Convergent Intelligence Classroom: An Innovative Classroom Form with AI LLMs

The new round of technological revolution and industrial transformation is driving the leap of social productivity, profoundly affecting the structure of talent demand and capability requirements, prompting us to re-examine how to cultivate new quality talents that match the new quality productivity. The cultivation of new quality talents requires new quality education, and the classroom is the main battlefield for cultivating new quality talents, as well as an important way to realize new quality education. Educational reform must touch the classroom level to truly open up the “last mile” and achieve a comprehensive innovation of educational concepts. AI LLMs, by providing interactive learning environments, customized learning materials, personalized learning experiences, and automated assignment assessments, are expected to become a lever to promote profound changes in classroom teaching and trigger a revolution in learning paradigms. However, it is worth noting that while large models empower classroom transformation, they also harbor risks such as misleading answers, integrity crises, and excessive dependence. Therefore, some schools, such as New York City public schools, have explicitly prohibited the use of AI LLMs in classrooms. However, as AI LLMs increasingly integrate into various aspects of education, the education sector is gradually realizing that rejecting them is not a wise choice. Today’s students will find themselves in an era where they dance with AI LLMs; they must possess the ability to work collaboratively with AI LLMs to better adapt to the development of the digital society. Therefore, how to guide teachers and students to use AI LLMs responsibly in the classroom has become an important issue that needs urgent attention. This study aims to explore specific strategies for deeply integrating AI LLMs into classroom teaching and to investigate the innovative learning paradigm of the convergent intelligence classroom that incorporates AI LLMs, clarifying its conceptual connotations and key features, discussing its key elements and application modes, with the hope that AI LLMs can become a trusted “thought partner” for teachers and students, effectively promoting the digital transformation of classroom teaching.

The classroom is the core arena of education and teaching, and its form evolves continuously with the changes in social demand and educational philosophy. The “convergent intelligence classroom” is a product of the development of the digital society and the choice of educational values. Accurately grasping its key connotations and features against the background of the digital intelligence era is a necessary path for integrating AI LLMs into classroom teaching.

(1) Conceptual Origin: The Classroom Practice Field of Convergence Education

Technological progress is accelerating changes in the structure of social capability demand. Increasingly sophisticated automation technologies are continuously taking over conventional, repetitive, and even highly knowledge-intensive tasks previously dominated by humans. This profound change poses severe challenges to the low-level educational concept that emphasizes memorization and simple application in traditional education, forcing us to re-examine the talent cultivation model. In this context, learning individuals must possess a series of new quality competencies: AI infiltration skills and technology ethics norms under technical thinking, high-awareness learning quality and cross-cultural capability supported by complex thinking, and boundary-breaking abilities and creative practical wisdom led by change-making thinking. This transformation urgently calls for the birth of a new quality education paradigm to respond to the new requirements for talent capability structure in the digital intelligence era.

Convergence education, as a frontier path for exploring new quality education, focuses on constructing rich technological learning environments and organizing cross-border collaborative innovative activities, leading individuals to face the complex challenges of the real world, breaking down disciplinary barriers, transcending knowledge boundaries, building interdisciplinary knowledge systems and integrated disciplinary mental structures, and inspiring individuals to create forward-looking solutions. The convergent intelligence classroom is a product of the application of convergence education concepts in the classroom. It cleverly integrates AI LLMs, which can not only utilize the rich data sources, progressive strategy chains, and diverse toolsets provided by large models to promote students’ construction of new knowledge and effectively address complex problems but can also decompose complex tasks into sub-tasks undertaken by different roles through multi-agent interactions based on large models, supporting real-time interaction and collaboration between students and intelligent agents to jointly solve problems. In this process, students continuously establish technical thinking through application experiences, cultivate complex thinking through interactive collaboration, and shape change-making thinking through solving poorly structured problems, continuously achieving cognitive upgrades and leaps in innovative thinking. Each collision of ideas and blending of wisdom in the convergent intelligence classroom is a manifestation of the emergence of collective intelligence, facilitated by the integration of human and artificial intelligence. This practical scenario not only provides fertile ground for the development of students’ new quality competencies but also lays a solid foundation for cultivating new quality talents that adapt to future societal challenges, effectively promoting the realization of convergence education goals.

(2) In-depth Explanation: A New Learning Form Integrated with Large Models

The deep integration of technology and education paints a picture of classrooms under technological innovation: from information-based classrooms based on electronic backpacks, to intelligent classrooms based on cloud network terminals, to smart classrooms based on big data analysis of learning conditions, the digital development of classrooms and the process of technological transformation complement each other. Currently, the emergence of AI LLMs is bringing revolutionary impacts on classroom teaching, not only extending the boundaries of traditional in-person interactions but profoundly changing the binary pattern of teacher-student dialogue. Moreover, by shaping diverse dialogue roles, it enables intelligent complementarity and collective intelligence collaboration among teachers, students, and AI LLMs. Against this backdrop, the “convergent intelligence classroom” has emerged, aptly interpreting the beautiful vision of harmonious coexistence between human intelligence and artificial intelligence in the classroom. “Convergence” means harmoniously aggregating different elements to create a brand new classroom ecology; “intelligence” emphasizes the collective wisdom that emerges from the collaboration and coexistence of human intelligence and artificial intelligence.

The convergent intelligence classroom, as a new form of classroom integrating AI LLMs, cleverly combines multiple interactions among teachers, students, and AI LLMs, driven by problems and facilitated through dialogue, continuously co-creating new knowledge, cultivating cross-border thinking, and agilely generating new solutions, realizing the emergence of collective intelligence and promoting the generation of wisdom in each individual. In the convergent intelligence classroom, the boundaries of dialogue are completely broken, and communication between teachers and students, as well as among students, flows naturally, while cross-border exchanges between teachers and AI LLMs become the norm, opening a new era of dialogic learning. Teachers can be a specific subject teacher or multiple teachers from different disciplines, who can not only guide students to interact with a single-role AI LLM but also cleverly organize deep dialogues between groups or classes and the large model, allowing the AI LLM to play multiple roles, collaborating with students to cross knowledge boundaries and explore unknown fields, injecting more flexibility and openness into students’ knowledge systems.

Unlike the “one device per person” model advocated by smart classrooms, which collects data in real-time to achieve efficient interaction and intelligent feedback, the convergent intelligence classroom is more flexible in its terminal configuration, which can be one device per person, one device per group, or even just a teacher’s device for projection. Regardless of the configuration, the focus is on leveraging AI LLMs to directly touch the essence of learning, where the classroom does not merely rely on data collection and analysis but pays more attention to the deep construction of students’ cognitive structures and the innovative development of thinking, helping students achieve deep understanding and interdisciplinary comprehensive application. Overall, the convergent intelligence classroom represents not only a profound transformation of the traditional classroom but also a forward-looking exploration of future educational forms, indicating that under the empowerment of AI LLMs, new educational forms will pave a path full of infinite potential for every student in a more digital, collaborative, and personalized way.

(3) Feature Analysis: A New Ecological Classroom with Multi-Agent Collaboration

Supported by AI LLMs, the convergent intelligence classroom constructs a vibrant new ecological classroom, with its core being multi-agent collaboration, showcasing six distinct features (as shown in Figure 1), driving profound changes in classroom teaching. First, group intelligence collaboration. The convergent intelligence classroom breaks the traditional single interaction model between teachers and students, establishing a “human-human” collaborative framework where teachers, students, and AI LLMs each assume different roles and tasks. Through deep collaboration, it not only promotes deep communication between teachers and students and among students but also establishes a new cooperative relationship between teachers and AI LLMs, as well as among multiple intelligent agents. This collaboration accelerates the flow and sharing of knowledge, making the learning process a dynamic process of gathering collective wisdom and jointly solving problems. Second, agile co-creation. In the convergent intelligence classroom, knowledge is no longer statically imparted but dynamically generated. Students, under the guidance of teachers, quickly design solutions, collect information, and analyze data with the help of AI LLMs. The large model dynamically adjusts information based on students’ questions, feedback, and interactions, while students engage in critical evaluations during dynamic interactions, continuously constructing new knowledge and achieving collective value co-creation. Third, strategic agreement. The convergent intelligence classroom encourages students to jointly formulate and adjust learning strategies with teachers and AI LLMs, with the large model acting as a personalized learning planner providing customized learning suggestions, assisting teachers in assessing students’ learning states, and adjusting teaching strategies. Students then assess the large model’s suggestions based on actual situations, ensuring that learning paths are reasonable and efficient. This process not only makes personalized learning in large-scale teaching possible but also promotes students’ autonomous learning and self-decision-making capabilities. Fourth, knowledge and ability co-construction. In the convergent intelligence classroom, teachers guide students to ask questions to AI LLMs, obtaining the resource libraries and reasoning chains necessary for cognitive construction. Through multi-directional interactions, students continuously build and refine their subject knowledge systems, even exploring boundary knowledge and expanding cognitive boundaries, generating more new boundary knowledge, and co-constructing a more complete knowledge and ability system. Fifth, creative co-generation. The multi-dimensional perspectives provided by the large model offer students a broad creative space, allowing them to freely imagine, boldly hypothesize, and repeatedly test, continuously generating novel viewpoints or ideas, which not only stimulates students’ innovative capabilities and change-making thinking but also accelerates the formation of an innovative culture. Sixth, thinking co-expansion. One of the deep goals of the convergent intelligence classroom is to expand students’ thinking boundaries. AI LLMs, through inspiring deep thinking and stimulating critical thinking, support students in breaking out of traditional thinking frameworks, examining problems from multiple perspectives, and reconstructing knowledge in multiple dimensions, promoting the deep development of students’ thinking.

Learning structure is the process by which learners gradually achieve corresponding goals in an appropriate space by selecting appropriate content and methods. From the perspective of learning structure, considering the core elements and basic structure of the convergent intelligence classroom not only helps to depict the picture of classroom transformation in the digital intelligence era but also assists teachers in systematically examining the core structural elements after integrating AI LLMs into the classroom, allowing for detailed analysis of the interweaving effects of each element and the practical logic of how the convergent intelligence classroom promotes students’ wisdom generation.

(1) Five-Dimensional Perspective: Deconstructing the Convergent Intelligence Classroom from the Learning Structure

From the perspective of learning structure, the key to examining the convergent intelligence classroom lies in exploring how to construct a classroom form that fully leverages students’ subjectivity and promotes genuine learning after the integration of AI LLMs. Different entry perspectives lead to different categorizations of structural elements. By focusing on the five dimensions of “people, objects, events, environment, and context” in learning design, we can establish a systematic understanding of the convergent intelligence classroom and better describe its basic structure.

First, the collaborative subjects integrated with AI LLMs (people). The subjects and their relationships in the convergent intelligence classroom include not only human teachers and students but also the multi-role humanoid subjects played by AI LLMs, forming various collaborative relationships. Among them, students, as cognitive subjects, deeply interact with the large model, leveraging the advantages of human intelligence and machine intelligence to achieve collective intelligence integration, forming a “smart brain” that transcends ontology. Simultaneously, through interactions with teachers and peers, they generate collective intelligence that exceeds individual intelligence, forming a “social brain” in cognitive construction and display exchanges, achieving knowledge complementarity and capability co-construction. The “smart brain” formed by human-machine collaboration and the “social brain” formed by interpersonal interactions together construct a “composite brain,” achieving joint understanding, thinking, and decision-making on problems.

Second, the rich technological intermediaries encompassing AI LLMs (objects). The convergent intelligence classroom creates an open learning space constructed by rich technology, providing the attributes of “residence” (the place where learning occurs) and “transportation” (the medium that facilitates understanding and expression for subjects) for cognitive construction and display exchanges. Digital platforms and mobile terminals provide ubiquitous interactive learning environments, supporting students to flexibly obtain personalized services online and offline, in and out of the classroom, achieving self-paced learning; supporting the generation of rich and diverse learning resources relying on AI LLMs, helping students deepen understanding and construct new knowledge; supporting the recording of students’ learning behaviors and performances, guiding students to select suitable learning paths based on their needs with real-time feedback and suggestions provided by the large model.

Third, the activity scheduling supporting multi-agent interactions (events). This refers to a series of events in which teachers, students, and AI LLMs conduct several activities in a certain time sequence within the convergent intelligence classroom. It includes not only key activities in traditional classrooms (such as content delivery, group discussions, etc.) but also innovatively incorporates deep interaction sessions with AI LLMs, such as question feedback and reflective corrections. The design of activities needs to follow students’ cognitive needs and patterns, carefully arranging activity sequences based on different needs, which can be organized based on the large model to reflect the sequential arrangement of knowledge progression, parallel arrangement of multiple perspectives, optional arrangement reflecting subject intentions, and compensatory arrangement promoting cognitive expansion, thereby serving the achievement of learning objectives.

Fourth, the learning scenarios constructed based on large models (environment). Students in the convergent intelligence classroom are placed in scenarios connected to the real world and directed toward real problems, closely aligning with learning events, achieving consistency between learning content and the real world and work scenarios. Students trigger direct and indirect experiences in real-time interactions, preparing knowledge for subsequent problem-solving. The learning scenarios constructed in the convergent intelligence classroom can be practical projects supporting students’ direct engagement with the real world or fully utilizing AI LLMs to expand broader situational simulations and role-playing, greatly extending the boundaries of learning. Additionally, it is necessary to design problem contexts closely associated with the real world, encouraging students to actively deconstruct complex problems, seek questions from the large model, and quickly reflect and respond, thereby triggering new questions and thoughts, gaining a richer learning experience through the iterative cycles of question chains and response chains.

Fifth, establishing a content context that complies with the digital intelligence era (context). The “context” of learning in the convergent intelligence classroom includes both the content logic designed by teachers according to curriculum standards and competency requirements and the knowledge spectrum formed by students through activity sequences. The organization of content logic involves finding a thread of clues within the learned content, organizing course content with specific logic, and establishing several connections between knowledge points and knowledge units, designing a coherent content system based on cognitive patterns. Its value lies in guiding students to transcend shallow learning, entering the deep learning called for in the era of AI LLMs, forming a structured and deep knowledge structure, and transforming it into a mental structure for solving real problems, thereby addressing infinite real problems with limited content learning.

(2) Structural Relationships: The Operational Model of the Convergent Intelligence Classroom Structure

Considering the structure of the convergent intelligence classroom from five dimensions is due to the fact that in current teaching practices, especially after the intervention of new technologies, many teachers often focus only on certain specific dimensions while neglecting others, leading to varying quality of teaching and even raising concerns about the “backlash” of technology on educational quality. The introduction of a five-dimensional perspective not only reflects respect for teachers’ existing teaching wisdom but also helps them better understand the key dimensions that need attention in classroom teaching after integrating AI LLMs, aiding them in accurately grasping the key elements of classroom teaching reform amidst continuous technological innovation. In practical teaching and learning, it is necessary to clarify the structural relationships of the five elements to create a high-quality convergent intelligence classroom (as shown in Figure 2). Specifically, it can start with the “context” of learning content, closely aligning with curriculum standards and students’ cognitive needs, using the transformation and decomposition of “course objectives – unit content” to design content logic; based on this, leveraging AI LLMs to create learning “environments” that are dependent on content and driven by problems, guiding students to establish a value understanding of the content; through the transformation of “content – activities,” designing the learning “events” that promote the mapping of the content “context” to students’ knowledge “context”; emphasizing deep cooperation among collaborative subjects in learning activities, clarifying the role positioning and collaboration mechanisms of learning “people”; and utilizing rich technological intermediaries to provide supportive conditions for scenario construction and activity development (learning “objects”), stimulating collective intelligence, and truly realizing the “convergence of intelligence” in the classroom.

Deeply integrating AI LLMs into classroom teaching has become an unavoidable important issue in the era of digital intelligence. Given the varying progress of digital infrastructure construction in different schools and the uneven degree of popularization of intelligent terminal devices, this study proposes three classroom structures for the convergent intelligence classroom based on the three-layer configuration of intelligent terminals, aiming to explore adaptable and flexible practical paths.

(1) Three-Layer Configuration: From Teacher Terminals to Student Terminals

When exploring practical paths for the convergent intelligence classroom, we focus on various terminal configuration strategies, aiming to guide teachers to flexibly configure intelligent terminals according to specific learning objectives, student needs, and existing resource conditions. We hope to ensure that every student has equal learning opportunities, regardless of the level of digital development in their school, allowing them to deeply experience the learning changes brought by AI LLMs and share digital benefits, thereby confidently integrating into and leading the future digital society. Based on the five-dimensional perspective, we conduct an in-depth analysis of the convergent intelligence classroom structure under different terminal configurations (see Table 1), using analytical logic to explain it along the logical chain of “objects → people → environment → events → context”.

The basic configuration classroom only configures teacher intelligent terminals, constructing an interactive bridge between teachers and AI LLMs. At this time, the large model acts as an auxiliary tool, providing rich learning resources (reading materials, interactive questions, examples) and instant information support, assisting teachers in efficient teaching. Although the direct interaction of students with the large model is somewhat limited, through centralized and standardized teaching by teachers, students can still indirectly experience the learning scenarios and experiences generated by the large model. Teachers control the pace of learning, ensuring all students delve into the content in accordance with established progress, while providing limited personalized support and assessments. This classroom requires lower technical resources, suitable for situations with limited equipment, helping students understand knowledge and master concepts while cultivating their critical thinking and analysis skills regarding the content generated by the large model.

The advanced configuration classroom provides intelligent terminals for groups, promoting deep collaboration between group members and AI LLMs. Here, the large model can assume dual roles as an intelligent tutor and learning partner, providing task guidance and expert knowledge, while stimulating students’ exploratory desires through multi-dimensional perspectives and exemplary mistakes. In group cooperation, students actively ask questions and collaboratively solve problems, with teachers guiding and coordinating from the side, facilitating efficient team operation. This configuration is especially suitable for scenarios requiring groups to complete complex tasks or projects (such as project-based learning, cooperative inquiry learning, etc.), where each group can undertake different tasks in real or simulated scenarios and engage in dialogue with AI LLMs to gain multi-faceted understanding. Groups have a certain degree of autonomy and flexible pacing, with teachers providing personalized support for groups and jointly evaluating group performance with the large model, ultimately cultivating students’ problem-solving and teamwork abilities.

The high-level configuration classroom provides intelligent terminals for each student, enabling independent dialogue between students and AI LLMs. At this time, the large model not only provides expert knowledge or multi-dimensional perspectives but also acts as a personalized learning planner, offering customized learning paths and instant feedback. Students are highly engaged in interactions, generating progressive question chains, autonomously asking questions to the large model and correcting existing understandings. Teachers, in this process, play more of a guiding and observing role, focusing on students’ personalized development. This classroom form is suitable for personalized tasks that require students to complete based on their interests or for classroom activities that require self-exploration of a specific topic, supporting self-paced learning, with teachers and the large model jointly providing personalized learning support and offering real-time feedback and individual assessment suggestions. This configuration provides students with a highly flexible learning environment, promoting the comprehensive development of autonomous learning abilities and critical thinking.

In summary, the three-layer configuration system provides flexible and diverse solutions for different educational scenarios. When resources are limited, the basic configuration can be chosen to ensure teaching quality; when aiming for team collaboration and problem-solving ability cultivation, the advanced configuration is an ideal choice; and when technical equipment is sufficient and personalized needs of students need to be met, the high-level configuration will provide a better learning experience.

(2) Three Degrees: From Auxiliary Teaching to Promoting Cognition

In the classrooms with profound integration of AI LLMs in the three configurations, the large model plays different roles, assuming different responsibilities and functions (see Table 2). In the basic configuration, AI LLMs become auxiliary teaching assistants, generating and displaying multi-modal teaching content (such as virtual digital humans) based on teachers’ needs, presenting learning content through diverse representations to help students understand knowledge more intuitively. They support text or voice interactions, responding instantly to teacher and student inquiries, and providing precise information support. They can also quickly generate tests and exercises to assess students’ knowledge mastery. Leveraging intelligent platforms, they can deeply analyze students’ learning behaviors, generating diagnostic suggestions in real-time. In the advanced configuration, the role of AI LLMs is further expanded, becoming collaborative guides and wise partners for group learning. They not only generate specific tasks and questions for groups, providing collaborative support and task guidance but also simulate experts or historical figures, offering multi-faceted insights and intellectual inspiration to help students deeply understand the content they are learning. During group discussions, the large model can also simulate a wise companion, contributing constructive viewpoints and feedback, effectively promoting interaction and cooperation among group members. Moreover, it can serve as a summarizer for meetings, organizing the viewpoints constructed by the group and recording the participation and contribution levels of group members. In the high-level configuration, AI LLMs not only act as role players, simulating expert and peer roles but also become each student’s exclusive learning mentor, customizing personalized learning paths and resources based on students’ interests, needs, learning levels, and feedback, dynamically adjusting plans to ensure the achievement of learning objectives.

Based on different configurations of classroom forms, this study explores the application styles of the convergent intelligence classroom incorporating AI LLMs, constructing three typical modes centered on problem-based, inquiry-based, and adaptive learning, aiming to accelerate the transformation of classroom teaching in the digital intelligence era.

(1) Basic Configuration: Problem-Based Learning under the Convergent Intelligence Classroom

This mode is rooted in problem-driven teaching philosophy, taking real-world problems as the starting point and core of learning, guiding students to master core concepts and principles while actively exploring and solving problems, cultivating questioning awareness, critical thinking, and problem-solving abilities. As the most basic and common application style of the convergent intelligence classroom, it mainly encompasses five key links: immersing in simulated situations, defining progressive problems, collecting and analyzing information, collaborative dialogue for group intelligence, and integrating to construct new knowledge (as shown in Figure 3). During the teaching process, teachers flexibly utilize intelligent teaching platforms, leveraging virtual digital humans or the virtual roles of AI LLMs to immerse students in simulated learning scenarios, stimulating their exploratory interests and clarifying learning objectives. Students refine and decompose driving questions, constructing personal or collective question logic frameworks, and jointly determining core question chains with teachers. Next, focusing on each question, they conduct research, not only forming insights and sharing them based on information provided by teachers but also engaging in dialogue with virtual roles through individual questions or group examples, continuously reflecting on and expanding cognitive boundaries, ultimately constructing a knowledge network based on question chains to achieve knowledge internalization and innovation.

For example, in a ninth-grade history lesson on the “First Industrial Revolution,” teachers can plan a cross-temporal “interview” activity, allowing students to act as reporters and explore “why the Industrial Revolution first broke out in Britain and its profound impact on social and economic structures” in front of the AI LLM playing the role of Queen Victoria. By playing a background video of the Industrial Revolution, teachers cleverly set the scenario, posing thought-provoking questions to guide students in gradually breaking down complex issues, sharing their questioning logic and forming a clear question chain (“Why Britain?” – “What was the process?” – “What were the impacts?”). For the first question, students rely on diverse historical materials provided by the teacher to analyze the prerequisites for Britain to carry out the Industrial Revolution (such as capital, economy, technological conditions, etc.). Teachers guide students to think beyond a single perspective, considering whether other countries also possessed these conditions, encouraging students to ask the large model about “the social backgrounds of other countries at that time and the missing elements of the Industrial Revolution,” thereby deepening their understanding of “primacy” comprehensively. When exploring the second question, students are divided into three groups, focusing on the three major themes of “textile industry and the spinning jenny,” “steam engine and factory system,” and “steam cars and railway era,” thoroughly sorting historical materials, constructing knowledge frameworks, and generating derivative questions based on this. During group presentations, they can not only share their insights with teachers and peers but also refine what they know, acquire new knowledge, and broaden the unknown through dialogue with multiple subjects, deeply appreciating the historical significance of the Industrial Revolution. For the third question, students use “Britain hosting the World Exposition” as an entry point to ask the AI LLM questions, where the model playing “Queen Victoria” will inform them of the unique advantages and widespread social effects brought by the Industrial Revolution to Britain. Subsequently, students can continue to ask the large model questions or reflect on the double-edged sword effects of the Industrial Revolution based on the historical materials provided by the teacher, thereby constructing a profound understanding of the social impacts generated by the Industrial Revolution from both positive and negative aspects. Finally, teachers guide students to summarize and review, sharing their deep understanding of the three core questions and collaboratively completing the key knowledge summary of the lesson with the AI LLM. This process not only enables students to master historical knowledge deeply but also cultivates their discipline literacy in empirical historical materials through interaction with AI LLMs.

(2) Advanced Configuration: Inquiry-Based Learning under the Convergent Intelligence Classroom

Inquiry-based learning, as a deeply engaging learning method, encourages students to establish connections between learned knowledge and the real world through active exploration and high-level questioning, thereby stimulating their intrinsic learning motivation. The 5E model, as a typical inquiry-based learning model, leads students through five stages of engagement, exploration, explanation, elaboration, and evaluation, promoting meaningful learning and knowledge transfer applications. However, when students directly participate in inquiry activities in the classroom, they may encounter inefficiencies and obstacles due to insufficient mastery of foundational concepts. The flipped classroom, by reversing the traditional process of knowledge delivery in-class and knowledge internalization out-of-class, provides a potential solution to break through the “ceiling effect” of inquiry-based learning. It allows students to establish preliminary understandings before class, laying the foundation for deepening and transferring knowledge during class. Therefore, we actively explore a new path that integrates the flipped classroom with inquiry-based learning within the convergent intelligence classroom framework, proposing a dual-loop inquiry convergent intelligence classroom model based on the 5E approach (as shown in Figure 4). In this model, the pre-class inquiry loop is task-oriented, relying on intelligent learning platforms and AI LLMs to guide students in self-exploration and forming preliminary understandings; the in-class inquiry loop focuses on higher-order challenges, deepening understanding through group collaboration and AI assistance, achieving knowledge internalization and creative application.

For instance, in the interdisciplinary integrated lesson “The Full Moon and Emotions during Mid-Autumn Festival,” this course integrates the comprehensive learning of ancient poetry in the eighth-grade Chinese textbook and the content of the ninth-grade English unit “Full Moon, Full Feelings,” co-taught by Chinese and English teachers, supplemented by guidance from AI LLMs. The aim is to cultivate students’ ability to narrate Chinese traditional cultural stories in both languages and promote cross-cultural communication. In the pre-class inquiry cycle, teachers first provide learning task sheets and self-study guides, presenting the task scenario: “Invite foreign teachers to participate in the class’s Mid-Autumn Festival dinner,” stimulating students’ willingness to participate. They then guide students to conduct preliminary explorations around various dimensions of the Mid-Autumn Festival, utilizing AI LLMs and online resources to collect materials, understanding English vocabulary, grammar structures, and explanations of traditional cultural knowledge related to the Mid-Autumn Festival, forming an understanding of the traditional culture of Mid-Autumn Festival and its English expression. Finally, through the “Everyone Connects” platform and pre-class quizzes, teachers grasp students’ mastery of foundational knowledge accurately.

In the in-class inquiry cycle, teachers first provide feedback on students’ pre-class quiz results and learning task sheets, explaining the inquiry context and learning objectives through video calls with foreign friends. In the collaborative inquiry phase, based on the foreign teacher’s questions and students’ interests, students are divided into three groups to explore the origins, customs, and poetry of the Mid-Autumn Festival. Each group, under the guidance of teachers, asks questions to the AI LLM to further refine their knowledge exploration. Secondly, for the difficult problems encountered during collaborative inquiry, the dual-teacher collaboration guides answers, instructing students on strategies for questioning the large model (“theme + details + form”) to enhance learning efficiency. Group members utilize learned patterns to inquire about the AI, organizing and integrating collected traditional cultural materials into mind maps, taking turns to present and narrate the Mid-Autumn culture. Finally, students use the English vocabulary and grammar learned beforehand, assisted by the large model, to annotate keywords and phrases in English on the mind map. For challenging points in the translation process, teachers provide precise explanations and strategic guidance, encouraging students to apply learned techniques to autonomously translate famous Mid-Autumn poems. By comparing students’ translations, expert translations, and those generated by the AI LLM, teachers guide students to deeply appreciate the artistry and diversity of English expression, enhancing their language application abilities. Finally, through simulating a real round-table dinner scenario, students rehearse on-site with two English teachers to test their learning outcomes. In this process, AI LLMs become powerful assistants for students’ knowledge acquisition and bridges for cross-cultural communication, enabling students not only to gain knowledge of Chinese traditional culture but also to improve their English language expression abilities and master effective questioning strategies toward AI LLMs, laying a foundation for lifelong learning.

(3) High-Level Configuration: Adaptive Learning under the Convergent Intelligence Classroom

Within the framework of the convergent intelligence classroom, adaptive learning is endowed with new vitality. It follows the concept of “students self-managing, with technology providing adaptive support,” encouraging students to autonomously determine learning themes and plan learning paths based on their interests and needs. The personalized support of AI LLMs flexibly adjusts learning pace and strategies, promoting deep development of knowledge and abilities through self-driven learning. In order to break through the limitations of shallow learning in adaptive learning and achieve deep internalization and flexible application of knowledge, we innovatively integrate the Feynman learning method into the adaptive learning system. The Feynman learning method emphasizes deepening self-understanding through the process of “teaching others,” where students, after mastering knowledge, attempt to explain it to others in concise language. This process not only helps students identify and fill cognitive gaps and blind spots but also prompts them to actively seek personalized compensatory resources for targeted reinforcement and practice. Therefore, we envision a Feynman learning-based adaptive learning convergent intelligence classroom model (as shown in Figure 5), where teachers play the role of guides, deploying Feynman intelligent assistants or dialogue partners to provide students with detailed Feynman learning method guidelines. Students, based on their understanding of the core steps of the Feynman learning method, efficiently construct new knowledge systems and test and deepen their understanding by “explaining” learned content to the large model.

For example, in a seventh-grade biology lesson on “Photosynthesis,” teachers flexibly configure Feynman intelligent assistants or teach students the Feynman strategy for questioning the large model. After mastering the questioning techniques, students can autonomously set learning themes (such as the basic principles of photosynthesis) and obtain personalized content and materials (such as overviews and related resources of photosynthesis) from the large model. After constructing a preliminary understanding, students attempt to “explain” the concept and process of photosynthesis to the AI LLM, which simulates the role of a beginner, asking questions and seeking clarification, prompting students to deepen their understanding through explanation. When faced with challenges and difficulties in explanation, such as how chlorophyll captures sunlight and how plants utilize glucose, students can instantly mark knowledge gaps, and the AI LLM will provide immediate feedback, accurately pointing out misunderstandings and providing targeted supplementary materials (such as videos, articles, virtual experiments, etc.). On this basis, students review their learning based on the feedback from the large model, deepen their understanding, optimize their explanation methods, and attempt to articulate complex concepts in simpler language, using test questions generated by the large model for self-diagnosis to ensure a firmer grasp of knowledge.

It is particularly noted that the above-mentioned cases related to configurations merely illustrate the dynamic variability of the classroom “intelligence” model and do not emphasize the “binding” of models and configurations. In practical classroom teaching, teachers’ teaching wisdom can create more possibilities for the “intelligence” model, while configuration differences may result in varying degrees of “feasibility.”

Technological progress is the driving force behind social change and a key force in promoting educational reform. The collision of AI LLMs and education infuses education with technological colors, injecting new vitality into classroom teaching. The proposal of the convergent intelligence classroom marks a new era of deep integration of AI LLMs and educational teaching, showcasing the beautiful vision of collaborative coexistence among teachers, students, and AI LLMs in the classroom, and foretelling profound changes in the educational ecology and the infinite possibilities of future education. However, we must also be clear that any technological innovation comes with challenges and risks. Although the integration of AI LLMs into classrooms brings numerous conveniences, issues such as technological illusions, traps of technological dependence, and potential academic misconduct cannot be ignored. Therefore, how to enjoy the benefits of technology while reasonably avoiding risks and ensuring the healthy and regulated application of AI LLMs in the field of education becomes a pressing issue we need to address. Looking ahead, the development potential of the convergent intelligence classroom in the era of digital intelligence is immeasurable, and the theoretical exploration and practical research on the convergent intelligence classroom are also new topics. This study aims to inspire further epistemological thinking and methodological exploration by more educational communities regarding the convergent intelligence classroom, jointly promoting the transformation of classroom teaching paradigms.

Editor: FAN Xiaohong

Proofreader: FANG Li