Obsolete Schooling and the Courageous Minority: Rethinking Educational Change in the Age of Artificial Intelligence

Introduction

Calls for fundamental change in schooling are not new. For decades, educators, policymakers, and scholars have argued that schools must be transformed to respond to changing social, economic, and technological conditions. Yet despite repeated reform efforts, policy waves, and pedagogical fashions, the basic structure of schooling has remained remarkably stable (Zhao, 2025a). In many parts of the world, schools continue to reproduce what Tyack and Cuban (1995) called the “grammar of schooling”: age-graded classrooms, standardized curricula, teacher-directed instruction, and assessment systems organized around uniform measures of performance (Tyack & Tobin, 1994). Reform has been constant; structural transformation has been rare.

Technology has long been expected to disrupt this pattern. From film and radio to television, computers, and the Internet, each major technological innovation has been accompanied by predictions that schooling would finally be transformed. Yet these predictions have repeatedly gone unfulfilled. As Cuban (1993) argued more than three decades ago, new technologies were largely absorbed into existing classroom routines and institutional logics rather than used to alter the underlying organization of schooling. More than 30 years later, that assessment remains largely valid (Zhao, 2025a; Zhao et al., 2015). Indeed, one of the great ironies of recent educational history is that governments and school systems invested enormous resources to connect schools and students to the Internet, treating digital access as essential to modernization and educational opportunity. Today, many of these same systems are investing substantial effort and policy attention in restricting or removing students’ access to smart devices connected to that same Internet (Ginsberg & Zhao, 2023). This reversal is telling. It reflects not only dissatisfaction with specific technologies, but also a deeper uncertainty about whether schools know how to make productive use of them.

The rapid development of generative artificial intelligence (AI) has brought this uncertainty into even sharper focus. Because AI can now perform many tasks traditionally assigned to students and teachers, it appears to offer yet another opportunity for educational transformation. But the real question is: Can schools actually change?

So far, the answer appears uncertain at best. Much of the emerging discussion of generative AI in education has focused on its risks: cognitive offloading, weakened student agency, academic dishonesty, shallow engagement, and new forms of technological dependence. These concerns have emerged alongside broader anxieties about digital technology, especially around screen time, distraction, and student well-being. In response, a growing number of schools and policymakers have moved toward restricting smart devices and limiting AI use in educational settings. Such responses are understandable, but also revealing. They suggest not merely concern about the dangers of a new technology, but also a broader loss of confidence in schools’ ability to incorporate powerful digital tools without fundamentally rethinking what counts as worthwhile learning, how learning should occur, and how it should be assessed in an AI-saturated world.

This article argues that many of these concerns are real, but often misdiagnosed. The negative outcomes associated with AI are frequently treated as evidence that the technology itself is inherently harmful to education. A more plausible interpretation is that AI is exposing the growing obsolescence of traditional schooling. When students use AI to complete routine assignments, bypass mechanical exercises, or generate acceptable performances with minimal intellectual engagement, the problem is not simply that they have access to a powerful tool. The problem is that schools continue to value and reward work that has become increasingly automatable. AI did not create this obsolescence. It revealed it. What appears to be a technological threat is, in important respects, a challenge to an educational model still organized around routine production, standardized performance, and narrow proxies for learning (Tyack & Tobin, 1994; Zhao, 2024; Zhao, 2025b).

If this diagnosis is correct, then the central question is not whether AI is good or bad for education. That question is too superficial to be analytically useful. The more important question is whether schools are willing to undertake transformational change at a historical moment when much of what they still ask students to do can now be done by intelligent machines. Addressing this challenge will require more than tighter regulation, more cautious implementation, or better guidance on tool use. It will require a more fundamental reconsideration of curriculum, pedagogy, learning environments, and assessment (Zhong & Zhao, 2025).

Yet transformational change in schooling is extraordinarily difficult. Traditional reform strategies, especially those seeking large-scale, top-down change across whole systems, have repeatedly failed to produce deep and lasting transformation. They often generate compliance, symbolic adjustment, or temporary enthusiasm, but rarely alter the underlying logic of schooling. If the age of AI demands not merely improvement but a new educational paradigm, then it also demands a different theory of change. In this article, I argue that such a theory must begin not with the expectation that the entire system will move at once, but with a courageous minority willing to create meaningful alternatives in the spaces they can actually control. Systemic transformation, I suggest, does not begin when everyone is ready. It begins when some are willing to build a future that the system has not yet authorized. This article contributes to current debates on AI and education by reframing the issue as one of educational obsolescence and by proposing a new theory of change centered on the courageous minority.

Evidence of Risk of Technology and AI

There is growing concern that technology can do harm to students. Horvath (2026) has argued in the U.S. Senate that children's cognitive development in many domains has “stalled” and in some areas “reversed,” alongside declines in literacy, numeracy, attention, and higher-order reasoning, and in public discussion surrounding that testimony he sharpened the warning by insisting that, in schools, the problem is not solved by changing the type of screen because screen-based technologies themselves can still undermine learning when they displace forms of teaching and interaction better aligned with human cognition.

Related concerns have been raised by others from different angles. Wolf (2018, 2025) argued that digital reading cultures can weaken the conditions for deep reading by eroding sustained attention, inference, reflection, and critical analysis. Haidt (2023) similarly contended that screen-saturated school environments undermine attention, relationships, and belonging, thereby weakening the social and cognitive foundations of learning. Twenge and Campbell (2018) added population-level evidence that heavier screen exposure is associated with lower psychological well-being among children and adolescents, including less curiosity, lower self-control, greater distractibility, and more difficulty finishing tasks. These scholars do not make identical claims, nor do they rely on the same evidence, but they converge on a common warning: increasingly screen-based forms of schooling may be reshaping attention, reading, well-being, and development in ways that should make educators cautious before assuming that more digital mediation necessarily produces better learning.

Recent evidence has raised serious concerns about AI's effects on students. Burns and Winthrop (2026) argued that AI poses substantial risks to student agency, deep learning, and emotional well-being if adopted uncritically, and framed the central educational task as helping students prosper, prepare, and be protected in an AI world rather than merely introducing new tools. Stanford Accelerator for Learning (2026) likewise concluded that the causal evidence on AI's effects remains thin in important areas, especially around emotional and social development, even as adoption has accelerated rapidly. Pew found that 12% of teens report using AI chatbots for emotional support or advice (Pew Research Center, 2026).

There is also growing concern about cognitive dependence and diminished independent performance. Stanford HAI (2026) reported that recent research suggests students may perform better while AI is available but lose the advantage when the tool is removed, and that some students begin to view the AI as more creative than themselves. A particularly influential recent study strengthens this concern. In Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task, Kosmyna et al. (2025) compared essay writing across groups using a large language model, a search engine, or no external tool. They found that participants in the large language model (LLM) condition showed lower neural connectivity during writing, weaker recall of what they had written, and a reduced sense of authorship over their essays relative to those writing without AI assistance, suggesting that convenience in AI-assisted writing may come with cognitive costs when learners offload too much of the composing process to the tool (Kosmyna et al., 2025).

Academic integrity concerns are equally salient. Pew Research Center (2026) found that 59% of U.S. teens think students at their school use AI chatbots to cheat at least somewhat often. The 2025 Higher Education Policy Institute Student Generative AI Survey reported that 92% of students had used AI in some form and 88% had used generative AI in assessment contexts, with common uses including explaining concepts, summarizing articles, and suggesting ideas for research and writing (Freeman, 2025; Pew Research Center, 2026).

Implications

This new approach to change has significant implications for everyone involved in schooling. For school leaders, the task is less to impose one grand reform on everyone at once than to create and protect spaces in which more relevant forms of learning can emerge. Leaders must identify where experimentation is possible, buffer it from premature standardization, and give legitimacy to work that may not yet look like conventional schooling (Grissom et al., 2021; Leithwood et al., 2004). In practice, this may mean creating a school-within-a-school pathway, protecting a grade level or interdisciplinary team to pilot new learning designs, allocating common planning time for inquiry-based work, or allowing selected teachers to use portfolios, exhibitions, and oral defenses alongside or in place of conventional tests. It may also mean publicly supporting teachers when parents or colleagues question why students are spending time interviewing community members, building prototypes, or revising long-term projects instead of completing more familiar worksheets and test-preparation tasks. Leadership in this sense is not simply managerial. It is ecological and political: leaders create the conditions under which new forms of education by the courageous minority can survive long enough to demonstrate their value.

For teachers, the implication is not diminished importance but transformed importance. Teachers matter less as distributors of information or producers of routine instructional materials, and more as designers of meaningful learning, cultivators of judgment, and mentors of student growth. Research on rigorous project-based learning suggests that intellectually demanding, authentic project designs can produce stronger learning outcomes than traditional instruction when tied to core disciplinary ideas. This means, for example, that a teacher might replace a standard persuasive essay with a project in which students identify a local issue, gather data, interview stakeholders, draft multiple proposals, receive critique, and present recommendations to a real audience. A mathematics teacher might ask students to analyze actual transportation patterns around the school and propose safer traffic flow solutions rather than simply solving decontextualized rate problems. A language arts teacher might organize a community oral history project in which students learn interviewing, transcription, interpretation, and narrative writing while preserving local memory. In each case, the teacher's role shifts from assigning uniform tasks to designing experiences in which students must use knowledge purposefully, make decisions, revise ideas, and create value for others.

For students, the implication is that schools should increasingly invite them to become active authors of learning rather than passive completers of assignments. Research on student voice and agency suggests that meaningful participation in decision-making can strengthen agency when it is substantive rather than tokenistic (Moore et al., 2022). This means students should have more opportunities to choose topics, frame questions, determine methods, and shape the final form of their work. For instance, instead of every student completing the same science fair project, students might identify issues they notice in their own lives or communities—air quality, food waste, loneliness among older adults, accessibility of public spaces—and decide how best to investigate them. Schools should also make more time available for students to explore interests, revise work, and collaborate across differences. In such settings, students are not merely being given “choice” within a predesigned assignment; they are being asked to take responsibility for identifying worthwhile directions, contributing different strengths, and working with others toward meaningful outcomes. This also helps shift success from individual competition toward human interdependence, where students learn that one person may excel at research, another at visual communication, another at organizing teams, and another at building community relationships.

For community partners, the implication is that schools must become more porous. Research on community schools and school–community partnerships suggests that well-implemented partnerships can improve attendance, school climate, belonging, and academic outcomes, especially when they are integrated into the core educational design rather than treated as add-ons (Maier et al., 2017; Swain et al., 2025). This means community partners should not be limited to guest speakers or field trip hosts. They can become co-designers of learning. A local clinic might work with students to develop multilingual health-information materials. A city planning office might provide real data for students to analyze neighborhood safety or walkability. A museum or historical society might collaborate on student-curated exhibits. A nonprofit organization might invite students to help investigate community needs and develop practical responses. These partnerships make student work more consequential because they provide authentic audiences, constraints, needs, and feedback that a school alone often cannot provide. In the age of AI, this matters even more: The easier it becomes for students to generate school-like products with digital tools, the more important it becomes for learning to be accountable to realities beyond school.

For policymakers and system leaders, the implication is not that large-scale policy no longer matters, but that it should shift from prescribing uniform solutions to creating conditions in which meaningful local transformation can emerge and spread. If system leaders continue to define success mainly through standardized testing, narrow accountability metrics, rigid compliance structures, and one-size-fits-all reform mandates, they will reinforce the very model of schooling that AI has made increasingly obsolete. Their role should instead be to remove unnecessary barriers, create flexibility for protected experimentation, legitimize alternative forms of assessment, and support schools in developing different pathways rather than forcing all schools into the same model of change. For example, state or district leaders can authorize pilot programs, allow schools to use portfolios, exhibitions, or performance assessments in place of some conventional measures, redesign accountability systems to value broader outcomes, and provide resources for networks of schools to learn from one another. They can also help by reducing the policy fear that drives AI use underground, replacing vague prohibition with clearer principles about ethical use, disclosure, judgment, and student development. In this sense, policymakers and system leaders should not try to engineer transformation entirely from the top down. Their more productive role is to create enabling conditions for courageous minorities to act, for small innovations to survive, and for promising local changes to influence the wider system over time.

These implications suggest that the change in the age of AI is not mainly about adding a new technology or enforcing new rules. It is about reorganizing relationships, time, tasks, and purposes so that learning becomes more personalized, more problem-centered, and more socially meaningful. Leaders protect the conditions for change, teachers design new forms of learning, students become active participants in shaping that learning, and community partners help connect schoolwork to real human needs. However, it is important to note that students and community leaders can also initiate transformational changes. Although the courageous minority is never everyone at once, it can include students, educators, leaders, parents, and community members whose actions are inspired and protected (Zhao, 2025a).

Conclusion

Recent studies and policy movements have made it increasingly difficult to ignore the risks of AI and smart devices for student learning, cognition, integrity, and well-being. But these risks are not surprising. They are symptoms of a deeper educational problem. AI has exposed the obsolescence of a school model that still teaches and assesses too much routine, externally defined work—the very kinds of work AI now performs well. As long as schools continue to reward those performances, students will keep using AI in ways schools call cheating, and restrictive policies will continue to drive much of that use underground rather than eliminate it (Freeman, 2025; Fu et al., 2026).

If schools want to remain relevant, they must change. Yet schools are difficult to change because they are stabilized by enduring institutional forms and wider social expectations—and because they function as a peace treaty among stakeholders whose interests do not fully align. That is why meaningful transformation is unlikely to begin through whole-system reform. It is more likely to begin with a courageous minority willing to redesign learning in the spaces they actually control. In the age of AI, the future of education may depend less on whether schools adopt the newest tools than on whether enough courageous people are willing to make schooling worth doing again.

An Update

Since developing the courageous minority approach as a theory of change, we have begun putting it into action through a global network of about 20 schools from different countries. Educators are joining meetings from around the world—some late at night, early in the morning, or while fighting jet lag—to design small, protected experiments within their own schools. These efforts include learner agency days, capstone projects, AI-supported inquiry, student podcasts, classroom-based problem finding, innovation groups, and “schools within schools.” Together, they show that educational transformation does not have to wait for whole-system reform. It can begin with courageous minorities who act within their own zones of possibility and generate evidence for broader change.