AI and the global classroom: governance, judgment, and the future of learning
As AI reshapes cognition, pedagogy, and societal structures, global leaders must confront the ethical, strategic, and diplomatic implications of human-machine collaboration in education and beyond
Photo: Prof. Alexander M. “Sasha” Sidorkin, California State University, Sacramento
Editor’s Note
As artificial intelligence transforms classrooms across the globe, its impact reaches far beyond teaching methods. In this feature, Alexander Sidorkin, Professor of Graduate and Professional Studies in Education and former Chief AI Officer and Director at California State University Sacramento, explores the profound ways AI is reshaping human cognition, pedagogy, and societal structures. Drawing on his experience in education and AI policy, Sidorkin examines the ethical, strategic, and diplomatic dimensions of human-machine collaboration in learning, highlighting both the risks and opportunities that emerge.
From national competency to cultural relevance, his insights illuminate how AI challenges traditional notions of expertise, equity, assessment, and governance. This conversation is part of Techplomacy Magazine’s special series, The Cognitive Frontier, which investigates how AI is redefining human potential, critical thinking, and the future of education.
Sidorkin’s perspectives offer policymakers, educators, and industry leaders a nuanced guide for navigating a rapidly evolving educational landscape where technology and human judgment intersect.
Olin Thakur, Editor-in-Chief, Techplomacy Magazine
Redefining expertise in the AI era
Expertise in an AI age cannot remain centered on procedural mastery that machines now execute flawlessly. We need to shift toward what I call Extended Executive Cognition: the ability to orchestrate cognitive resources across human and artificial agents. This involves developing strategic judgment about task decomposition, knowing what to delegate to AI and what requires human insight, and maintaining metacognitive awareness about when machine outputs need human correction.
For countries with limited technological infrastructure, this creates both challenge and opportunity. The challenge is obvious: access gaps threaten to widen global inequities. The opportunity is more subtle. Nations can leapfrog traditional educational sequences that consumed decades in developed countries.
Just as many African nations bypassed landline telephone infrastructure and moved directly to mobile networks, educational systems can skip the long march through procedural automaticity and move directly to teaching strategic thinking with AI assistance. Supplementing limited supply of highly qualified teachers with AI-powered tutor bots can also boost educational attainment in developing countries.
This requires reconceptualizing what schools deliver. Instead of spending years drilling multiplication tables or grammar rules that AI handles instantly, education should focus on developing discerning thinking: the ability to evaluate AI outputs for accuracy, relevance, and contextual appropriateness. Students need to recognize eloquent emptiness; the fluent but substantively hollow content AI produces so convincingly. They need a theory of alien mind to understand how AI systems process information in ways fundamentally different from human cognition.
The infrastructure question becomes not whether countries have universal high-speed internet, but whether students have sufficient access to practice human-AI collaboration under guided conditions. Even intermittent AI access, when paired with strong pedagogy, can develop the executive cognition that matters most for future work.
Extended executive cognition as a national competency
Extended Executive Cognition is more than an individual skill. It is a fundamental capability for navigating an AI-integrated world, and nations that develop it systematically will possess significant advantages. But framing it as a national competency requires careful thought.
The capability itself involves several interconnected dimensions. Strategic allocation: breaking complex projects into components and deciding what humans versus machines should handle. Cognitive load distribution: preventing human overload while maintaining efficiency across the extended human-AI system. Input specification design: crafting prompts and providing context that aligns AI processing with human goals. Discerning thinking: evaluating output quality and adding genuine value through human insight. These skills form the master capacity for orchestrating distributed cognitive systems.
Policy should prioritize this development through educational transformation rather than narrow vocational training. The goal is not producing prompt engineers but cultivating workers and citizens who can think strategically with technological partners.
Curricular reform must overcome what I call the curriculum curse, those years spent practicing prerequisite skills whose purpose remains opaque. AI enables engaging students with meaningful complexity before they master all foundations, using AI to bridge knowledge gaps just-in-time rather than requiring exhaustive prerequisites.
Diverse educational systems will develop extended executive cognition differently, reflecting local contexts and values. Some nations may emphasize collaborative approaches where students work in teams with AI, others may focus on individual mastery. What matters is ensuring all students develop sufficient executive cognition for meaningful participation in AI-integrated work and civic life. The inequality danger lies not in varied approaches but in populations receiving no systematic development of these capabilities while others do. Compute is expensive, and countries that invest heavily in expanding their compute capacity should plan to share that capacity with the rest of the world.
Equity and access in AI-assisted learning
The equity challenge in AI-enhanced education operates on multiple levels. The most visible is technological access: reliable internet, capable devices, and quality AI tools. Yet focusing only on infrastructure misses deeper concerns about how AI integration could entrench or disrupt existing inequalities.
First, we must acknowledge what I call the pedagogy of grace rather than merit-based thinking. Traditional education pretends students start from equal positions and rewards those who demonstrate the most effort. This fiction obscures massive differences in access to resources, preparation, and support. AI makes these inequities more visible. The student with extensive AI experience, sophisticated digital literacy, and home support leverages these tools far more effectively than the one encountering them only in under-resourced classrooms.
The disability dimension reveals the accommodation paradox most starkly. AI functions like eyeglasses or hearing aids, making functional barriers disappear rather than curing underlying conditions. A student with dyslexia can interrogate text through speech while AI converts spoken thoughts into polished prose. Students with ADHD receive organizational support without human judgment or impatience. A specific diagnosis does not matter. Gatekeeping to receive special support becomes indefensible when AI transforms accommodation from scarce resources requiring rationing to abundant support that costs nothing to scale.
The student who struggles with writing mechanics because of documented disability and the student who struggles because they attended under-resourced schools both need the same assistance. Why privilege one over the other? Traditional procedural barriers, handwriting requirements, spelling tests, calculation speed, served as exclusionary mechanisms disguised as academic standards. AI exposes this historical function by making these barriers obsolete, shifting competition from mechanical facility to actual thinking. When support becomes abundant rather than scarce, diagnostic gatekeeping loses its justification.
Governance frameworks must address several dimensions. Infrastructure equity requires ensuring baseline access, but not necessarily identical technology everywhere. Strategic deployment might provide robust AI access in educational institutions even when home connectivity remains limited. The key is structured time for students to develop executive cognition under guidance.
Pedagogical equity matters more than technological equity. Teachers need frameworks for teaching with AI that work across diverse contexts. This means practical guidance on designing learning activities that remain meaningful when AI can perform traditional academic tasks. It means assessment approaches that capture genuine learning rather than just polished outputs. It means understanding how to teach Extended Executive Cognition to students with varying backgrounds.
Content and cultural equity requires ensuring AI systems provide meaningful feedback in multilingual contexts without imposing Western pedagogical frameworks. Current AI models were trained predominantly on English content reflecting specific cultural assumptions. Governance must support development of locally relevant AI educational tools and approaches that honor diverse epistemologies.
The deepest equity challenge concerns what gets valued. If education continues prioritizing procedural tasks AI performs effortlessly, those with access gain automatic advantages. But if it shifts toward authentic complexity, contextual problem solving, and collaborative intelligence, then AI can level the playing field, allowing students from less privileged backgrounds to produce sophisticated work demonstrating genuine intellectual capability.
Human oversight versus automation in assessment
The automation question in assessment is one of the most consequential policy choices education faces. AI promises efficiency, consistency and cost savings, but the danger lies in reducing assessment to what machines can measure while missing what matters for learning.
Much current assessment serves what I call educational theater, performances of rigor that satisfy external audiences without genuinely measuring learning. Multiple-choice tests, five-paragraph essays, standardized formats work well for machine scoring because they reduce complex thinking to simple procedures. But education’s purpose is not producing outputs machines can grade efficiently.
Human judgment remains essential for evaluating capabilities that matter in an AI age. Extended Executive Cognition requires assessing how students orchestrate human and machine resources, make strategic choices about task allocation, and demonstrate metacognitive awareness. Discerning thinking demands assessing how students evaluate AI outputs critically, recognize eloquent emptiness, and add genuine value through human insight. Machines cannot assess their own limitations effectively.
Different cultures hold different expectations about what education should accomplish and how fairness manifests in assessment. Some emphasize individual achievement, others collective advancement. Some value innovation and risk-taking, others mastery of established knowledge. These cultural differences should inform assessment design rather than being erased by standardized algorithmic approaches.
The balance involves using AI strategically. Let it provide rapid feedback on procedural elements like grammar, citation format or computational accuracy, freeing human attention for higher-order capabilities evaluation. Let it surface patterns in student work for closer examination, but reserve judgment on intellectual sophistication, creative insight, and ethical reasoning for human evaluators who understand cultural context and educational purpose.
Policymakers should resist the efficiency seduction. Yes, AI can grade essays quickly. But education’s value lies not in processing student work efficiently but in developing human capabilities that resist automation. Assessment must capture those capabilities through approaches that prioritize meaningful evaluation over mechanical efficiency.
AI, pedagogy, and societal trust
When AI mediates learning and evaluation, it transforms the social contract between institutions and the public, threatening trust but also creating opportunities to rebuild confidence through transparency and authentic assessment.
Trust erosion appears in several ways. Parents and employers question whether credentials reflect genuine human capability or AI use. Students doubt their own competence when machines contribute significantly to their work. Faculty lose confidence in their ability to evaluate learning when traditional evidence becomes unreliable. This crisis deepens when institutions respond through detection and prohibition rather than meaningful integration.
An example of this erosion is a case at Texas A&M, where a professor failed an entire graduating class based on faulty AI detection. Students protested their innocence, but institutional trust in machines exceeded trust in students. This incident reveals the brittleness of assessment systems that depend on distinguishing humans from machine work rather than evaluating actual capability.
The path toward rebuilding trust requires several shifts. First, transparency about AI use rather than prohibition. When institutions acknowledge that AI has become part of intellectual work, they can focus on teaching effective collaboration rather than policing boundaries. Make AI integration explicit in course design, assignment structure, and assessment criteria. This honesty helps students, parents, and employers understand what capabilities are actually being developed.
Second, shift from product-focused to process-focused assessment. Instead of inferring learning from polished outputs that AI can generate, require documentation of thinking processes. Collaboration logs showing prompt iterations, decision-making rationales, and reflection on strategy effectiveness provide richer evidence of learning than final papers. This metacognitive transparency serves both assessment and pedagogical purposes.
Third, emphasize authentic complexity that grounds learning in specific contexts where machine outputs require human judgment. When students work with real community organizations, navigate actual ethical dilemmas, or solve problems with unique local constraints, their AI-assisted work demonstrates capabilities machines alone cannot provide. This contextual grounding makes assessment more meaningful and trust more justified.
Institutions gain trust not by pretending AI does not exist, but by developing robust frameworks for teaching, learning, and assessment in an AI-integrated world. The societies that navigate this transition successfully will be those that face the disruption honestly rather than defending obsolete practices.
Cognitive evolution versus cognitive loss
Framing AI-related changes in traditional skills as either evolution or loss reveals more about our ideological commitments than about actual cognitive development. Both framings contain truth, and policymakers must resist simplistic narratives in either direction.
What appears as loss often represents strategic reallocation of cognitive resources. I propose the External Automaticity Hypothesis: that just as internal automaticity (fluent execution of procedures through practice) frees cognitive capacity for higher-order thinking, external automaticity (fluent use of AI for procedural tasks) may achieve similar benefits. When navigation apps handle route planning, drivers do not lose spatial reasoning capacity; they redistribute cognitive effort toward safe vehicle control. Similarly, when AI handles citation formatting or computational procedures, students can focus on argument development or conceptual understanding.
Yet external automaticity carries risks. Not all cognitive processes can be externalized. Embodied skills, cultural intuitions, and interpersonal capabilities resist technological delegation. Students need sufficient foundational understanding to detect AI errors and inappropriate responses. The boundary between strategic delegation and problematic dependence remains genuinely unclear and likely varies across domains and individuals.
The adaptation framing helps when it focuses on what new capabilities emerge. Extended Executive Cognition represents a genuinely novel form of thinking that previous generations did not need. Students must learn to orchestrate cognitive resources across human and artificial agents, a metacognitive demand that may exceed traditional executive functions. Theory of alien mind for AI requires understanding how these systems process information through mechanisms utterly unlike human cognition, a form of cognitive flexibility that is genuinely new. Eloquent emptiness detection demands resisting processing fluency bias in contexts where machines produce impressive-sounding nonsense.
Policymakers should avoid both panic about cognitive decline and naive celebration of enhancement. Instead, invest in empirical research on actual cognitive development in AI-integrated contexts.
We need longitudinal studies tracking whether external automaticity produces lasting benefits or creates dependencies that emerge when technological support is removed. We need domain-specific investigations of which cognitive tasks benefit from internal versus external automaticity. We need developmental research on optimal timing for introducing AI assistance at different ages and competency levels.
Education policy should focus on capabilities that remain valuable regardless of technological change: adaptive problem solving in novel contexts, ethical reasoning in complex situations, creative synthesis of diverse perspectives, and collaborative intelligence that leverages both human and machine strengths. These are true adaptations, not mere reactions to AI disruption.
AI literacy as diplomatic leverage
The notion that AI-literate cognition could function as soft power deserves serious consideration, though not in the narrow technological sense that dominates current discourse. Nations will not gain diplomatic advantage primarily through producing more prompt engineers or AI technicians. The leverage comes from developing populations capable of sophisticated Extended Executive Cognition that shapes how societies integrate AI across all domains.
Nations with citizens and institutions skilled in human-AI collaboration can model integration that balances technology and human values. They can export educational frameworks, offering approaches to teaching, learning and assessment, that other countries can adapt. This pedagogical leadership represents real soft power.
Technical alliances increasingly depend on trust in how partners deploy AI systems. Nations with populations that critically evaluate AI outputs become attractive partners and can meaningfully shape AI governance. Global negotiations on standards, accountability, and data sharing benefit from broad AI literacy, not just elite expertise.
The diplomacy angle also involves cultural dimensions. Different societies approach human-AI collaboration through distinct epistemological frameworks. Some emphasize individual agency, others collective intelligence. Some prioritize efficiency, others context and relationship. These differences should inform AI development and deployment rather than being erased by Western-dominated technical standards. Nations that successfully integrate AI while maintaining cultural identity and values offer models of technological adoption that honor human diversity.
Regulatory foresight from educational AI
AI in education is a test case for broader governance challenges because education’s unique vulnerabilities expose risks that other sectors may face eventually. Lessons learned should inform regulation across sectors while respecting context-specific differences.
The erosion of assessment tools in education reveals a broader pattern: AI threatens the core functions of institutions when it can perform the very tasks these institutions rely on to measure capability. Education depended on essays, problem sets, and exams to evaluate learning. Once AI could produce these artifacts, the measurement system faltered.
This is not unique to schools. Professional credentialing relies on exams and practical demonstrations that AI may soon handle. Healthcare quality metrics depend on documentation and diagnostic accuracy that AI now assists. Legal practice involves research and drafting that AI can augment. Every domain faces its own version of the calibration problem: how to distinguish human contribution from machine input when both are legitimately present.
The lesson from detection failures extends beyond education. AI detection tools in schools proved catastrophically unreliable, producing false positives that disproportionately affected non-native speakers while missing actual AI use through simple evasion. Other sectors are experimenting with similar detection systems for AI-generated content, synthetic media, and automated decision-making. Experience in education suggests these approaches will fail in the same way. What works instead is transparency and collaboration between humans and AI, rather than trying to enforce segregation.
AI also amplifies existing inequalities. Students with more resources, preparation, and support leveraged AI far more effectively than those without. This pattern repeats elsewhere: in healthcare, digitally literate patients with access navigate AI-enhanced medicine more successfully; in legal systems, clients who can afford AI-assisted representation gain advantages; in labor markets, workers skilled at collaborating with AI tools command premium pay.
International coordination faces similar challenges. Education needs AI-assisted credentialing standards that travel across borders while respecting local values. Healthcare requires agreements on AI-assisted diagnosis and treatment that balance innovation with safety. Labor markets need policies for AI-enhanced work that prevent a race-to-the-bottom. Without such coordination, nations risk competing by lowering standards, approving questionable AI deployments, or failing to protect vulnerable populations.
Cultural context also matters. AI systems trained mainly on Western populations may make inappropriate recommendations for other genetic or cultural groups. Legal AI reflects the precedents of specific jurisdictions. Agricultural AI assumes conditions that may not hold globally. Regulators must ensure AI development includes diverse perspectives from the outset, rather than treating adaptability as an afterthought.
Most importantly, education shows that AI integration demands a fundamental rethinking of institutional purpose, not just superficial technological addition. Schools cannot simply add AI tools to existing practices; they must reconsider what learning means, how capabilities develop, and what outcomes truly matter. The same applies to healthcare, which must rethink diagnostics, treatment, and patient relationships, and to legal systems, which must reconsider adversarial procedures, precedent interpretation, and justice itself. Regulation should insist on this foundational rethinking rather than accepting shallow integration that preserves outdated assumptions.
Cultural context and AI feedback
Providing meaningful feedback across multilingual and multicultural contexts without imposing dominant frameworks is one of AI’s hardest governance challenges. Current AI systems, trained predominantly on Western educational content and assessment practices, risk becoming tools of epistemological colonization.
The challenge of AI feedback in education has three dimensions. First, linguistic complexity goes far beyond translation. Effective educational feedback must account for cultural communication norms, rhetorical styles, and pedagogical expectations. Some cultures favor indirect feedback to preserve face, others prefer direct critique. Some emphasize individual achievement, others collective growth. AI trained on Western directness may feel harsh or inappropriate in contexts that value subtlety.
Second, pedagogical diversity matters. Different societies have distinct learning traditions that deserve respect. Indigenous knowledge systems stress relational learning and community accountability. Confucian traditions value mastery through repetition. Progressive Western approaches prioritize creativity and critical thinking. African philosophies often integrate spiritual and communal dimensions. No single AI model can capture all these epistemologies.
Third, content relevance is critical. AI judging student work by generic standards can undervalue local knowledge, cultural practices, or community-centered perspectives. A history essay highlighting indigenous viewpoints may score lower despite its rigor, and scientific explanations blending traditional and Western knowledge may be flagged as confused even when sophisticated.
Governance must respond thoughtfully. AI systems should provide transparent documentation of training data and assumptions so educators and students can evaluate cultural biases. Development of locally relevant AI tools is essential, with training data reflecting regional pedagogies and values. AI feedback should be limited to procedural tasks like grammar, citation, and computation, leaving intellectual quality, argumentation, and creativity to human evaluators who understand context. Multilingual and multicultural assessment frameworks should recognize diverse forms of excellence, using AI for organization and pattern recognition but never replacing human judgment.
The goal is not to eliminate AI from educational feedback but to deploy it in ways that support local pedagogical wisdom. Achieving this requires ongoing collaboration between AI developers, educators, and communities to ensure AI serves learning, not standardization.
Preparing for cognitive interdependence
Education policy faces a fundamental challenge in preparing students for cognitive interdependence, where decision-making spreads across humans and machines, blurring traditional lines of responsibility. This is not a distant future, it is already emerging and demands immediate pedagogical action.
Moving from individual cognition to distributed intelligence is more than adding tools to existing practices. It requires what I call autonomous agency: the ability to direct human-AI collaboration while maintaining meaningful choice and purpose. This is at the heart of what I term Diacognitive Mode: thinking through technological extensions while keeping human judgment central.
Several capabilities are essential. Extended Executive Cognition serves as the master skill, orchestrating cognitive resources across human and artificial agents. It includes task decomposition, cognitive load distribution, input specification design, and dynamic attention allocation. Students must develop an intuitive sense of what to delegate to AI and what requires human oversight, constantly adjusting as both their skills and AI capabilities evolve.
Ethics of answerability becomes critical. Students collaborating with AI must take full responsibility for outcomes, verifying accuracy and ensuring ethical use. Excuses like “the AI made a mistake” are not acceptable. This standard mirrors professional realities where workers remain accountable for AI-assisted work.
Understanding AI itself is also necessary. Students need a theory of alien mind: how these systems process information through statistical patterns and probabilistic outputs, not human reasoning. Recognizing these differences helps students collaborate effectively while judging when to trust machine outputs.
Education policy should integrate these capabilities across the curriculum rather than treating them as optional. Courses from elementary through higher education must explicitly teach human-AI collaboration. Assessment should capture distributed cognitive processes, not just final products. Teachers need professional development to navigate cognitive interdependence.
Institutions must provide infrastructure for guided AI access, policies that encourage transparency, and frameworks recognizing responsible AI use as legitimate. Ethical reasoning about distributed decision-making is vital. When algorithms influence medical treatments, loan decisions, or criminal risk assessments, humans must exercise judgment, understanding when to accept, modify, or override machine recommendations.
Preparing students for cognitive interdependence means embracing AI’s transformation rather than resisting it. It means teaching them to think with and through machines while remaining fully human. It requires developing metacognitive sophistication to navigate a cognitive landscape no previous generation faced. Education that achieves this transformation secures its relevance in an AI-integrated world.
Alexander M. Sidorkin is Professor of Graduate and Professional Studies in Education at California State University Sacramento. He previously served as Dean of the College of Education and as Chief AI Officer and Director of the National Institute on AI in Society at the same institution. He provides consulting services to educational leaders and organizations at every stage of AI adoption, from initial assessment to seamless integration into instruction. He also advises startups developing AI-driven solutions for the education market.
The views expressed in this article are those of the interviewee and do not necessarily reflect the views of Techplomacy Magazine or the Techplomacy Foundation. Articles may be republished in full, without alteration, with credit to Techplomacy Magazine (techplomacyfoundation.org).
In Case You Missed It
AI Readiness and Partnership Priorities in Türkiye
Brazil’s Vision for AI Diplomacy: Sovereignty, Scale, and Safety in a Multipolar World
The state of enterprise AI report (OpenAI)
State of AI-Assisted Software Development report to dive deep into how AI is impacting technology-driven teams (Google)
Republishing: All our articles may be republished in their entirety, without alterations, to prevent misinterpretation or misuse. Please credit via Techplomacy Magazine (techplomacyfoundation.org).