Re-examining the educational application of generative AI from the perspective of system theory: Coupling, change and symbiosis (http://doi.org/10.63386/619474)
Zheng Bofei
Institute of Marxism, Fudan University, Shanghai,200433,China;
e-mail：peterzbf@163.com
Abstract: With the rapid advancement of large language model technology, generative AI applications, such as ChatGPT, are increasingly emerging and being widely applied in the education sector, presenting both new opportunities and challenges to the education system. From the perspective of Niklas Luhmann’s systems theory, there is a dual coupling mechanism between’ generative AI, ‘a subsystem of self-generated technology, and’ education, ‘a subsystem of the social system. This mechanism involves’ indirect coupling-direct coupling, ‘where the stimulation from generative AI triggers’ self-stimulation’ in education. On one hand, the application of generative AI in education brings about profound changes, harboring new development opportunities. On the other hand, it may also introduce potential educational risks. Only by organically integrating educational logic with technological logic can we achieve mutual promotion and organic coexistence between the two in the structural coupling of multiple systems.
Key words: system theory; generative artificial intelligence; education; structural coupling; change; symbiosis
Chinese Library Classification number: G641 Document identification code: A
- Introduction
From ancient times to the present, human activities across various [①]cultures have always been intertwined with technology. Technology shapes every aspect of human life, from our ways of thinking and communicating to our behaviors and social structures, all of which are closely linked to technological innovation. In recent years, the rapid advancement of large language models (LLMs)has led to a significant leap in generative artificial intelligence (AI), particularly in areas such as language interaction, reasoning, and problem-solving. This has injected transformative power into fields like politics, economics, culture, and education. ChatGPT, a prime example of a generative AI system based on large language models, has achieved groundbreaking progress in natural language understanding and generation by processing vast amounts of data and complex algorithms, leading to widespread social impact and reshaping people’s perceptions of machine intelligence. However, it is undeniable that the complexity, opacity, and uncertainty of generative AI also pose numerous challenges to the education sector.
People’s perception of technology is crucial to the role of technology in human society and the development of their interactive practices. When technology is seen merely as a tool, it serves to enhance work efficiency and reduce labor intensity in specific tasks. However, when people’s understanding of technology transcends its material aspects, they begin to focus on the broader, deeper, and more complex interactions between technology and society. For [②]instance, Heidegger posits that ‘instrumentality’ is just the surface definition of technology[③]; the essence of technology lies in revealing the world, gradually unveiling what has been obscured . Don Idem argues that technology extends human experience and serves as a medium for changing and shaping the relationship between humans and the world . As society evolves, the connotations and extensions of technology continue to expand. To fully and deeply analyze the profound impact of generative artificial intelligence (AI) on education, it is essential to avoid viewing it merely as a tool in the educational domain. Instead, one must delve into the internal structure of generative AI as a technology, clarify the intrinsic logic of its interaction with education, and thus integrate and clearly understand the interactive and co-constructive relationship between generative AI and education. In this context, this paper aims to introduce Niklas Luhmann’s System Theory, using the social analysis framework provided by Luhmann. It conceptualizes generative artificial intelligence as a subsystem of self-creative technology systems and education as a subsystem of the broader social system. By examining the structural coupling between generative artificial intelligence (such as ChatGPT) and the educational system, the paper explores the deep connections between these two systems and analyzes the opportunities and risks that generative AI presents to education. Furthermore, it provides theoretical insights for the organic coexistence of generative AI and education, exploring feasible pathways.
Second, the structural coupling between generative AI systems and education systems
Systems theory, proposed by Luhmann, is a theoretical framework that observes society from the perspective of the distinction between ‘system and environment.’ The [④]objects of observation in systems theory are either systems (and their components) or environments (and their components) . Each system operates, maintains, and evolves based on its unique elements, media, and binary codes. Binary codes serve as one of the foundational elements of system operation, distinguishing the system from others through their binary nature. For example, the legal system operates with a binary code of ‘legal’ and ‘illegal,’ which provides clarity, consistency, and direction for each system, ensuring its effective functioning within its domain. [⑤]Luhmann argues that modern society is a functionally [⑥]differentiated society, where the entire society is divided into multiple independent subsystems, such as political, economic, and educational systems (Social Systems). Each social system is self-generated [⑦](Autopoiesis), operates in a self-referential (Self-Reference) closed loop, and constructs itself recursively through its constituent elements . Communication (Communication) is a fundamental element of social systems . Each social system is a communication system, continuously generating new communication from existing communication to ensure its ongoing operation and construction. Any event in society is a self-referential process of ‘communication creating communication,’ and the cessation of communication signifies the end of the social system .
(1) As a generative artificial intelligence system of self-created technology
Luhmann uses the concept of ‘structural coupling’ to describe the dynamic structural relationships between self-creating systems, which are situated within each other’s environments.’ Systems within [⑧]a system environment contribute to the formation of the system ‘.’ Structurally coupled systems are interdependent —— and thus not self-sufficient —— but they operate [⑨]autonomously, making them each other’s environment’ . Each system provides its complexity to another structurally coupled system. So, are technical systems self-creating systems? The concept of self-creation was initially introduced by Chilean biologists Matulana and Varela, referring to living organisms that produce and sustain themselves. Luhmann abstracted [⑩]this into the general concept of ‘systems manufacturing and sustaining themselves through the production of elements’ , applying it to social system theory. Later, following Luhmann’s lead, scholars like Andre Reichel further introduced the concept of self-creation to technical systems. When technology is conceptualized as a self-creating system, a self-referential recursive cycle is generated. Social systems and humans (including life systems and psychological systems) collectively form the environment of technical systems. Operability (the loose coupling of elements) serves as the medium of technical systems, [11]and’ working/failing’ (work/fail) acts as the binary code guiding the interaction between the system and its environment. Systems achieve self-renewal and evolution through coupling with the external environment.
Within the framework of system theory, generative artificial intelligence, such as ChatGPT, can be seen as a subsystem within self-generated technology systems. The self-generation capability of ChatGPT stems from the GPT (Generative Pre-trained Transformer) model. Transformer, a revolutionary neural network architecture for processing natural language, is a bionic model composed of a large number of interconnected artificial neurons (nodes). In this neural network architecture, neurons are arranged in a hierarchical structure, forming input, hidden, and [12]output layers. Data is processed at each layer and then passed to the next, establishing a recursive relationship. This transformation is achieved through components and technologies such as word embedding, position encoding, attention mechanisms, feedforward [13]neural networks, residual connections and layer normalization, encoder-decoder architectures, etc. Artificial neurons, due to their operability, serve as the system’s medium. Transformer, a trainable function composed of different functions, where each function is executed by a neuron, and each neuron contains a weighted link. By continuously adjusting the link weights, the trainable function completes its assigned tasks, thereby acquiring pattern recognition capabilities . With its pattern recognition capabilities, ChatGPT captures and identifies specific structures, patterns, or trends in input data, such as various linguistic features, grammatical structures, semantic relationships, and hidden logic in text, and generates meaningful output content. Pattern recognition is a key part of large language model training. Through a large amount of data and complex algorithm optimization, generative artificial intelligence can show strong language understanding and generation ability in multi-language scenarios.
In summary, ChatGPT can adapt to its input content (such as user queries) during natural language processing. It operates in a recursive cycle of ‘generation-evaluation-correction,’ using binary codes to determine whether the output is effective or not, ensuring the coherence and validity of the output. This makes it a ‘self-referential self-generation system’ that operates within specific causal logic and meaning domains. Additionally, the social and cognitive systems continuously enhance the complexity of ChatGPT’s self-construction through pre-training and interactive Q&A sessions.
(2) Education as a subsystem of the social system
Luhmann argues that, within the complex context of a functionally differentiated society, education can be seen as a subsystem of a social system with specific [14]functions. A key feature [15]of the educational system is its structural coupling with the psychological system . Education appears to be ‘impossible’ because the psychological system, being self-referential and closed, is autonomous and opaque, and individual consciousness cannot be shaped through the communication of the educational system. However, education is ‘useful’ because it can influence the self-construction of the psychological system through its coupling, thereby achieving the goal of influencing the psychological system. Based on the coupling relationship between the educational [16]system and the psychological system, Luhmann posits that the role of education is to consciously alter the psychological environment of society, which is composed of the psychological systems of each individual. The outcome of education is not about cultivating ‘better’ individuals but about enhancing personal abilities, knowledge, and skills, leading to changes in the psychological systems of individuals, improving their ability to communicate, and promoting the achievement of consensus in social communication. In other words, the function of education involves the socialization of individuals, and education can complement and correct the process of socialization.
As education has evolved, the media within the educational system have shifted from ‘students’ to ‘life course.’ This transformation in media has made it [17]difficult for education to maintain a strict binary code. Luhmann argues that the primary binary code in the educational system is ‘communicable/uncommunicable.’ Through this code, education distinguishes between teachable and non-teachable content. However, due to the closed and opaque nature of the psychological system, teachers cannot accurately determine the effectiveness of their teaching. Instead, they assess whether [18]teaching outcomes deviate from their expectations based on students ” progress/regression ‘in academic performance. Consequently, selective codes (Selection) have emerged as secondary codes in the educational system. These selective codes operate in complex ways and are characterized by strict binary choices, such as’ progress/regression, ” inclusion/exclusion, ” admission/non-admission, ‘and’ pass/fail. ‘These selective codes influence [19]the system’s operation and guide the construction of careers. Luhmann abstracts career development into the chronological structure of personal resumes, emphasizing the importance of early stages for later stages. This perspective leads students to view their educational journey as a crucial part of their ‘career,’ integrating career planning into the educational system. Students must create a favorable starting point for their future career at school. The randomness and uncertainty of career will encourage students to continuously strengthen personal accumulation during the school period, and enhance their development potential and risk resistance ability by obtaining “capital” such as degrees, diplomas and awards.
From the above analysis, it is evident that education is a system centered on communication. Its primary goal is to influence learners ‘psychological systems through the transmission of knowledge, the development of skills, and social guidance. Generative AI, which generates natural language text based on large-scale data training, has a deep connection with the communication characteristics of educational systems, thus offering the possibility of structural coupling between the two. In this process, generative AI provides a new form of communication. For instance, it can engage in’ dialogue-style ‘communication with users, generate suitable learning materials, and continuously optimize content based on user feedback, thereby enhancing communication efficiency within the educational system. Additionally, it can adapt to the requirements of the educational system through self-generation and self-referential recursive processes, providing more targeted and effective educational feedback. On the other hand, systems theory offers a macro perspective on understanding education. The functions and roles of the educational system, along with its complex operational mechanisms based on multiple binary codes, provide explanations for the behavioral logic of key subjects such as teachers and students within the system. Many contradictions exist in the educational system, such as the’ impossibility ‘and’ effectiveness ‘of education, and the’ inclusiveness ‘and’ exclusiveness’ of the educational system. These paradoxes also highlight important issues in contemporary educational construction, including teaching effectiveness, educational equity, and personalized education. In this context, the development of technology system is particularly remarkable. In the process of coupling generative artificial intelligence system and educational system structure, the enabling effect of new technology can provide new possibilities for alleviating traditional contradictions in education and solving key problems.
(3) Dual coupling: the coupling mechanism between generative AI system and education system
Throughout history, it is evident that technological advancements and educational reforms have always been intertwined. From the invention of papermaking and printing to the advent of the Internet, technological innovations have driven significant changes in the education system. As technology evolves to meet educational needs, it continuously expands its applications and adapts to new scenarios. The deep integration of generative AI with the education system not only drives development, transformation, and reshaping but also enhances the AI’s ability to adapt to and optimize real-world social contexts, with both systems evolving dynamically through mutual adjustments.
The co-evolution of generative AI as a subsystem of the technology system and education as a subsystem of the social system is achieved through the “indirect-direct” dual coupling path between the two.
The first is indirect coupling. Generative artificial intelligence (AI) transforms itself from a ‘technical entity’ into the subject of communication with the educational system through the medium of ‘language’ that serves as a bridge between human thought and the educational system. This transformation forms an indirect coupling through the intermediary of ‘language.’ Currently, the development and application of generative AI are attracting increasing attention and have become a significant topic of discussion in the education sector. Media reports, scholarly discussions, and interactions among teachers, students, and parents are gradually shaping social trends, research perspectives, and public opinions regarding generative AI, all of which influence its future direction. For instance, under the influence of technological optimism, people tend to view the educational value and potential of generative AI positively. Conversely, technopessimists may fear the decline of human intelligence in education and resist the integration of AI technology with education. Techno-neutralists, however, advocate recognizing the opportunities that generative AI brings to education while maintaining the central role of humans in education and cautiously integrating AI technology with education.
Secondly, direct coupling: the language generation capabilities embedded in generative [20]AI act as a ‘trigger’ for social communication . This provides conditions for direct coupling between generative AI and educational systems. The more social interactions mediated by generative AI, the more frequent its direct coupling with educational systems becomes. Generative AI is deeply rooted in various real-world educational scenarios, where it gathers large-scale data and diverse feedback. This data helps optimize and enhance the algorithms and text generation capabilities of generative AI, further improving its applicability to educational systems. For example, by training on interaction data accumulated from educational settings, ChatGPT can better understand educational terms, concepts, subject knowledge, and problem structures, resulting in content that better meets educational needs. This not only ensures the effectiveness of technology applications but also provides clearer direction for the development and optimization of ChatGPT.
It is crucial to note that in systems with structural coupling, changes and instability in one system can cause excitation (Irritation) in the systems it is structurally coupled with. This external stimulus does not directly affect the excited system; instead, the excited system changes according to its own operational logic, rather than undergoing linear changes due to the external stimulus. Thus, in a strict [21]sense, the system excites itself, and the concept of excitation “indicates the internal workings of the system, not the causal relationship between the system and its environment” . The structural coupling between generative AI systems and educational systems means that when generative AI systems change, it can trigger self-excitation in the educational system ——, leading to both profound positive changes and serious challenges with potential risks.
Third, the educational reform caused by generative artificial intelligence system
Currently, the widespread application of generative artificial intelligence in various educational settings has triggered a self-revolution within the education system, leading to profound changes. These changes not only involve the innovation of teaching tools and the reshaping of teaching methods but also touch on deeper aspects such as the system’s functions, operational mechanisms, and goals, offering new opportunities for the transformation and development of the education system.
(1) Enhance the basis of functional differentiation and promote the innovation and development of education
Luhmann posits that functional differentiation is a hallmark of modern society. Functional differentiation means that each subsystem in society performs a specific function. For instance, the education system is defined as a functional system because it continues to differentiate based on its educational role. The differentiation of the education system is closely tied to its autonomy. Enhancing the autonomy of the education system helps solidify and strengthen the foundation of its differentiation, thereby improving its ability to adapt to complex environments, address various challenges, and achieve sustainable development. The autonomy of the education system relies on the independence created within the system, which helps it avoid interference from other functional systems and external decisions. Some ‘technological’ [22]inventions during the process of system differentiation can help enhance the system’s independence. In the education system, the’ technological ‘invention that fosters autonomous educational communication is the classroom interaction system. The primary setting for education is interactive courses, and’ classroom interaction,’ as the form of this interactive system, is a crucial foundation for the functional differentiation of the education system. Today, the development of communication technologies such as the Internet and generative AI has expanded the boundaries of classroom interaction, establishing new connections between the classroom and the external world, placing classroom interaction in a broader social context, and thus providing new possibilities for the functional differentiation of the education system.
Specifically, the development and application of generative artificial intelligence have transformed traditional classroom interactions, solidifying the operational models of knowledge transmission and human socialization within the education system, thereby strengthening the foundation of social differentiation. Firstly, generative AI can provide real-time Q&A in class, reducing the issues of teachers being unable to respond promptly or students struggling to keep up with the teaching pace. This enhances the reflexivity of classroom interactions, reduces uncertainty in teacher-student interactions, and addresses the dual contingency (Double Contingency) in the social system, where both teachers are uncertain about student responses and students are uncertain about the state of teachers ‘teaching strategies. This improves the efficiency of knowledge transfer [23]between teachers and students and enhances teaching effectiveness. Secondly, generative AI can customize the’ dialogue ‘format through personalized instructions, allowing for the recreation of classroom interactions after class. Through one-on-one Socratic’ maieutic ‘(a method of interactive teaching[24]) interactions, it helps cultivate students’ communication skills. Furthermore, generative AI can collaborate with teachers to create a ‘human-machine collaborative’ learning ecosystem, redefining the role of teachers. This collaboration is achieved through four forms: ‘AI agent + teacher’, ‘AI assistant + teacher’, ‘AI mentor + teacher’, and ‘AI partner + teacher’, which establish a bilateral division of labor . These forms help teachers manage the numerous basic tasks for students, freeing them from tedious and repetitive mechanical teaching tasks. This allows teachers to focus more on personalized teaching and the design of the ‘invisible curriculum’ (Invisible Curriculum) based on classroom interactions. The invisible curriculum subtly conveys social norms, values, and behavioral standards that are not explicitly stated in school education but are widely present, thereby promoting students’ socialization process.
(2) Suppress the negative impact of codes and promote the reform of education mode
The binary nature of selective codes in the education system essentially serves as a standard for evaluating students ‘learning processes and performance. While this binary evaluation mechanism provides a simple and effective operational framework, playing a crucial role in maintaining the fairness and efficiency of the education system, it may also limit students’ individual development. In traditional teaching, influenced by the underlying logic of ‘pass/fail’ and ‘admit/deny’ binary codes, students often focus on achieving ‘pass’ or ‘qualified’ results to meet the standards. This focus on passing scores can lead them to ignore their personal interests and strengths, hindering their free development. Additionally, teachers often base their instruction on these binary codes, leading to the creation of standardized content and teaching plans that fail to cater to individual needs and the diverse learning paces of students.
The application of generative AI technology in the education system offers new possibilities for personalized educational development. On one hand, generative AI can cater to students ‘diverse needs during [25]knowledge transmission by acting as intelligent teaching assistants, learning companions, and mentors , providing each student with more targeted learning support. This enables students to fully tap into their potential through personalized adaptive learning. Additionally, generative AI can dynamically adjust learning content based on learners’ feedback, helping students find a suitable learning pace within their personalized learning paths. On the other hand, the diversity of generative AI applications in education can provide multi-level, multi-channel, and full-process support for students ‘personalized development. For example, the ChatGPT app store Explore GPTs offers a variety of educational applications, such as EduPath, which provides personalized school selection suggestions based on students’ needs and abilities; Real-time voice translator, which offers one-on-one oral and interpretation training; and Resume Polisher, which provides customized resume editing and polishing services. It is expected that generative AI technology will continue to adapt to users’ personalized learning needs, generating more specialized and precise educational applications, further promoting the innovative development of personalized teaching models.
(3) Reconstruct the orientation of education and training, forcing the restructuring of the education system
In a functionally differentiated society, occupation is the most significant manifestation of personal identity and a modern means of integrating individuals with society. When [26]the continuum of personal life is digitized into various ‘thresholds’ and different ‘stages,’ occupations emerge. The temporal structure of career paths dictates their close connection to the education system, where the practical and economic potential of various social occupations influences the construction of the education system. To meet the demands of social development and students ‘career growth, schools must dynamically adjust their educational systems and training objectives. The rise of generative AI has transformed the existing ecosystems in various industries. Today, tasks such as data processing, text generation, and data analysis can be efficiently completed by generative AI, leading society to place greater emphasis on’ soft ‘qualities (such as critical thinking, creativity, media literacy, and problem-solving skills) rather than’ hard ‘skills that can be replaced by AI. Generative AI has altered the skill structure required for career development, making career choices more flexible and dynamic, and redefining the starting point of the education system through the restructuring of phased’ endpoints.’
As social demands undergo structural changes, traditional career paths are being deconstructed and restructured. The education system must reassess its role in students ‘career development and adapt to these changes by reshaping the educational framework. In terms of curriculum design, it is essential to introduce AI-related courses at all educational levels—primary, secondary, and higher—to guide students in the scientific use of AI, enhance their AI literacy, and help them integrate into the AI era early on. Teaching objectives should focus on fostering students’ continuous learning abilities, shifting from a utilitarian ‘admission’ goal to a ‘lifelong learning’ vision. This approach helps students adapt to various challenges at different stages of their lives through adaptive learning. In terms of skill development, there should be a greater emphasis on cultivating high-level skills such as creativity, critical thinking, and problem-solving, as well as interdisciplinary and cross-domain capabilities. This enables students to tap into their potential and unique strengths with AI assistance, developing innovative capabilities in complex professional environments. Regarding the role of teachers, they should evolve from ‘knowledge transmitters’ to ‘learning facilitators’ and ‘career planners.’ In this process, teachers should not only use various generative AI applications for teaching innovation, guiding students to collaboratively complete personalized learning projects with AI, but also actively participate in students’ career planning, helping them understand the impact of generative AI on career development and providing career selection advice. In terms of the organizational structure of the school, it is necessary to explore the transformation from the traditional discipline-oriented organization to a more flexible interdisciplinary organization. By establishing interdisciplinary education centers and integrating teaching resources in multiple fields, students can be provided with more interdisciplinary courses and projects, so that they can acquire more comprehensive skills and adapt to the changing and complex career life in the future.
Fourth, the organic symbiosis of generative AI system and education system
In today’s context, while generative AI has injected new momentum into the development of educational systems through systematic integration, opening up a new landscape, its inherent complexity also poses numerous ethical challenges to education, such as uneven development, weakened subjectivity, and distorted values. Only by considering education within its broader system can we understand how it can organically integrate educational logic with technological logic, achieving mutual promotion and organic coexistence in the structural coupling of diverse systems.
(1) Leverage the choice code switch to promote balanced development of education
The operation of selective codes in the education system plays a fundamental role in achieving educational equity. For instance, the ‘inclusive/exclusive’ selection code aims to provide more people with opportunities for educational advancement, interaction, and socialization. However, it can also exclude some individuals through various screening mechanisms, such as school districts, economic status, and academic performance. Notably, the introduction of generative AI into educational activities has led to new forms of exclusion, known as the ‘AI digital divide.’ Specifically, while the development of generative AI offers new opportunities for educational equity, such as personalized teaching and adaptive learning systems that provide one-on-one support to students who are at a disadvantage in traditional classrooms (such as those with learning difficulties or special needs) and offer online education support to remote areas with limited teaching resources and imbalanced teacher-to-student ratios. However, there are significant differences in the ability to access and use generative AI among different social groups, which may further widen the digital divide associated with technology diffusion. The digital divide is not only about the differences in physical devices or network access but also about the gap in understanding and using technology. Schools and families in developed regions typically can more quickly adopt and utilize new technologies like generative AI, whereas students in less developed areas may be constrained by the lack of software and hardware facilities and the insufficient professional competence of teachers, making it difficult for them to benefit from these new technologies. This inequality exacerbates the existing digital divide, leaving the digitally disadvantaged even further behind in the age of AI.
From the perspective of systems theory, to bridge the new digital divide caused by generative AI, it is essential to explore how to leverage the ‘inclusive’ side of selective codes through multiple measures, promoting balanced educational development. First, we need to enhance the accessibility of hardware. This involves further improving Internet infrastructure to ensure that schools and families in remote areas can access the digital world. Second, we need to expand the availability of software. We should build open and shared educational platforms for generative AI and online AI literacy courses, promoting them nationwide. Additionally, we need to create user-friendly human-computer interaction interfaces to lower the barrier to using generative AI applications. Third, we need to improve the inclusiveness of training programs. We should provide widespread AI digital literacy training for teachers across different regions, enhancing their ability to use and even develop generative AI applications, and integrating these technologies scientifically and reasonably into teaching practices. Overall, within the framework of the ‘inclusive/exclusive’ selective codes in the education system, we need to increase resource allocation to underdeveloped areas and populations, enabling more students to integrate into the educational environment of generative AI, thereby promoting balanced educational development.
(2) Go beyond the limitations of machine communication and maintain the independence of educational subjects
From the perspective of systems theory, generative artificial intelligence (AI) is essentially a self-generated technology system that uses artificial neurons as its medium. In the educational system, the self-generation capabilities, strong generative abilities, and high degree of ‘human-like’ characteristics of generative AI have transformed it from a ‘tool’ and ‘semantic’ intermediary into an alternative ‘self-generated entity,’ challenging human subjectivity in educational activities. By leveraging deep learning mechanisms, generative AI generates text, images, videos, and other content that mimics human creativity through extensive data support and complex algorithm optimization. This often gives users the illusion of interacting with human intelligence during human-computer interactions. However, while generative AI exhibits human-like features, it cannot truly understand the meanings and emotions embedded in language like humans can. Over-reliance on generative AI may weaken students ‘creative thinking and independent exploration skills, reducing them from’ knowledge acquirers ‘and’ creators ‘to mere’ receivers ‘and’ consumers ‘of generated content. Additionally, the content generation method of generative AI, which is oriented towards user preferences, needs, and habits, may lead students to unconsciously project their real emotions into human-computer interactions. The technical nature of generative AI means it cannot establish genuine emotional connections with students. Over time, as real interpersonal communication is increasingly replaced by human-computer interaction, the deep interactions and emotional bonds between students and teachers, parents, and friends will gradually fade and dissolve, potentially leading to the machine replacing the human’s central role in the educational system.
From a systems theory perspective, the medium of the education system has shifted from ‘students’ to ‘life processes,’ leading to the concept of lifelong learning. The impact of generative AI on educational subjectivity is not confined to schools but extends throughout the entire process of socialized learning. To solidify the central role of individuals in the education system, it is essential to build a comprehensive, human-centered lifelong learning framework at both school and societal levels. By offering a variety of educational resources, we can ensure that each individual, at different stages of their growth, can access education and training opportunities that align with the evolving times based on their needs and development goals. This approach not only ensures students ‘autonomy in their studies during their time at school but also enables individuals in society to proactively and continuously enhance their skills in an AI-driven environment. Only by adopting a’ human-centric’ approach and grounding educational reforms in the principles of human life development (Su Huili et al., 2024) can individuals, as the subjects of education, achieve free and comprehensive development in a society where humans and machines coexist.
(3) Organically integrate the logic of technology to avoid the alienation of educational value
Technology is often seen as a tool to enhance work efficiency and streamline workflows, with its underlying logic often hidden in the guise of ‘technological neutrality.’ However, in reality, as technology systems evolve, they establish a coupling relationship with their environment through a binary code of ‘effective/ineffective,’ allowing each system [27]to operate according to its own logic while also striving to adapt to this binary code. In other words, to some extent, technology constructs reality by adapting to the social construction of technological logic. The [28]rapid advancement of generative artificial intelligence (AI) technology poses a risk of technological logic intrusion and value orientation distortion in the education system. If generative AI is continuously introduced into teaching practices without critical evaluation, the instrumental rationality inherent in generative AI will inevitably permeate every aspect of intelligent education, leading the education system to unconsciously internalize the instrumental rationality of the technology system, such as prioritizing efficiency and maximizing utility. This can result in educational goals, values, and processes being constrained by the utilitarian shackles of instrumental rationality, with quantifiable indicators like pass rates, admission rates, and employment rates becoming the standards for measuring educational success. Meanwhile, non-quantifiable indicators such as moral qualities, critical thinking, responsibility, and creativity are gradually marginalized. This leads to the simplification of the complex connotations of education by quantifiable standards, causing the educational goal to shift and fall into the trap of narrow development. For example, schools may increasingly focus on cultivating students into ‘qualified’ workers who can adapt to technological development and market demands, while neglecting the attention and cultivation of students’ multidimensional abilities.
New technological interventions often introduce a new dimension. Generative AI, with its inherent autonomy and unique patterns, not only serves as a tool and means in educational practices but also integrates its technical logic into the education system, influencing classroom teaching models, resource allocation, and management systems. As the education system undergoes self-stimulation, it must leverage the empowering effects of generative AI technology, critically accept and adapt to its technical logic, promote the organic integration of technical and educational logics, and ensure that the value rationality of education guides the instrumental rationality of generative AI. The ultimate goal of designing educational plans and conducting educational activities is to achieve the free and comprehensive development of individuals.
[US] Don Id: Technology and the World of Life: From Eden to the World of Dust, translated by Han Lianqing, Beijing: Peking University Press, 2012, p. 22.
Marti Heidegger，The Question Concerning Technology and Other Essays ，Translation by William Lovitt，New York: Harper & Row, Publishers, p.3-36.
[Tang Id: Technology and the World of Life: From Eden to the World of Dust, translated by Han Lianqing. Beijing: Peking University Press, 2012, p. 22.
Georg Kneer, Armin Nassehi. “Luhmann’s Theory of Social Systems”, translated by Lu Guixian, Taiwan: Juliu Book Company, p. 85.
Georg Kneer, Armin Nassehi. “Luhmann’s Theory of Social Systems”, translated by Lu Guixian, Taiwan: Juliu Book Company, p. 76.
Claudio Baraldi, Giancarlo Corsi& Elena Esposito，Unlocking Luhmann A Key-word Introduction to Systems Theory，Translation by Katherine Walker，Wetzlar: Majus-kel Medienproduktion GmbH，2021，p.45.
[Niklas Luhmann, Social Systems: An Outline of a General Theory, translated by Lu Guixian and Tang Zhijie, Taiwan: Nuan Nuan Book House Cultural Enterprise Co., LTD., 2021, p. 60.
Luhmann, N，Social systems，Stanford: Stanford University Press，1984，p.213.
Georg Kneer, Armin Nassehi. “Luhmann’s Theory of Social Systems”, translated by Lu Guixian, Taiwan: Juliu Book Company, p. 83.
Georg Kneer, Armin Nassehi. “Luhmann’s Theory of Social Systems”, translated by Lu Guixian, Taiwan: Juliu Book Company, p. 76.
André Reichel，“Technology as System: Towards an Autopoietic Theory of Technology，” International Journal of Innovation and Sustainable Development，Vol. 5, Nos. 2/3, 2011, pp.105-118.
Zhang Haoyan, Lu Wentao, Yu Runze, and Deng Zhijiang (2024). “Research Status of Large Language Models”, Radio Engineering, 2024, No.9, pp. 1-15.
He Bin, “Analysis of the Influence of ChatGPT on Education —— A Study Based on Ecological Perspective”, Jiangsu Educational Research, 2023, No.23, pp. 21-26.
Claudio Baraldi, Giancarlo Corsi，Niklas Luhmann Education as a Social System，Cham: Springer Nature，2017，p.39.
Luhmann, N，Das Erziehungssystem der Gesellschaft，Frankfurt: Suhrkamp:2002，p.80.
Claudio Baraldi, Giancarlo Corsi，Niklas Luhmann Education as a Social System，Cham: Springer Nature，2017，p.49-50.
Claudio Baraldi, Giancarlo Corsi，Niklas Luhmann Education as a Social System，Cham: Springer Nature，2017，p.88.
Claudio Baraldi, Giancarlo Corsi，Niklas Luhmann Education as a Social System，Cham: Springer Nature，2017，p.62-67.
Claudio Baraldi, Giancarlo Corsi，Niklas Luhmann Education as a Social System，Cham: Springer Nature，2017，p.68.
André Reichel，“Technology as System: Towards an Autopoietic Theory of Technology，” International Journal of Innovation and Sustainable Development，Vol. 5, Nos. 2/3, 2011, pp.105-118.
Georg Kneer, Armin Nassehi. “Luhmann’s Theory of Social Systems”, translated by Lu Guixian, Taiwan: Juliu Book Company, p. 90.
Claudio Baraldi, Giancarlo Corsi，Niklas Luhmann Education as a Social System，Cham: Springer Nature，2017，p.72-76.
Mao Gang and Wang Lianghui, “Man-machine Synergy: Understanding and Constructing the Way of the Future Education World”, Educational Development Research, No.1, 2021, pp. 16-24.
Yang Zhongying, “The core ideas, connotation and application implications of the ‘Implicit Curriculum’ educational paradigm”, Social Scientist, No.4, 2023, pp. 147-154.
Wu Di, Li Huan, Chen Xu, Analysis of the Influence of General Large Model Education Application on Artificial Intelligence, Open Education Research, No.2, 2023, pp. 19-25.
Claudio Baraldi, Giancarlo Corsi，Niklas Luhmann Education as a Social System，Cham: Springer Nature，2017，p.69.
André Reichel，“Technology as System: Towards an Autopoietic Theory of Technology，” International Journal of Innovation and Sustainable Development，Vol. 5, Nos. 2/3, 2011, pp.105-118.
Yu Mengqian, He Xiaofang and Niu Shuying, Between the Way of the Instrument: On the Instrumental Rationality and Value Rationality of the Application of New Technology in University Teaching, Contemporary Education Forum, No.6, 2024, pp. 68-75.