control and Automation Science and Engineering (CASE) is often taught through a small set of familiar examples: inverted pendulums, cruise controllers, tank-level systems, and similar benchmark problems. These examples are useful, but they no longer reflect the diversity and pace of modern automation practice. Developing new cases is difficult because it requires control-theoretic expertise, domain knowledge, access to realistic operating data, and repeated pedagogical calibration. This paper introduces CASE-Gen, a generative-AI-assisted framework for making teaching cases out of control and automation source materials. The framework follows three steps: scenario extraction from technical sources, schema-guided case structuring, and multi-dimensional quality assurance. Unlike direct prompting, CASE-Gen constrains generation through a CASE-specific schema and verifies outputs using retrieval grounding, Bloom’s taxonomy checks, difficulty calibration, and symbolic consistency tests. We describe the design rationale, implementation workflow, an illustrative case package, and an evaluation plan, then show how the framework can support accreditation-oriented curriculum development, professional upskilling, and cross-institutional case sharing. The paper is a framework proposal rather than a completed empirical evaluation. The central argument is modest but practical: generative AI should not replace expert case authors, but it can turn the most time-consuming parts of case development into a supervised, reviewable workflow.",33,40,202,{"id":75,"title":76,"authors":77,"abstract":78,"keywords":-1,"type":22,"page_start":79,"page_end":80,"order_in_issue":14,"published_date":9,"view_count":81,"download_count":38},89,"iCDIOs and the Shift Toward Human AI Co Creation in Engineering Education",[55,59],"Against the backdrop of the deep integration of generative artificial intelligence into engineering activities, the traditional CDIO-based engineering education model is increasingly facing pressure for structural transformation. This study systematically analyzes the limitations of CDIO from the perspectives of agency structure, competency focus, collaborative mechanisms, and governance requirements. Building upon this analysis, the paper further develops the iCDIOs framework from the perspective of Human-AI Co-Creation, emphasizing the educational significance of the “i” dimension in innovation and intelligent literacy, as well as the “s” dimension in service orientation, safety governance, sustainability, and responsibility. Furthermore, the study discusses how Human-AI Co-Creation functions as a core operational mechanism that reshapes engineering competency models, curriculum organization, classroom interaction, and evaluation processes. Representative teaching practices involving courses on Large Language Models (LLMs), AI Agents, and Automated AI are further introduced to illustrate how AI can participate throughout lecture preparation, project development, tutoring, and assessment under interdisciplinary educational settings. The study argues that engineering education is undergoing a transition from a traditionally human-centered paradigm toward a Human-AI Co-Creation paradigm, with iCDIOs providing a systematic framework for engineering education transformation in the AI era.",41,49,187,{"id":83,"title":84,"authors":85,"abstract":88,"keywords":-1,"type":22,"page_start":89,"page_end":90,"order_in_issue":14,"published_date":9,"view_count":91,"download_count":92},91,"Smart Navigation Developing AI Agents for K12- Students",[86,87],"Siji Ma","Chenxing Fan","This paper proposes Smart Navigation, an ACP-based multi-agent framework for personalized and interest-aware learning navigation in K-12 education. The core challenge addressed in this paper is not the absence of isolated technical tools, but the lack of a structured framework that can simultaneously organize interdisciplinary learning content, model learners’ knowledge states and evolving interests, support adaptive strategy generation, and maintain clear safety boundaries for young learners. Building upon the iSTREAMS educational framework—where i denotes Inspirations, Innovations, Interdisciplinary, Intelligence, and International; STREAM denotes Sciences, Technology, Robotics\u002FResearch, Engineering\u002FEconomy\u002FEcology, Arts\u002FAI, and Math\u002FManagement\u002FManufacturing; and s refers to Safety, Security, Sustainability, Sensitivity, Service, and Smartness—this paper treats iSTREAMS as the content and value backbone of Smart Navigation. The ACP methodology provides its operational logic through artificial learning systems, computational experiments, and parallel execution. On this basis, three software-defined knowledge robots—the Descriptive Intelligence Agent, Predictive Intelligence Agent, and Prescriptive Intelligence Agent—are introduced to support knowledge representation, learner-state estimation, and learning-strategy generation, respectively. The framework also defines Parallel Teachers as a teacher–agent collaborative structure rather than a replacement of human teachers. Three illustrative scenarios covering early primary, school-transition, and high-school project learning demonstrate how Smart Navigation can support differentiated K-12 learning needs. The contribution of this work lies in providing a theoretically integrated, mechanistically interpretable, and educationally extensible framework for AI-agent-based learning navigation systems, while clarifying limitations and future directions for classroom deployment, multimodal modeling, and data governance.",50,59,171,17,{"id":94,"title":95,"authors":96,"abstract":101,"keywords":-1,"type":22,"page_start":102,"page_end":103,"order_in_issue":14,"published_date":9,"view_count":104,"download_count":105},94,"An Empirical Study on the Guiding Capacity of Parallel Kindergartens Based on ACP Theory",[97,98,99,100],"Jimei Ma","Peixin Lin","Tingting Li","Yan Li","In the context of the comprehensive implementation of the digital education strategy and the deep empowerment of people’s livelihood by artificial intelligence technology, smart education has become a core tool and inevitable path to promote the high-quality development of preschool education. Traditional kindergarten management models have long relied on manual experience, offline circulation, and fragmented management, and generally suffer from a series of practical governance dilemmas such as the separation of physical space and management data, experience-based decision-making, lagging emergency response processes, and passive handling of potential safety and emotional risks. In order to systematically solve the management shortcomings of preschool education scenarios from the perspective of complex system governance, this paper uses ACP parallel intelligence theory as the core supporting framework, combined with the governance logic of social physical information systems (CPSS), to construct a new management paradigm for parallel kindergartens adapted to preschool education scenarios. The research systematically designs a four-layer architecture with descriptive intelligence, predictive intelligence, and guiding intelligence as the core and multi-level structure synergistic support, focusing the research on the model construction and quantitative empirical testing of kindergarten guiding capacity under the parallel system.\n\nThis paper relies on the AgentScope distributed multi-agent development framework and large language model technology, taking the full set of communication data from WeChat groups for home-kindergarten collaboration in real kindergartens in the autumn semester of 2024 as the research sample. Following ethical guidelines, data anonymization and cleaning were completed to construct a high-fidelity artificial social system for kindergartens. Three control schemes were set up: a baseline control experiment, a parallel management intervention experiment, and a quasi-parallel execution experiment with multi-dimensional expert evaluation, to complete multi-scenario, multi-dimensional quantitative comparative analysis. The study fully verifies the adaptability and application value of ACP theory in typical complex open systems such as kindergartens, proving that the parallel governance model can effectively promote the preschool education management system from traditional single-point information tool empowerment to a systematic intelligent governance upgrade driven by data, supported by models, and linked by virtual and real systems. The research results can not only enrich the application system of parallel intelligence theory in the sub-fields of education, but also provide solid theoretical basis, technical implementation path and empirical data support for the top-level design, scenario implementation and standard construction of smart kindergartens.",60,72,198,19,{"id":107,"title":108,"authors":109,"abstract":111,"keywords":-1,"type":22,"page_start":112,"page_end":24,"order_in_issue":14,"published_date":9,"view_count":113,"download_count":38},90,"Smart Journey Deploying AI Agents for K12+ Students",[110],"Sangtian Guan","Education at every level, from K-12 through university, confronts an irreducible structural barrier: one human instructor cannot provide personalized, real-time attention to every student who needs it. When a student ventures beyond familiar material and becomes stuck, she enters a queue for instructor help. As class size grows, the queue lengthens, and the expected cost of exploration eventually exceeds its expected benefit. Rational students abandon active learning for safe, superficial routines. This is not a motivational failure. It is a congestion game. It scales with every student added to the classroom. We formalize the traditional classroom as a Constrained Exploration Game (CEG) and prove that Smart Journey’s AI Agent deployment converts an unstable congestion trap into a stable exploration equilibrium. Three theorems establish: (i) a congestion threshold beyond which exploration collapses in AI-free classrooms; (ii) an attention restoration condition showing that even a modest number of AI Agents restores exploration at any scale; and (iii) a regret bound demonstrating an order-of-magnitude improvement in learning efficiency. Simulations across class sizes from 20 to 500 students confirm a twenty-seven-fold improvement in concept mastery with AI Agents deployed. An ablation experiment decomposes AI’s contribution into structural and cognitive components, finding that 83%–97% of the total learning gain comes from structural unbundling of attention. The central finding is that AI Agents do not need to teach better than humans to transform education at scale. They only need to teach without queuing.",73,197,{"id":115,"title":116,"authors":117,"abstract":122,"keywords":-1,"type":22,"page_start":123,"page_end":49,"order_in_issue":14,"published_date":9,"view_count":124,"download_count":125},101,"Parallel Education with AI Agents: Teaching Innovation with New AI and Creativity Science",[118,119,120,121,20],"Qinghua Ni","Tengchao Zhang","Fei Lin","Jun Huang","Higher education still largely relies on the traditional model of classroom lecturing, student assignments, and examination-based evaluation, which faces limitations in timely support, personalized learning, collaborative organization, and continuous teaching optimization. To address these limitations, this paper proposes Parallel Education, a framework for intelligent teaching and learning in the AI era. Inspired by parallel intelligence and the ACP approach, the framework constructs an artificial teaching society that operates in parallel with the real teaching system. It extends professors, teaching assistants, and students into digital professors, digital teaching assistants, and digital students, enabling educational roles, learning states, and collaborative relations to be represented and updated in digital environments. A five-layer system architecture is designed, including the physical teaching system, data collection, artificial teaching society, parallel intelligence, and application and governance layers. The paper further discusses representative scenarios, including online learning, parallel classroom teaching, offline practice, collaborative learning, and creative learning. The proposed framework aims to move intelligent education from fragmented AI tools toward a virtual-real integrated and human-centered educational paradigm.",85,83,8,1783694007797]