Yan Ruifeng: Ethical Risks and Mitigation Paths of Generative AI Empowering Ideological and Political Education—An Investigation Based on the Perspective of the Subjectivity of Educational Objects
At present, the wide application of generative artificial intelligence (AI), represented by Large Language Models (LLMs), has significantly enriched and expanded the practical scenarios of "AI + Education," driving the continuous evolution of the traditional educational ecosystem toward digitalization and intelligence. In recent years, China has deeply implemented the national strategy for educational digitalization, actively exploring the deep integration of intelligent technology and educational practice. Top-level designs and policy documents concerning the digitalization and intellectualization of education have been continuously issued and systematically implemented. The necessity of the "technologized existence" of education is prompting a rapid transformation toward a new state of integrated virtual-real patterns. Empowered by generative AI, Ideological and Political Education (IPE) has gained a powerful transformative force. Leveraging comprehensive technical advantages such as natural language processing, big data analysis, deep learning, and multi-modal fusion, generative AI has broadened the space for innovation in IPE regarding personalized learning, the construction of knowledge graphs, and immersive interaction. Consequently, it exerts an important influence on the value shaping, knowledge acquisition, and capacity cultivation of the educated. However, "technology is a sharp tool for development, but it can also be a source of risk" [1]. The one-dimensional expansion of instrumental rationality in intelligent technology may also dissolve the foundation of value rationality in IPE, causing a deep "colonization" of the educational process by technical logic. This triggers a reactive alienation (yìhuà) of the subjectivity of the educational subjects and induces a series of ethical risks, necessitating an active search for countermeasures at both theoretical and practical levels.
I. The Technical Logic Leading to the Dissolution of Subjectivity in Generative AI-Empowered Ideological and Political Education
Marx pointed out: "The productive life is the life of the species. It is life-engendering life. The whole character of a species—its species-character—is contained in the character of its life activity; and free, conscious activity is man's species-character" [2]. The cultivation of the subjectivity of the educational subject in IPE is an important path to achieving the free and comprehensive development of human beings. In practice, subjectivity is reflected in multiple dimensions, including autonomy, self-activity (zìwèixìng), agency, consciousness, and freedom. The contradiction between generative AI technology and IPE is, to a large extent, embedded in the technology's function of dissolving subjectivity. By constructing a technical explanatory framework based on data-driven discipline, algorithmic cognitive substitution, and technical mediation, we can provide the necessary theoretical fulcrum for deconstructing the ethical risks of IPE in the era of digital intelligence.
(1) Data-driven discipline triggers the technical deconstruction of value subjectivity
Values education is an important way to maintain the subjective status of the educational subject and achieve identification with the mainstream ideology. Through interaction mechanisms based on intersubjectivity, it forms a value consensus within the subject’s cognitive processes of evaluation, judgment, and choice, thereby achieving the goal of cultivating value rationality. At the level of philosophical ontology, value subjectivity links the value of the object with the existence of the subject. The diachronic [3] evolution of value forms always unfolds around the historical existence of the subject, thereby constructing a transformation mechanism of "subjective awakening—value identification—conscious practice" in the dimension of educational practice, ensuring that individual value choices remain highly consistent with the guidance of core social values. However, in the context of generative AI's deep intervention in IPE, data-driven discipline reshapes the value structure of the educational subject, gradually forming a new mode of power operation. Utilizing technical devices such as data analysis, algorithmic recommendation, behavioral modeling, and digital profiling, it performs fine-tuned regulation of individual cognition and behavior, thereby transforming the behavioral trajectories of the educational subject into quantifiable, predictable, and controllable data flows. This subjects the independence carried by value subjectivity to an unprecedented technical deconstruction, where the individual’s ability to independently perform value reflection and choice is deeply interfered with or even controlled by technical logic.
First is the data-driven locking of the value-cognitive framework. In teaching practice, when information pushing, knowledge acquisition, and autonomous learning are all dominated by algorithms, the breadth of an individual's value cognition is limited by "information cocoons," the underlying logic of cognition is reshaped by data, the subject's critical thinking is dissolved by instant feedback mechanisms, and the freedom of value choice is coerced by data. Second is the temporal disappearance of value reflection. The real-time data correction function of AI systems gives rise to "granular existence" (wēilìhuà shēngcún). Amid continuous behavioral adjustment, individuals lose the temporal depth required for value sedimentation; internal value standards are surrendered to external data indicators, leading to a humanistic predicament of "value dilution" in IPE. At this point, educational platforms use matching algorithms to define emotional value, and teaching evaluation systems quantify morality procedurally; the intersubjective construction of values is alienated into a process of debugging technical parameters. Third is the algorithmic shaping of values through data. Data-driven discipline uses technical means to transform human behavior, emotions, and thoughts into calculable, predictable, and intervenable data objects. Affective computing, through multi-modal recognition mechanisms (such as facial expressions, voice, and breathing), converts complex value judgments into quantifiable data symbols, perceiving and analyzing them through psychological, behavioral, and physiological measurements. For example, a system might dynamically assess a student's degree of identification with a specific political issue through facial expression intensity coefficients and adjust teaching strategies accordingly—this is essentially a hypothesis of a strong correlation between biological signals and value judgments. The deconstruction of value subjectivity by data-driven discipline reveals the technical aspect of the crisis of modernity; however, this deconstructive effect is not the inevitable destiny of technological development, but rather the consequence of humans surrendering the power of value judgment to technical systems.
(2) Algorithmic cognitive substitution leads to the technical deprivation of cognitive autonomy
In the field of IPE, maintaining the autonomy and independence of thinking constitutes an inherent requirement for the cultivation of a modern personality and serves as the cornerstone for achieving the free and comprehensive development of the human being. From the perspective of Marxist technological critique, the essence of algorithmic cognitive substitution is the epistemological extension of labor alienation in the intelligent era, thereby forming a new type of epistemological predicament based on digital alienation. When generative AI deeply erodes the essence of IPE with an alienated technical logic, the fact of instrumental rationality’s encroachment upon value rationality causes the autonomy of the subject's thinking to be technically deprived. The subject passively converges toward a cognitive paradigm manipulated by algorithms, resulting in the technical domination of subjectivity—this is the technical deprivation of autonomy of thought caused by algorithmic cognitive substitution.
First is the technical reconstruction of the cognitive architecture. On one hand, digital interfaces implement cognitive capture through the mechanism of the attention economy. Algorithmic recommendation systems, through a closed-loop domestication mechanism of "information cocoon—cognitive inertia—algorithmic dependence," cause the educational subject to gradually lose the ability for autonomous information filtering, thereby forming a disciplinary effect dominated by interface cognition and a cognitive domestication effect dominated by algorithmic authority. On the other hand, the deep intervention of intelligent technology in human thinking processes leads to a structural estrangement between the cognitive subject and their own thinking ability. This is essentially an extension of Marx's theory of "alienated labor" into the cognitive dimension in the digital age. When AI assumes high-level cognitive labor such as knowledge retrieval, data analysis, and value judgment, technical tools are alienated from cognitive aids into dominant forces, easily resulting in the outsourcing of thought processes. At this point, core cognitive links such as knowledge construction, contradiction analysis, and systematic reflection are forced to be surrendered to intelligent systems, leading to a systematic degradation of the subject’s metacognitive abilities. Through the closed-loop domestication mechanism of algorithmic recommendation systems, the subject's autonomous cognition becomes an appendage of the algorithmic cognitive system, causing the "algorithmization" of cognitive patterns. Second is the algorithmic colonization of the thinking process. The algorithmic substitution of critical thinking dominated by intelligent technology prompts a shift in human thinking from dialectical logical deduction to pattern-matching responses; from value-reflection cycles to instant-feedback dependence; and from historical contextualized understanding to fragmented information processing. When educational subjects rely on intelligent devices to complete high-level thinking tasks, their metacognitive cycles—replaced by technical agencies—fall into an "outsourcing trap" from which they cannot extricate themselves. The instant answers provided by algorithms weaken the motivation for autonomous reflection, making it easy for individuals to exhibit "algorithmic truth" and "either-or" mindsets (sīwéi dìngshì) when analyzing problems. The hidden disciplinary effect of algorithmic logic on human thinking simplifies complex human thought processes into calculable and quantifiable standardized paths, gradually weakening autonomous thinking ability within information cocoons, leading to the quiet collapse of cognitive autonomy during algorithmic domestication. The reasoning mode of generative AI is essentially a processual simulation of human non-linear thinking; its technical mimetic quality obscures the Marxist Law of the Negation of the Negation [4], causing the subject’s ability for contradiction analysis to become "flattened."
(3) Technical mediation causes a structural transformation of intersubjectivity
Effectively handling the relationship between the subject and object of education is an important aspect of ideological and political work. The communicative paradigm, represented by intersubjectivity, is the key link for achieving value transmission and emotional connection. Generally speaking, subjectivity and interactivity are the prerequisites for intersubjectivity: "Intersubjectivity views education as a communication between subjects, thereby confirming education as an authentic way of human existence." In traditional contexts, technical artifacts exist only as objects to be transformed by humans; the subject-object relationship between the two is relatively clear, and there is no intersubjectivity based on subjective interaction between humans and things. However, in the process of human-AI interaction, the quasi-subjective characteristics of intelligent systems enable them to gradually possess the practical ability to autonomously transform the objective world. This transforms the traditional "Subject–Object" relationship between humans and technology into a "Subject–Subject (Quasi-subject)" relationship, thus leading to a structural transformation of intersubjectivity. In the scenario of IPE driven by AI, the original "Subject (Teacher)—Subject (Student)" binary structure is transformed into a "Subject (Teacher)—Subject (Quasi-subject)—Subject (Student)" ternary structure. Changes in the interacting subjects likewise cause a structural transformation of intersubjectivity. First, the algorithmic reconstruction of educational relationships triggers the alienation of intersubjectivity. The two-dimensional embodied interaction between "teacher and student" in traditional IPE is alienated into a three-dimensional technical mediation mode of "Human—Machine—Human." Interaction is reduced to a symbolic exchange game centered on data flows. The multi-modal embodied interactions crucial to interpersonal communication—such as the transmission of emotions through fluctuations in mood and tone—are weakened at the virtual interface, leading to the predicament of "poverty of interaction." Second is the double-migration crisis of the subjective structure. The quasi-subjective development of intelligent systems gives rise to a zero-sum game in the distribution of subjectivity, causing the professional authority of educators to migrate toward algorithmic systems (e.g., teaching evaluation rights being surrendered to learning analysis systems), and the value construction of the educated to migrate toward data models (e.g., ideological dynamics being simplified into emotional analysis curves), resulting in the "hollowing out" and "fragmentation" of subjectivity.
In scenarios where intelligent technology intervenes in education, the structural transformation of intersubjectivity generally takes two forms. On the one hand, under the influence of Weak AI, technical mediation expands the subject's space for subjectivity and enables free transition between the virtual and the real across time and space. As the Canadian scholar Marshall McLuhan pointed out, the extension of human cognitive processes by technology is "just as our sense organs and nervous systems are extended by various media." Technology as a mediatory existence not only reconstructs the interface between physical and digital spaces but also dissolves the essential distinction between organisms and artifacts, expanding the extension of human subjectivity. On the other hand, the ontological mediatory effect of Strong AI drives intelligent systems to break through the objective boundaries of traditional technical artifacts, thereby generating cognitive entities with quasi-subjectivity. This paradigmatic leap not only triggers a cognitive surrender of human subjectivity to algorithmic systems (such as the surrender of decision-making) but also leads to a fundamental change in the subjective structure. Through the continuous evolution of deep learning and autonomous decision-making capabilities, Strong AI gradually breaks through the objective positioning of traditional technical artifacts, giving rise to a dual-structured entity that possesses both tool-objectivity and cognitive-subjectivity, forming a complex interactive system that breaks the traditional subject-object binary framework. Although it can be regarded as "a high-level object in the guise of subjectivity" or "pseudo-intersubjectivity," "a Strong AI with autonomous consciousness is no longer a tool for executing programs; its will and actions no longer represent the consciousness and actions of a master or humanity."
II. The Ethical Risks to Subjectivity Brought by Generative AI-Empowered Ideological and Political Education
The dual-natured character of technology provides an important insight for the technologization of education: in practice, we must fully explore the positive efficacy of technology in promoting educational development, while remaining constantly alert to the various risks and challenges it brings. In reality, the alienation of educational technologization is prone to causing a crisis of subjectivity—the "humanization" of machines and the "machine-ization" of humans—resulting in problems such as symbolic identity, interpersonal alienation, and obstacles to consensus-building. Furthermore, in the context of generative AI-empowered IPE, the excessive display of technology's one-dimensional instrumental rationality will cause the subjectivity of the educational subject to be dominated by technical logic and undergo reactive alienation, thereby deriving multiple ethical risks under the maladjustment of the relationship between technology and ethics.
(1) The weakening of subjective conviction leads to the disorder of value beliefs
Ideal and belief education is a crucial path for guiding students toward establishing a correct worldview, outlook on life, and system of values. Belief mainly refers to the subject's firm conviction in a certain thought, idea, or culture; its formation and change are profoundly influenced by the state of ideology. Value faith primarily refers to the value ideals or commitments collectively chosen by subjects based on shared expectations of value goals; its formation relies mainly on specific ideals and beliefs—or, one might say, belief is the foundation of faith, while faith is a higher-order manifestation of belief. As an inevitable presence in class rule, ideological and political education systematically shapes specific systems of value faith through the value-integration function of the dominant social ideology.
"Firmness in ideals and beliefs comes from firmness in thought and theory." The thoughts and theories within ideological and political education possess strong ideological attributes and serve as an important source for systematically shaping the subject's ideals, beliefs, and value foundations. The powerful knowledge-production capabilities of generative AI can create favorable conditions for studying thought and theory. However, its generated content possesses a high degree of autonomy, creativity, and uncertainty. When algorithmic models deviate from or run counter to the mainstream ideology, intelligently generated content may transmit heterogeneous values to the bridgeable subjects of education through "soft indoctrination," thereby shaking the foundations of their ideals and beliefs and weakening their faith in the mainstream values of society.
The ideological risks produced by generative AI are inextricably linked to the sources of training data and technical defects. On the one hand, there are ideological risks stemming from data information rooted in negative value forms. In practice, algorithm design is deeply influenced by value positions such as utilitarianism, virtue ethics, and deontology. AI agents laden with pluralistic ethical values, cultural forms, and subjective biases can "control social power through a technical logic that seemingly controls nature, thereby shaping people's value concepts and behavioral patterns." For example, the training corpus for the ChatGPT large language model primarily comes from Western value systems, and the value orientation carried by its generated content is dominated by mainstream Western values. If the training data used by the AI contains political bias, racial bias, historical nihilism [6], money worship, extreme liberalism, or other content that violates Socialist Core Values, then the intelligently generated or algorithmically recommended content is highly likely to match these views. The AI agent then becomes an intelligent "source point" for a large number of ideological risks. Once such harmful content spreads widely in the practice of ideological and political education and deeply affects the students' worldview, outlook on life, and values, the mainstream social ideology and values risk being marginalized. This could cause a disorder of value faith among the student population and weaken their positive ideals and beliefs. On the other hand, there are ideological risks caused by technical defects. First, the "filter bubbles" constructed by algorithmic recommendation systems through user profiling cause the scope of information contact to undergo hypergeometric contraction and information homogenization, leading to a decay in cognitive diversity and a blunting of value-based critical faculties. Second, the technical defect of "AI hallucinations" leads to "technically unconscious" lying and the fabrication of facts, creating a crisis of ontological trust and shaking the rational basis of value judgments, which poses a potential threat to values education. Third, cultural biases in the training data of large models may undergo latent infiltration through intelligent teaching systems and transmit related implicit biases to students. This causes values to degenerate from an intersubjective network of meaning into an algorithmic parameter space, thereby planting hidden dangers for value distortion.
(2) Inertia in Subjective Thinking Leads to the Obscuring of the Innovative Spirit
The cultivation of innovation capacity is not only an essential requirement for educational modernization but is also a key focal point for ideological and political education to realize its mission of "molding the soul and nurturing the person" [7]. Ideological and political education takes the reconstruction of human ideas as its value orientation, and the enhancement of the subject's innovation capacity fundamentally depends on the cultivation of an innovative spirit centered on innovative thinking. Innovative spirit, innovative personality, and innovative talent are important contents of ideological and political education. Ideological and political workers should play the role of "engineers of the soul" who inspire students' creative thinking and creative imagination and cultivate creative talents. Precisely for this reason, the importance of ideological and political education in cultivating students' creative thinking is concentrated in providing Marxist philosophical guidance, establishing correct value orientations, shaping creative personalities, and providing scientific methods of thinking. However, the excessive reliance of students on intelligent technology during the process of innovative practice allows the algorithmic recommendation, data modeling, and automatic generation of technology to replace innovation elements such as problem discovery, the process of critical thinking, and original expression. This leads to the degradation of innovative activities from subjective creation to technical operation; innovative practice thus forms a tension-filled contradiction between "technological enhancement" and "the defense of subjectivity." From the perspective of knowledge innovation, generative AI currently finds it difficult to complete high-level innovation tasks and can only simulate human thinking to engage in low-level creative work. The innovation capacity it possesses belongs to the type of "creating something from something already there" (yǒu zhōng shēng yǒu) [8]; it cannot yet produce new knowledge "from nothing to something" (cóng wú dào yǒu) as humans do. Furthermore, although generative AI can complete combinatorial innovation based on data correlations, it is primarily "low-level innovation" or "pseudo-innovation." It still possesses essential limitations at the level of breakthrough or original innovation and cannot yet meet the needs of high-level innovation.
Faced with the convenience and high efficiency brought by the deep integration of intelligent technology and ideological and political education, if the subjects of education lack sufficient discernment or self-control, they easily develop an excessive reliance on AI agents. This obscures and weakens the active role of Marxist scientific thinking and creative personality, fostering and magnifying the subject’s cognitive inertia and leading to a proliferation of shallow and weak thinking. This generates a widespread crisis where the innovative spirit is obscured. On the one hand, excessive technological reliance triggers a "dimension reduction" in the subject's cognition and a structural atrophy of critical thinking. Relying without limit on generative AI for knowledge production easily fosters a trend of unthinking "borrowism" (ná lái zhǔ yì) [9] in knowledge, causing the subject's powers of thought, judgment, and creativity to gradually weaken, eventually leading to the inertia of subjective thinking and the erosion of the critical spirit. Algorithmic path-dependence causes alienation in the subject's cognitive process; the instantaneous supply of answers by AI causes the subject to skip the necessary cognitive stage of "suspending the question for dialectical reflection," leading to the continuous solidification of cognitive inertia. Sustained reliance on AI to complete knowledge integration and value judgment causes the mechanisms of self-monitoring and self-regulation to gradually degrade, resulting in the continuous atrophy of the subject's metacognitive abilities. The machine-led mode of knowledge production is, in essence, an algorithmic reshaping of the dialectical cognitive process, enticing the user to fall from an active subject of practice to a passive consumer of technical products, thereby producing the objectification of the subject. On the other hand, the proliferation of "intelligent creation" deeply erodes the subject's innovative agency. The subject's innovative agency is replaced by rampant "intelligent creation," and an innovation cycle where "bad money drives out good" weaves a low-level "innovation cocoon," further leading to a lack of creative drive, the weakening of creative ability, and the thinning of innovation consciousness. Consequently, the continuous reinforcement of low-level innovation coexists with a crisis of innovation homogenization; innovation degenerates into a technical game of parameter adjustment, and innovation energy is locked within a limited space, resulting in a creative poverty of knowledge production.
(3) The Virtualization of Subjective Perception Causes the Dilution of Humanistic Emotion
Humanistic emotion is, in essence, an embodied representation of intersubjectivity and an important bond for maintaining social ethical relations and achieving individual development. Within this, the subject's capacity for perception is the epistemological prerequisite for generating emotional capacity, playing an irreplaceable role in emotional transmission within ideological and political education. Ideological and political education is a piece of educational work full of thoughtfulness and humanism; it places particular emphasis on emotional exchange and humanistic care brought about by firsthand experience based on actual perception. Through this, it continuously promotes the emotional development and moral cultivation of the subjects of education. The embodied emotional cultivation paradigm of traditional ideological and political education relies on the "intersubjective dialogue" of physical presence. It is able to use contextualized perception and concrete interaction to create a clear advantage of "presence" where the subject and object blend, realizing a rational transition from sensory stimulation and emotional arousal to value reflection and behavioral internalization. This blending of subject and object makes perception both real and three-dimensional, facilitating the expression and transmission of humanistic emotions between teachers and students and achieving emotional exchanges with "educational warmth." However, AI agents themselves do not possess the capacity for emotional experience; they are merely rational interacting subjects. An AI agent's cognition and perception of emotion are built upon rational, objective data; its emotional simulation is essentially a process of the algorithm reproducing emotional outward behaviors in a statistical sense. It lacks the real emotional capacities possessed by humans, such as joy, anger, sorrow, happiness, empathy, affection, or a sense of beauty. On the one hand, the "standardized empathy" generated through data analysis and affective computing by virtual teachers generated by AI systems easily severs the organic link between emotional expression and internal values. In the illusion where "the simulacrum replaces the truth" or where there is a "misalignment between simulacrum and truth," it induces a crisis of the symbolization of emotional exchange. On the other hand, although existing AI technology can simulate sensory stimulation to a certain extent, it cannot reproduce embodied elements such as the transmission of warmth and the arousal of mood found in actual perception, thereby leading to the dissolution of embodied cognition.
Virtual teachers lack real emotional feedback and the capacity for emotional empathy. Over-reliance on them will seriously hinder the embodied cultivation of students' humanistic emotions. The proliferation of virtual scenes easily leads to phenomena such as fractures in emotional cognition and the "disembodiment" of empathetic experience, which can result in "emotional blunting" when facing real social problems. To a certain extent, this will dissolve the humanistic nature of ideological and political education. The rich humanistic emotional exchange and personality-based education that originally existed between teachers and students may be replaced by cold, dull, and unfeeling symbols. The meaning and value of the subject's existence are gradually dissolved and assimilated by the machine, spreading toward the extreme direction of the virtualization of subjective perception. Although the human-like capabilities of AI technology are constantly increasing, it lacks the unique human elements of free will, inspiration, emotion, and ethics. This means its positive role in emotional understanding, spiritual empathy, and mental feeling remains limited. Human-machine interaction merely simulates human behavior through forms of judgment and reasoning such as data, algorithms, code, and models; ultimately, it cannot understand the human spiritual world or the essence of emotion. The excessive control of emotional education by AI causes "values and meanings that should have been possessed to be filtered out; the collision of thinking and values is simplified into programmed, singular algorithmic operations; and interpersonal relationships that were originally full of wisdom and warmth become a cold, mechanical process of information-symbol interaction. This will inevitably lead to a decline in individual empathy and a reduction in the sense of experience." AI cannot replace human emotional power, imagination, or moral power. The alienation of the human-machine relationship—leading to human-machine homogenization or scenarios where humans depend on machines—will cause ideological and political education, which should be full of "human touch" (rén qíng wèi), to become cold "machine education."
(4) The Fading of the Subject's Sense of Responsibility Prompts the Deviation of Moral Behavior
A sense of responsibility plays an important role in shaping an individual's moral behavioral norms, and the guidance and cultivation of the subject's sense of responsibility is an important aspect of ideological and political education. The encroachment of the tool-rationality of intelligent technology upon subjectivity both enhances the controllability of the educational process through means like data and algorithms and creates a dilemma of the dissolution of subjectivity at the level of the ethics of responsibility. This forms a structural tension between technical empowerment and the ethics of responsibility. When technical rationality breaks through its instrumental boundaries and evolves into a dominant force, the technical deprivation of human autonomy by intelligent technology will cause the subject to fall deep into a digital virtual mirror. The subject's sense of responsibility is reconstructed by algorithmic logic, causing it to degenerate into a passive reaction to external discipline and lose the practical foundation of "self-legislation," thereby triggering a crisis in the ethics of responsibility. In the educational field empowered by generative AI, the multidimensional responsibility network composed of subjects such as system developers, teachers, managers, and students exhibits characteristics of dispersed responsibility subjects, blurred responsibility boundaries, and fragmented responsibility cognition. This creates "responsibility blind spots" or "responsibility vacuums." For instance, system developers focus on product performance and responsibility for design defects, while teachers focus on classroom application and responsibility for use. In a network ecology where multiple responsibility subjects coexist, the phenomena of the suspension and dilution of responsibility intertwine, forming a "bystander effect" mechanism of responsibility diffusion, which leads to a series of moral deviations as the subject’s sense of responsibility is technically deconstructed.
First is the moral deviation brought by platform anonymization. Platforms strip away users' real identity markers through technical means, constructing a virtual space that seems free and equal. However, this invisibly provides the subjects of education with a means of escaping accountability...
The "digital mask." Anonymized intelligent platforms may weaken the sense of moral responsibility, inducing the targets of education to evade accountability on the grounds that "virtual behavior has no real-world consequences." For instance, the incidence of deviant behavior—such as linguistic violence in platform discussion areas—is significantly higher than in physical classrooms. To a certain extent, this confirms the negative effect that identity concealment or blurring exerts by suspending the responsibility of the subject. Second is the ethical harm caused by the calculability of morality. When moral evaluation is transformed into data indicators within an intelligent environment, one may fall into the trap of "calculating virtue" [10], leading students to simplify learning outcomes into "data tags." When moral behavior is reduced to quantifiable "data tags" (such as interaction frequency and emotional fluctuations), students easily fall into the "Dataism" cognitive trap, thereby alienating moral practice into the technical manipulation of "data farming." For example, to pursue platform points or algorithmic recommendations, they may deliberately manufacture high-frequency interactions, forge emotional feedback, or even reduce moral reflection to a test-taking strategy of "matching the standard answer." Third is the crisis of academic integrity. When technical mediation breaks through instrumental boundaries to restructure the logic of knowledge production, the subjective foundation of academic creation is shaken, causing creative labor to tend toward alienation. This leads to a paradigm shift in the cognition of responsibility from "author-centered" to "human-machine hybrid" or even "machine-led." Currently, the widespread application of generative AI in the academic field has catalyzed an efficiency revolution in scenarios like thesis writing, design creation, and report writing, but it has simultaneously triggered academic misconduct such as plagiarism, cheating, fabrication, and copyright disputes. In particular, functions like intelligent generation, rewriting, and polishing have blurred the boundaries of originality in academic creation, spawning misconduct such as direct copying, latent [N] patching, and AI-coordinated fabrication. These phenomena are rooted in the "responsibility cognition drift" effect caused by the "black box operation" [11] of technology on the subject’s responsibility, causing academic subjects to exhibit a clear attitude of denial and a tendency toward indifference when facing issues such as fact-checking, copyright attribution, and data authenticity. Fourth is the crisis of the debasement of the subject's personality. In the digitized educational field dominated by AI technology, the tension between the proliferative construction of the subject's virtual identity and the continuous atrophy of their real-world identity causes digital persona to exert a strong squeeze on real-world personality. Under technical discipline, real-world personality is gradually weakened or loses its capacity for moral commitment, becoming debased or alienated. This leads to the deformity of personality development and the impoverishment of the spiritual world, running counter to the free and well-rounded development of the individual. On one hand, digital twins cause the self-personality to split between the virtual and the real; the coexistence of multiple personas leads to the fragmentation of personality and a rupture in self-reflection, making it easy for the same subject to exist in a state of severe severance between their online and offline personalities, finding it difficult to form a stable system of moral principles. On the other hand, in the process of the digital restructuring of the subject's personality as it "shifts from the real to the virtual," the disorderly expansion of virtual identity leads to the hollowing out of the real-world personality.
III. Paths for Resolving the Ethical Risks Derived from Generative AI-Empowered Ideological and Political Education
The empowerment of ideological and political education (IPE) practices by generative AI is a dialectical process in which positive efficacy and negative risks coexist. We should seek a scientific path for ethical governance based on the laws of IPE and the technical characteristics of AI. AI-empowered IPE must emphasize the subjective status of the human being; we must both promote technological development with a humanistic philosophy and integrate technology into educational practice with a responsible attitude. In the New Era, to prevent and resolve the ethical risks brought by generative AI to IPE, we must base our efforts on implementing the fundamental task of "fostering virtue through education" [12], adhere to the basic principle of "putting people first and using intelligence for good," and incorporate the technical rationality of AI into the framework of the laws governing IPE, ensuring that technical efficacy is released in an orderly manner according to the value rationality of education.
(1) Consolidating the Value-Guiding Foundation of IPE through Ideological Security Construction
The mainstream values of a society profoundly shape the ethical characteristics of the mainstream ideology, while the ethical traits of the mainstream ideology, in turn, deeply influence the state of mainstream values. IPE shoulders the arduous task of disseminating mainstream social values. In practice, we should use ideological security construction as a lever to compel the technical rationality of AI to serve the value rationality of IPE, "improving the capacity for online education and solidly performing school ideological, political, and ideological work in the Internet age." First, we must continuously enhance the leading power of the mainstream ideology to ensure that the value output of AI conforms to mainstream social values. We must uphold the Party's absolute leadership over ideological work in the education system, establish and improve corpora that align with our country's mainstream ideology and values—particularly by deeply embedding Marxist value concepts, the Socialist Core Values, and fine traditional Chinese culture into AI systems—and improve the capacity for value-sensitive design in algorithms. We must strengthen the effort to "clear the source and rectify the root" [13] of cyberspace, purify the soil for the spread of harmful ideas and erroneous trends of thought, and strengthen proactive data "feeding" and content pushes based on mainstream social values. This will allow generative AI to become a booster and multiplier for consolidating the mainstream ideological front during the process of empowering IPE. Simultaneously, relevant departments can explore the construction of intelligent platforms specifically for ideological and political work and embed them into other existing major Large Language Models (LLMs) in the form of public responsibility, creating a favorable platform environment for IPE. Second, we must explore the establishment of a full-process ideological supervision mechanism to ensure that every link of AI-empowered IPE meets the requirements of mainstream social values. The value characteristics of AI technology possess clear plasticity, and it is largely feasible to achieve alignment with specific values. Therefore, in the design process of AI agents, we must adhere to the principle of value alignment based on a holistic and systemic view, effectively embedding mainstream social value content and relevant value principles into every stage of AI, including data collection and processing, machine learning, algorithmic design, and content generation. We must also establish agile supervision, feedback, and correction mechanisms for the pre-event, mid-event, and post-event stages. Concurrently, we should establish an ethical review committee for intelligent educational technology specifically for IPE, formulate a "negative list" system covering links such as data collection, algorithmic application, and human-machine interaction, and regularly conduct ethical impact assessments and practical guidance for technical applications.
(2) Strengthening the Construction of a New Educational Mechanism Oriented Toward Innovation Capacity
Innovation capacity is increasingly becoming an important standard for measuring the value of talent. The high-level innovation capacity unique to humans is a vital aspect for judging the essential difference between humans and machines, while the infinite knowledge production capacity of AI stands in sharp contrast to its limited innovation capacity. In an era where AI technology is deeply reshaping IPE, it is necessary to establish a dialectical framework for technical cognition: we must use the instrumental rationality of intelligent technology to improve innovation efficiency, while remaining vigilant against the risk of the dissolution of subjectivity caused by over-reliance on technology. Therefore, traditional modes of innovation face the requirement for transformation under the impact of technology. We must strengthen the construction of new innovation capacities based on human-machine interaction. In particular, we must persist with an orientation toward the spirit of innovation, strengthen education on the Marxist view of innovation at the cognitive and practical levels, improve the corresponding innovation evaluation standard system, and use a series of corrective measures to rectify the phenomenon and risk of the degradation of the subject's innovation capacity caused by technology addiction. First, we must return to the fundamental standpoint of the Marxist critique of technology. While utilizing intelligent technology to enhance educational efficacy, we must protect cognitive autonomy through institutional design, embedding necessary reflective designs within technical systems to maintain the continuous reproduction of the subject's critical thinking and build a human-oriented digital-intelligent educational ecosystem. Second, we should carry out educational and training work based on the use of general AI technology to focus on improving the "intelligence literacy" (AI literacy) of the targets of education. AI literacy primarily consists of three parts: the awareness to choose and use AI, the ability to operate AI flexibly, and the responsibility to use AI according to norms. In practice, the goal of shaping a new spirit of innovation should run through the educational and training work of "IPE + AI." We should place responsibility ethics education in a prominent position and cultivate students' comprehensive ability to use intelligent technology for innovation, especially high-level innovation capacity assisted by intelligent tools. Third, we must strengthen the educational assessment and evaluation mechanism with innovation capacity as the core value orientation. As an important means for IPE to cultivate innovative talent, educational assessment and evaluation play a key guiding and motivating role in the educational process. The low-level innovation characteristics of AI determine the necessity of stimulating the subject's innovation vitality and raising the level of innovation. In reality, we should emphasize the weight of innovation in educational assessment and evaluation, setting key evaluation goals, indicators, methods, records, and feedback around innovative thinking and practice, thereby achieving a virtuous cycle between innovation capacity cultivation and assessment. Fourth, we must build a new cultural ecosystem that supports subject innovation through intelligent technology assistance, coagulate ethical centripetal force around innovation principles and norms, and emphasize the cultivation of the subject's innovative awareness, capacity, and responsibility. Against the background of an era where intelligent technology deeply empowers education, the innovation of educational mechanisms is shifting from a single focus on knowledge transmission to the cultivation of comprehensive quality centered on innovation capacity. Building a new cultural ecosystem with innovation at its core has become a key proposition for educational modernization. In practice, we should take the cultivation of the high-level innovation capacity of the targets of education as the core mission, accelerate the construction of an innovation culture based on human-machine collaborative innovation capacity, and promote the comprehensive leap of the subject's innovation literacy—including awareness, capacity, and responsibility—by reshaping innovative concepts, principles, and corresponding ethical norms, thus forming a broad atmosphere and fashion of innovation within the educational field and society at large.
(3) Improving the Humanistic Environment of Intelligent-Assisted IPE
Xi Jinping has pointed out: "Ideological and political work is, fundamentally, work concerning people. It must center on students, care for students, and serve students, continuously improving their ideological level, political awareness, moral character, and cultural literacy, so that students become well-rounded talents." IPE carries the core tasks of ideological and political work. In educational practice, it focuses on "the work of shaping the spiritual world and uplifting the morale and spirit of people," emphasizing the transformation of the human spiritual world and soul, thereby highlighting humanistic characteristics. In educational practice, we must adhere to humanistic value concepts that put people first, strengthen the humanistic care and human-oriented design of digitization and intelligence, and reasonably integrate the instrumental rationality of intelligent technology with the value rationality of IPE. This ensures that in the process of empowering IPE, generative AI possesses not only precision, breadth, and effectiveness, but also reflects warmth, emotion, and sentiment. First, we should explore effective paths for embedding "affective intelligence" based on humanistic values into intelligent technology, focusing on improving the comprehensive technical capacity for affective computing, perception, recognition, simulation, and expression. We should emphasize the comprehensive use of cross-modal and multi-media means—combining virtuality and reality, human-machine complementarity, scene immersion, the integration of emotion and reason, discourse innovation, and the unity of knowledge and action—to continuously improve the affective interaction function based on efficient intelligent empathy between humans and machines, thereby realizing the humanistic value and care of intelligent IPE. According to the laws of fostering people in IPE, we should integrate technologies such as Virtual Reality (VR), Augmented Reality (AR), Mixed Reality (MR), and Extended Reality (XR) into intelligent design to create panoramic reproductions and educational narrative scenes where scenes and emotions blend. For example, by constructing a humanistic care mechanism for intelligent education, technical designs for multi-modal emotional recognition and regulation can be effectively embedded into the entire process of human-machine interaction to prevent the emotional alienation caused by the subject's excessive reliance on technology. Second, we must explore healthy human-machine interaction modes within IPE scenarios to ensure the subjective status of both educators and students. Generative AI empowering IPE should be a beneficial human-machine collaborative educational mode under human leadership, rather than an educational mode where interpersonal relationships are alienated due to the "overstepping" of the machine. In practice, we must actively advocate for the return of human subjectivity, give full play to the role of teachers as the main force in humanistic and emotional education, and construct a new healthy interactive mode based on the trinity of "teacher-machine-student." This allows the human warmth of education and teachers to organically fuse with the work efficiency of the machine, using human wisdom and emotion to lead intelligent technology in better serving the cause of IPE.
(4) Establishing and Improving the Moral Responsibility Education Mechanism Based on Intelligent Scenarios
The phenomenon of moral anomie triggered by the deep embedding of artificial Intelligence (AI) technology into ideological and political education is, in essence, a crisis of humanistic values caused by the encroachment of technical rationality upon value rationality. Establishing a moral responsibility education mechanism compatible with intelligent scenarios has become an effective path to solving this difficult problem. First, we must refine the moral responsibility education system to adapt to the era of digital intelligence. This involves implementing AI literacy education projects and constructing a moral responsibility cultivation model that conforms to the characteristics of the digital intelligence era. Specialized courses on AI responsibility ethics should be established to cultivate and strengthen students' awareness of responsibility ethics when utilizing AI. Digital twin technology should be employed to conduct "mirror training" [14]; through comparative analysis of moral decision-making simulations in virtual scenarios and feedback from real-world behavior, the identity-cognition of virtual and real responsibilities can be reinforced. Simultaneously, human-machine collaborative decision-making training should be enhanced, designing intelligent interactive scenarios characterized by moral dilemmas to cultivate students' autonomous responsibility-control capabilities under technological assistance. Relying on cutting-edge technologies such as Virtual Reality (VR) and Augmented Reality (AR), we should develop immersive and high-fidelity ideological and political education teaching systems based on multimodal technology. This allows the educated to experience moral dilemmas in realistic intelligent interactive environments, continuously strengthening the subject’s moral judgment and sense of responsibility through repeated practice.
Second, we must construct a distributed responsibility network system. On one hand, an algorithmic responsibility traceability system should be established, embedding ethical impact assessment modules into the algorithm development stage to achieve full-process responsibility tagging from code writing to educational application. On the other hand, a multi-subject responsibility-sharing mechanism should be constructed and the moral responsibility assessment system perfected. This distributed responsibility network system should clarify the responsibility boundaries of developers, educators, managers, and learners. Accordingly, multi-dimensional assessment indicators—including technical ethics, educational ethics, and academic ethics—should be developed, with the assessment results incorporated into the education quality monitoring system.
Third, we must establish a new type of integrity education mechanism based on intelligent scenarios. We should give full play to the leading role of ideological and political education to effectively alleviate the integrity crisis through ethical education, institutional norms, supervisory means, and cultural construction. This involves cultivating students' awareness of integrity and sense of responsibility in using AI technology, while strengthening their moral judgment and capacity for self-discipline. We must improve scientific research normative mechanisms, clarifying the ethical boundaries of AI in knowledge production and application, and establishing targeted punishment mechanisms for academic misconduct. Furthermore, we should enhance the capacity to use technology to counter the "evils" of intelligent technology by developing and promoting the application of new academic misconduct detection technologies targeted at AI. Finally, the construction of an "integrity culture" within intelligent education should be strengthened to create a new trend of integrity and ethical education.
(The author is an Associate Professor at the School of Marxism, Beijing University of Technology) Source: Ideological Education Research (Sīxiǎng Jiàoyù Yánjiū) Issue 6, 2025 Web Editor: Ma Jingren