Abstract
Purpose
This article reexamines Fukuyama's “End of History” thesis, arguing that artificial intelligence (AI) has reopened ideological competition and created an unprecedented “legitimacy gap” between exponentially advancing technology and linearly evolving governance. Bridging this gap has become pivotal to addressing many of the global crises we confront today.
Design/Approach/Methods
The study employs a problem–analysis–solution structure: identifying the legitimacy gap via a “Technology
Findings
The analysis identifies three AI-driven transformations: ideological resurgence, the reshaping of national capacity through compute, data, and intelligence as “political surplus,” and the recasting of individuals from the “last man” to the “predicted man.” Drawing on Latour's actor-network theory, the article proposes building trust networks through dual educational reconfiguration—cultivating AI literacy alongside civic literacy, enabling citizens to both collaborate with and govern intelligent systems.
Originality/Value
The article repositions education as the decisive mechanism for resolving the technology legitimacy crisis. Beyond novel frameworks such as “silicon-based economics” and “algorithmic constitutionalism,” it argues that dual-track educational reconfiguration is the only scalable pathway to rebuild trust networks for human–machine symbiotic governance.
Keywords
Introduction: Reexamining the “End of History” With the Artificial Intelligence (AI) Variable
From the launch of ChatGPT on November 30, 2022, to the present, more than 3 years have passed—this period marks not only an explosive breakthrough moment for AI as a transformative technology, but also a singularity in which AI becomes deeply embedded into society. Foundation models represented by ChatGPT and DeepSeek have spread rapidly worldwide; the cost of reasoning and generation with large models has been dropping every 3 months; and the embedding of intelligent systems across entire industrial chains has been accelerating at full speed.
In the United States, the White House in 2025 unveiled America's AI Action Plan, a comprehensive blueprint with more than ninety initiatives organized around three pillars: accelerating innovation, building domestic AI infrastructure, and maintaining leadership in diplomacy and national security (White House, United States, 2025). In August 2025, China's State Council issued the Guiding Opinions on Deepening the Implementation of the AI + Initiative, calling for the vigorous development of “AI-native technologies, products, and service systems,” the cultivation of AI-native enterprises, the exploration of new business models, and the creation of emerging intelligent ecosystems (State Council of the People's Republic of China, 2025). At the United Nations, the General Assembly passed its first global resolution on AI in March 2024, emphasizing the protection of personal data, the monitoring of AI risks, and the safeguarding of human rights (United Nations, 2024). Yet precisely at this juncture of global, societal, and mass-scale AI adoption, we are also witnessing a void of thought—a “legitimacy rupture.”
In The End of History and the Last Man, Francis Fukuyama proclaimed that the triumph of liberal democracy after the Cold War signaled the endpoint of mankind's ideological evolution—no viable alternatives remained to contest its legitimacy (Fukayama, 1992). In this sense, we had all become the “last men.” His thesis rested on two premises: First, the Cold War's end confirmed a systemic victor; second, the global expansion of technology and capital would universalize liberal democracy.
Three decades later, however, unprecedented breakthroughs in AI have not consolidated a firmer democratic consensus. Instead, they have opened new fissures of legitimacy. Platform algorithms, large language models, and autonomous agents now constitute an emergent “governance reality”—AI can already steer attention, rewrite narratives, and stands on the verge of autonomously executing tasks. These systems no longer embed themselves within the traditional human orders of power allocation, accountability, and ethical imagination.
What emerges from these transformations is not the reaffirmation of institutional victory, but the reopening of institutional alternatives. Against this backdrop, it becomes urgent to reassess Fukuyama's “end of history”: AI is not merely a technological breakthrough—it is actively challenging the very notion of the end of history. The endpoint has been reopened, and the intertwined trajectories of technology and governance will determine the path forward.
This dual-singularity moment of “technology and governance” is not simply about compute and models—it reveals an accelerating asynchrony between technology and institutions, education, and the humanities: While the technological curve rises exponentially, the governance curve advances only linearly. The wider this gap grows, the more acutely society feels the rupture of legitimacy.
At precisely this moment, more than ever before, we need new forms of explanation, ability, legitimacy, and governance to fill the void of thoughts, trust, and policy. This is the fundamental reason for the humanities and social sciences (HSS), and education to reconstitute themselves as the “new-quality infrastructure” of a human–machine symbiotic society in front of us all.
Three Singular Transformations Brought by AI: From Ideological Resurgence to the Reinvention of Human Condition
The leap of AI is not merely a technological advance; it is, as Žižek might put it, a singular event endowed with “monstrous power.” At a deep level, it unsettles the core paradigms of our political philosophy, state governance, and civic identity, while radically reshaping the structures of power, the logic of legitimacy, and the very ways in which humanity understands itself.
Ideological Resurgence: A New Narrative of Liberal Democracy Versus Techno-Authoritarianism
In The End of History and the Last Man, Francis Fukuyama offered a radical thesis: The end of the Cold War marked the conclusion of ideological struggle, and liberal democracy would, on a global scale, become the sole form of legitimate political and social organization. At that time, the collapse of the Soviet bloc, the triumph of Western capitalism, and the advance of globalization together provided a “historical crescendo” for the liberal narrative.
Yet in the era of AI, new technological variables are reversing this trajectory. With large language models, compute infrastructures, pervasive data monitoring, and social credit systems forming the backbone of social life and its governance, technology itself is becoming the foundational substrate of state governance. This highly integrated “algorithm-state” model offers two political advantages derived directly from technology:
Together, these constitute a new challenge to liberal democracy: the temptation of “technological legitimacy” as a substitute for “value-based legitimacy.” Faced with this condition, Fukuyama's thesis must be revised—History has not ended, it is reopening now; ideological competition has shifted its terrain. The struggle has moved from the axis of “ideals–institutions” to that of “technology–governance.”
We now see two divergent paths:
A “liberal–algorithmic” path, grounded in openness, transparency, privacy protection, and public participation.
An “authoritarian–algorithmic” path, characterized by centralized data collection, closed black-box decision models, and uniformed order prevailing over personal expression.
Which path a nation takes is a core challenge put in front of every nation. In this sense, AI is rapidly becoming the very infrastructure of new ideological competition.
Reshaping National Capacity: Compute, Data, and Intelligence as the New “Political Surplus”
In the context of a traditional economy, national capacity has typically been measured through fiscal revenues, military power, institutional resilience, and the ability to mobilize society. Yet in the silicon-based economy (Shao, 2025a, 2025b), national capacity is no longer determined merely by industrial output or capital accumulation. Instead, it is being redefined by three strategic resources of a new type: compute, data, and intelligence. Compute in the 21st century is like coal and steel in the 19th century; data today is equivalent to oil in the 20th century; and intelligence, in the decades ahead, will resemble electricity or money, serving as a foundational force circulating throughout society and underpinning governance. They are not merely factors of production, but the “hard core” that sustains national comparative advantage, international bargaining power, and domestic governance capacity (Figure1).
The three elements of the new national capacities.
First, compute—as the “core of digital infrastructure,”—is generating new gaps between nations. Supercomputing centers, cloud platforms, and AI-specific chips not only determine the speed of national research and innovation but also set the boundaries for industrial upgrading and security defense. Nations with compute surplus can front-run knowledge creation and decision-making in critical domains such as AI, life sciences, climate modeling, and financial regulation, thereby seizing the opportunity in institutional design and global agenda-setting. In this sense, compute surplus becomes a form of “political surplus”—analogous to today's trade surplus—delivering structural power advantages.
Second, data, as the “raw material of social cognition,” possesses high externalities and irreplaceability. Whoever commands larger volumes of high-quality, multimodal data can train stronger intelligent agents and secure advantages in both value creation and risk governance. Data does not exist in isolation; it is transformed into national capacity through institutional arrangements. Privacy protections, data sovereignty regimes, and rules governing cross-border flows all become ways for nations to assert legitimacy within the global order. Nations with data surplus can project their domestic models outward—through technical standards and international regulatory frameworks—turning them into global norms.
Third, intelligence—as the “automated output of cognitive labor”—is becoming a strategic surplus on par with compute and data. Intelligence means more than expanding model parameters or improving reasoning capacities; it signifies the ability of agents to substitute for humans in performing complex tasks in real-world scenarios. Nations that accumulate an “intelligence surplus” earlier—across education, research, industry, and governance—will hold greater power to shape institutions in an era of “human–machine cogovernance.” Unlike compute and data, intelligence embodies the applied transformation from algorithms to tasks to institutions; it is the totality of replicable capabilities that AI embeds within society.
Finally, the combination of compute, data, and intelligence shifts the axis of competition from “trade surplus” to a new type of “political surplus.” This surplus is not measured in the volume of goods, but in cognitive leadership, institutional spillover, and global narrative dominance. Under this logic, the reshaping of national capacity hinges on whether nations can incorporate compute allocation, data governance, and intelligence enablement into long-term strategies—building full-chain systems that extend from research to industry, from security to ethics.
Thus, future patterns of nation's competition will no longer be defined solely by GDP or military expenditure, but by Intelligent Domestic Product (Shao, 2025a, 2025b) and the “compute–data–intelligence balance sheet.” Those who achieve structural surpluses in these two dimensions will secure the power in the new global order and sustain influence in the form of a new type of “political surplus.”
Recasting the Individual Condition: The “Last Man” or the “Predicted Man”?
Fukuyama's notion of the “last man” depicts a depoliticized citizen at the supposed end of history—immersed in consumption, fearful of sacrifice, and withdrawn from public life. This condition feels strikingly familiar today: algorithm-driven entertainment, platformized short-video economies, and fragmented opinion ecosystems all encourage individuals to “lie flat,” weary of critical thought. The rise of AI intensifies this concern: Individuals risk not only becoming “last men” sheltered within algorithmic greenhouses, but also “predicted men”—transparent, controllable data objects whose futures are preprogrammed.
First, AI's pervasive capture of behavioral traces, consumption patterns, and emotional preferences renders individuals increasingly transparent. Predictive algorithms reduce the person to a data profile, and future choices appear prelocked by probabilistic models. Under this logic, the individual forfeits the space of “indeterminacy”; freedom is reshaped into “predictable variation.” Here emerges the “predicted man”: whose life is circumscribed by recommendation systems, algorithmic credit scores, and risk models—where even self-knowledge is subtly steered by feedback loops.
Second, this algorithmic shaping of subjectivity easily compounds the condition of the “last man.” When individuals surrender to passive acceptance, instant gratification, and algorithm-optimized convenience, the impulse to resist or transcend gradually dissolves. It is a gentle domestication: outwardly efficient and comfortable, yet inwardly eroding the very horizon of possibility. The haunting question of whether humanity will be caged by AI keeps resurfacing nowadays (Tegmark, 2017).
The challenge of human self-understanding has evolved from escaping the trap of the “last man” to breaking free from the shackles of the “predicted man.” If AI's cognitive variables can act as an antidote to the poison of modernity's decline, then the symbiosis of human and machine, in reopening the so-called end of history, must also reopen a higher realm of indeterminacy for the human condition. Within such a framework, the “last man” at the end of history will reclaim self-agency to explore new possibilities—and these possibilities will be coconstructed by humans and AI alike (Wu, 2024).
The Tension Space of the “Technology–Governance” Dual Curve: Constructing Trust in the AI Era
Since 2022, the world has entered an intelligence transition period driven by large language models. AI technologies have advanced rapidly—from autonomous content generation (generative AI), to the steering of attention (algorithmic order), to the capacity to organize collective action (agent economy). Yet parallel to it is the lag of governance mechanisms: the absence of institutional design and the vacuum of public ethics. The fastness of technology versus the slowness of governance is producing an ever-widening “legitimacy gap”: At its core lies the reconstruction of trust in the AI era (Figure 2).
Technology–governance dual curve and the legitimacy gap.
In today's global context—where nations are racing in AI and societies are widely embracing its applications—trust is the first institutional asset we lost, and the hardest to rebuild. We see this clearly: AI-assisted decision systems are poised to partially replace human judgment in critical fields such as law, medicine, and finance, but they often lack a closed-loop mechanism of explainability–auditability–remedy. Algorithmic bias, model hallucinations, and unclear lines of accountability will erode public confidence. In a reality we are evolving to where “the algorithmic platform becomes the state,” even the most advanced models cannot secure institutional trust without effective governance mechanisms.
A 2025 study by KPMG and the University of Melbourne, covering 47 countries, 48,000 respondents, found that while a majority of people remain optimistic about AI, only about 46% expressed willingness to trust AI systems. The public overwhelmingly believes AI requires regulatory oversight and governance (KPMG & University of Melbourne, 2025). Similarly, researchers at UC Berkeley, surveying thousands of large firms, concluded that the “level of trust”—together with factors like differentiation of data, strength of digital core, rate of learning, depth of capacity reinvention and strength of external partnership networks—constitutes the key to sustainable competitive advantage in the age of AI (University of California, Berkeley, Haas School of Business, 2024).
The world's leading AI companies are themselves shifting the battlefield—from a race over “who has the most advanced AI” to “whose AI can be trusted.” Google, for example, has built governance structures into its product release cycles, implementing risk taxonomies, prerelease evaluations, and mitigation measures. This embedding of responsibility and trust into the Research and Development and product pipeline—governance plus risk evaluation—is essential for earning the confidence of enterprise clients, regulators, and the public (Google, 2024). OpenAI has likewise conducted a global consultation with over a thousand participants—users and experts alike—to revise its Model Specification based on public feedback (OpenAI, 2025). Anthropic has centered its approach around Constitutional AI, explicitly encoding normative principles into training processes so that model behavior can be audited against a set of transparent rules, and publishing iterative updates to its “Constitution” to build public legitimacy (Anthropic, 2023).
“Can we trust an AI?”—This is perhaps the most fundamental institutional question of our time. Bruno Latour, in Science in Action, reminds us that science does not endure because it unveils some universal truth, but because it becomes stably embedded in an “actor-network” of humans, technologies, institutions, documents, and instruments. In other words, trust is not given a priori to any technology or institution; it is sustained through the collaborative construction and maintenance of networks of actors (Latour, 1987). For AI to be embedded into our social machinery, building a trust network supported by technologies, standards, documentation, institutions, and platforms is indispensable. Trust is not a matter of belief—it is a matter of network stability.
“Algorithmic Constitutionalism”: From New Standards of Legitimacy to Networks of Trust
In the context of AI's deep involvement in governance practices, the traditional foundations of legitimacy—such as democratic authorization through the ballot box, protection via property rights regimes, and the procedural justice of the rule of law—are undergoing a structural challenge. Algorithmic models have already penetrated critical domains such as credit approval, educational assessment, and medical triage. Yet their operational logics, training data, and judgment mechanisms often remain invisible and unaccountable “black box systems.” In this context, public trust in institutions can no longer rest solely on democratic voting or expert judgment; it must increasingly shift toward the supervision and constraint of algorithmic behavior itself (Figure 3).
Latour's actor-network of trust in science and AI and society.
This gives rise to a new “standard of legitimacy”—one that no longer only asks, “Who governs?” but goes further to ask, “How do algorithms govern? Are they fair, transparent, and correctable?” Its core elements include:
This shift does not imply that technology substitutes politics; rather, it means that algorithms must themselves be incorporated into governance structures and subjected to something akin to “constitutional review” to test their legitimacy—what we may call algorithmic constitutionalism. Drawing on Latour's analysis in Science in Action, we might say that, trust is not about whether technology is correct, but about whether technology is stably embedded in the social order, and whether it can continuously undergo correction and accountability.
Constructing institutional trust in the age of AI, therefore, requires not only better models, but also embedding them within an institutional network that is explainable, auditable, and reparable—so that technology becomes a node of cooperative governance rather than the isolated core of a new authoritarianism.
The Dual Return of the HSS, and Education
Triple Gaps and Triple Supply
In the rapid ascent of AI, a paradox has emerged: The exponential growth of technological capacity has not been accompanied by a corresponding upgrade in social order, value norms, and public governance. AI is reshaping the boundaries of knowledge, the structure of labor, and the modes of social collaboration. Yet our answers to fundamental questions—why we act, whether we should act, and how we assume responsibility after acting—remain vague, belated, and fragmented.
Against this backdrop, the HSS are undergoing a profound revaluation. Their legitimacy must be reclaimed not only within the academic system but also within institutional frameworks, where they are tasked with the critical functions of interpretation, norm-setting, and governance.
AI is a tool, yet it continuously generates institutional problems: algorithmic bias, model hallucination, privacy violations, attention manipulation, disinformation, labor displacement, and governance paralysis. The essence of the problem has never been AI itself, but rather who uses it, for what purposes, whether it can be controlled, and how accountability is ensured. These questions cannot be answered by technology, nor can they be resolved through business self-regulation. Ultimately, they lead us back to the reaffirmation of core values such as power, justice, responsibility, and community, which are precisely the fundamental questions of the HSS.
The accelerating pace of technology has opened three structural gaps: the interpretive gap, the normative gap, and the order gap. These gaps are the key fault lines where social systems fail to evolve in sync with technological systems, and they urgently require the supply of research from HSS.
The Interpretive Gap: Facts are Abound, Meaning is Scarce
In the AI-driven era, the processes of discovery–generation–retrieval have reached an unprecedented level of efficiency. Large language models and multimodal generative systems can not only extract structured information from vast oceans of data but also instantly generate coherent and seemingly realistic texts, images, sounds, and chains of reasoning. What once required years of scholarly accumulation or the coordinated production of specialized institutions can now be simulated, recombined, and disseminated in a matter of seconds. The barriers to producing and distributing knowledge have been dramatically lowered; facts appear to be within immediate reach, and the acquisition of information is no longer a scarce resource.
Yet accompanying this radical convenience is an equally radical erosion of meaning. Surrounded by layers upon layers of AI-generated outputs that are “realistic enough,” the boundary between truth and fabrication grows increasingly porous. Information expands without limit in quantity, but explanatory depth and normative orientation lag behind. This phenomenon constitutes the essence of the explanatory gap: a widening rift between the capacity to supply knowledge and the capacity to interpret it. We can access facts more quickly than ever, but our ability to contextualize, evaluate, and derive meaning from them is increasingly thinning.
This is not simply a problem of “content overload.” It is, more fundamentally, a precursor to the disintegration of cognitive consensus. In an era where virtually any narrative can be trained into a model and any public opinion can be synthesized at scale, facts no longer speak for themselves. They must be stabilized through interpretive frameworks that compete for legitimacy. Thus, the most basic questions resurface with renewed urgency: What counts as true? What deserves trust? And why should we believe? When societies can no longer maintain a minimal consensus on these questions, both the public sphere and institutional governance risk profound destabilization.
Thus, the role of the HSS becomes indispensable. Disciplines such as economics, political science, sociology, and cultural studies must not only analyze how AI reshapes labor relations, capital structures, sovereignty, and identity politics, but also provide the interpretive frameworks that make sense of such transformations. Economics, for instance, can help explain how intelligent productivity alters the balance between capital and labor. Political science can illuminate how data and compute capacity are becoming new foundations of state power and legitimacy. Sociology can track the reorganization of collective coordination and social stratification in the wake of algorithmic systems. Cultural studies, meanwhile, can interrogate how generative content transforms narrative authority and identity construction.
Absent such interpretive interventions, individuals and societies will be left adrift in an endless torrent of fragmented information. They will react passively to piecemeal news cycles and short-term opinion waves, rather than cultivating holistic understanding and foresight. In such a condition, governance becomes shortsighted and fragile, lacking the capacity to anticipate systemic consequences. Put differently: If explanatory capacity does not keep pace with the generative power of technology, society will be trapped in the paradox of having “facts at our fingertips” yet suffering from an acute “scarcity of meaning.”
The Normative Gap: What Models Can Do Versus What They Should Do
Algorithms, data, and models are permeating everyday life at an unprecedented pace. From credit scoring in finance to sentencing recommendations in criminal justice, from triage and diagnostics in healthcare to recruitment screening, advertising, and personalized recommendation systems, AI is no longer confined to the role of a mere tool. Increasingly, it is positioned as a quasi-decision-maker. Precisely at this juncture, societies must confront a fundamental normative question: Where should the boundary of algorithmic decision-making be drawn, and on what grounds can its legitimacy be established?
This is not a question that can be answered by scaling up parameters, optimizing architectures, or refining loss functions. Technical progress, however impressive, cannot substitute for value judgment. Issues of legitimacy belong to the domains of political theory, jurisprudence, and ethics. They involve whether rights are respected, whether justice is preserved, whether proportionality is observed, and whether accountability mechanisms exist. In other words, the question is not simply whether AI systems can make predictions with accuracy, but whether they do so within a framework that safeguards human dignity and institutional legitimacy.
Law and ethics have long provided principles for drawing such boundaries. Consider a few examples: In criminal law, if AI is used for sentencing prediction, does it respect the rights of the accused, or does it entrench existing biases embedded in training data? In education, if AI evaluates student potential, does it enable upward mobility or reinforce social stratification? In advertising, when recommendation systems capture and manipulate attention, do they cross the line into covert coercion, eroding individual autonomy? These are not technical problems to be solved with higher accuracy scores; they are normative dilemmas that strike at the heart of legitimacy, fairness, and human freedom.
At a deeper level, these dilemmas point to the indispensable role of the HSS. What is required is not merely better-performing algorithms, but a system of normative adjudication that establishes a dialogical and dynamic “gray zone” between efficiency and justice, between predictability and individual dignity. Such a framework cannot emerge from algorithms themselves; it must be consciously constructed, codified, and embedded into governance structures. In short, the boundaries of AI decision-making are not discovered by machines but assigned by societies.
The so-called “normative gap” is therefore not an abstract academic concern, but a structural void rapidly materializing in real life. If left unaddressed, AI risks advancing efficiency at the expense of justice and exploiting predictability at the cost of diminishing human agency. This is not only a matter of technological governance, but also a decisive question for the renewal of social trust and institutional legitimacy in the age of AI.
The Governance Gap of Social Order: How Can Technology Be Embedded Into Social Processes?
AI is no longer a laboratory demonstration but is steadily embedding in the very processes of governance and institutional decision-making. Yet the greatest shortfall we face today is the scarcity of capacity within social systems to integrate such a breakthrough technology into institutional routines in the right way. In other words, society is not prepared to treat AI as part of its public order. Even when people are aware that algorithms can be biased, few know how to lodge an effective appeal. Even when it is clear that models may fail, there is rarely a transparent and universal “review–redress” mechanism in place. This absence reveals the third structural deficiency: the order gap.
The order gap demonstrates that AI poses not only technical challenges but also challenges to social reproduction itself. Bridging this gap cannot be accomplished through more scientific papers or incremental technical advances; its true substitute is the educational system. In the tradition of classical educational sociology (Bernstein, 1975; Bourdieu & Passeron, 1977; Durkheim, 1956), education has always served not merely as a vehicle of knowledge transmission but as a mechanism of social order reproduction: By conveying norms, distributing recognition, and embedding institutional codes, education ensures continuity and legitimacy across generations. Today, as AI generates new modes of governance, education must once again take on the role of sustaining and correcting order. The sustainability of order in the future will depend not simply on what laws and policies declare, but on whether all members of society are equipped with the capacities of how to participate, how to challenge, and how to repair.
This point is crucial. Only education can scale, institutionalize, and normalize the abilities of audit, appeal, and correction across an entire population, creating systemic resilience. If such capacities remain confined to a narrow circle of experts, AI will remain a “black box,” and citizens will be left with no option but passive acceptance of technological imposition. But if these capacities are widely diffused through education, auditing, appeals, and corrective practices can become part of everyday social practice. Thus, the sustainability of order does not rest solely on the existence of institutional documents; it rests on whether education can transform corrective capacity into a universal social competence.
This also means that we must no longer equate education narrowly with “schooling.” In the age of AI, education must become a broad and continuous process of lifelong learning, civic education, and AI literacy. It should target not only students but all members of society, enabling them to navigate, supervise, and coshape the technological systems that increasingly govern their lives.
Accordingly, the future focus of education will not be on the transmission of knowledge alone, but on cultivating agents and fostering a new form of civic literacy for the AI era. Citizens must learn to ask: Is this system auditable? When confronted with an algorithmic judgment, how can I raise a challenge? How can communities organize to collectively monitor technological deployments? And how can local, everyday practices of “AI ethics” be developed at the workplace or within neighborhoods? The aim of this reform is not to add a few AI courses to existing curricula, but to embed resilience and accountability into society itself.
In other words, what is at stake is not a mere update to educational content, but a structural reconfiguration of the educational paradigm. It determines whether societies in the age of AI can sustain legitimacy, maintain trust, and construct a future order where human agency is preserved while collaboration with intelligent agents becomes both possible and productive.
Structural Challenges to the Educational Paradigm in the AI Era
Traditional education has historically evolved around three fundamental functions. The first is cultural transmission: Through curricula and classrooms, the accumulated knowledge, values, and symbolic systems of one generation are passed on to the next, thus sustaining historical continuity and identity. The second is capacity cultivation: Systematic curricula train basic skills such as logical reasoning, language expression, numerical ability, and collaborative competence, enabling individuals to find their roles within the division of labor. The third is social stratification: Examinations, credentials, and diplomas serve as mechanisms for selecting talent and distributing scarce social resources and career pathways. These three goals have largely presupposed a “slow-changing society,” in which social rules remain relatively stable, knowledge accumulates linearly, and the pace of technological progress broadly matches institutional adaptation.
The rise of AI, however, has thoroughly disrupted this balance. In particular, the spread of generative AI and large models has made knowledge no longer scarce. Traditional education operated under the logic of “scarce knowledge–long training–intergenerational transmission.” Today, however, models can instantly generate, combine, and even creatively extrapolate knowledge, allowing students to access within minutes what previously required years of study. Knowledge transfer is no longer the bottleneck; the new scarcity lies in higher-order capacities of definition, evaluation, and governance.
Similarly, examinations and diplomas are no longer the sole gateways to social mobility. In the contemporary labor market and circulation of talent, new evaluation standards are emerging: one's compute ownership (whether one can mobilize advanced computational tools), the richness of one's intelligent portfolio (the ability to creatively apply AI), and one's influence within the digital society (the capacity to construct reputation and networks online). Traditional educational outcome indicators are thus being eroded, while new systems of assessment are quietly taking shape. This cannot be solved by simply “adding an AI course” to the curriculum; it strikes at the structural functions of education itself.
We must therefore clearly recognize the current predicament of education as a structural reconfiguration, not a mere pedagogical upgrade. This requires returning to the foundational question of pedagogy: What is the ultimate goal of education? According to Bourdieu and Passeron (1977), education serves two ultimate goals: social reproduction and the distribution of capacities. In the AI era, these goals have not disappeared, but their meanings have fundamentally shifted. Social reproduction now depends less on the linear replication of cultural traditions and more on whether civic education can be rebuilt to uphold legitimacy and restore public trust. The distribution of capacities no longer centers on academic performance alone, but on cultivating agents who can both collaborate with AI and retain human autonomy. This transformation alters the very contract between education and society, rather than tweaking curriculum design at the margins.
History shows that moments of social upheaval—whether brought by war or by technology—are almost always accompanied by reinventions of civic education. Ancient Greece's democratic experiments required new forms of rhetorical and public education; the French Revolution's republican reconstruction required new national education to forge solidarity; and the post–World War II era relied on education to repair trust and rebuild legitimacy. Civic education has always been the linchpin of social trust and institutional legitimacy. The rise of AI reopens the end of history and places us at another turning point: the need to reinvent civic education. As algorithms, models, and intelligent agents penetrate domains such as justice, finance, healthcare, and education, questions of “who governs” and “whether governance is trustworthy” are pressing. In this context, education's mission is not only to impart technical skills but to equip citizens with the ability to audit algorithms, to raise challenges, and to press for institutional remedies. “How to become a citizen of the AI society” and “how to master AI skills and tools” are equally crucial—in my opinion, the former may be even more foundational.
The structural reconfiguration of education requires a dual reinvention of content and form. Education in the AI era cannot rely solely on traditional disciplinary structures; it must shift toward interdisciplinary, task-driven approaches. A new paradigm should revolve around two core capacities: (a) civic literacy for the AI society, including the ability to determine whether algorithms can be audited, to challenge AI decisions, to supervise deployments through collective action, and to establish localized “AI ethical practices” in everyday life and work; and (b) human–AI collaboration, including the ability to co-complete complex tasks with intelligent agents, to enhance creativity through AI, and to retain human leadership within collaboration. These capacities cannot be cultivated through mere pedagogical upgrades; they demand systemic rethinking and structural transformation of the educational institution itself.
The most difficult challenge lies in the fact that this is an institutional reconfiguration. If curriculum reforms and skills training can be considered “content updates,” then the deepest and most urgent task for education in the AI era is the redefinition of education's position and function within the social system. Education must simultaneously fulfill two roles: It must provide the mechanisms of social reproduction and capacity distribution for the AI society to sustain fast productivity growth and create a new order, while also embedding corrective mechanisms to fill the legitimacy gap and safeguard it.
In sum, the rise of AI does not herald a mere revolution in teaching methods, but a structural reconfiguration that extends into the social contract, institutional legitimacy, and the very order of human–machine coexistence. Education is no longer simply an intermediary for transmitting knowledge and allocating status; it must now shoulder the fundamental task of guiding societies through legitimacy crises, rebuilding public trust, and cultivating a new civic literacy for the AI era. This challenge runs far deeper than adding new courses or updating a handful of skills—it is a transformation at the very foundation of education itself.
Footnotes
Funding
The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Shanghai Municipal Education Commission's “AI-Driven Scientific Research Paradigm Reform for Disciplinary Advancement Program” (Grant No. 2024AI01005).
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.