Anthropocentrism and learnification: resisting the dominant discourse of human-centred Artificial Intelligence in Education

Synthesis: Wang and Jin argue that the near-axiomatic insistence that AI in education be human-centered is not a neutral ethical baseline but a discourse with costs. Reading policy documents from UNESCO, the OECD, and the Montréal Declaration alongside the post-humanist critique of anthropocentrism and Gert Biesta's account of 'learnification', they contend that human-centered AIED substitutes an instrumental human–machine relationship for an educational one. In their reading, the demand that students learn about AI, and the learner-centered claim that education should facilitate learning through AI, together marginalize the experience of resistance through which, on Biesta's account, a student becomes a subject. The paper's contribution is a critical vocabulary and a call for prudence, not a body of evidence: it is a philosophical argument about the framing of personalized learning, and its claims are conceptual rather than empirical.

Key Findings

1. The human-centered orientation of AI ethics is presented by the authors as a documented policy consensus, not a contested position: the Montréal Declaration, the OECD's 2019 recommendation, and UNESCO's Beijing Consensus, policy guidance, ethics recommendation and generative AI guidance all frame AI as serving human users, and the paper reads UNESCO's own headings — how can AI be 'leveraged' to enhance education, how can it be 'exploited' for the common good — as encoding the human as end and AI as means.
2. The paper's first critique is philosophical: Enlightenment humanism posits a norm of what it means to be human and thereby excludes those who do not meet it, so education built on humanist foundations becomes socialization into an existing definition of the human. The authors extend this to anthropocentrism, arguing that human-centered AIED risks reaffirming human exceptionalism and obscuring the co-agential dynamics of educational environments.
3. Transhumanism is rejected as a remedy rather than a correction: on the authors' reading, it preserves rationality, autonomy, mastery and perfectibility while seeking to amplify them, so it retains an anthropocentric ontology and a purely instrumental conception of technology.
4. Learnification appears in AIED in two forms. The demand to learn about AI — illustrated by Mark Cuban's 'dinosaur in three years' warning and by ministerial pronouncements that workers must embrace AI — treats learning as a natural and intrinsically desirable process while actually prescribing one particular kind of learning, making those who decline or resist it responsible for their own marginalization.
5. The learner-centered version concerns learning through AI. Against the 'guide on the side' teacher, the authors follow Biesta in holding that learning is accidental to education while teaching is essential to it, and they use Biesta's robot vacuum cleaner comparison — a system that adapts 'autonomously' yet cannot selectively resist what it is supposed to learn — to argue that adaptation is not education.
6. Personalized, adaptive AI tutoring is charged with eroding the common learning experience that grounds shared understanding, and with removing the experience of resistance — 'difficult and frustrating' by Biesta's account — through which the child encounters a world that does not simply mirror the ego.
7. A political-economy layer runs alongside the philosophical one: citing Komljenovic, Williamson and Knox, the authors argue that human-centredness can function as a narrative legitimizing venture capital, Big Tech infrastructure, and state strategies for technological supremacy, with schools left holding the burden of ethical and regulatory compliance.

The human-centered consensus and its philosophical root

The paper's first move is descriptive. Drawing on Asimov's three laws of robotics as an early fictional statement of the same logic, it shows that user-centredness has been promoted from an axiom of AI ethics to an axiom of AIED. The Montréal Declaration's responsibility principle, the OECD's definition of AI partly by its human-centredness, and UNESCO's repeated emphasis on empowerment and enhancement all, in the authors' reading, position the student as a user who manipulates a tool. The authors note that 'empowerment' and 'enhancement' are left undefined in these documents, and that human-centered systems are designed to resist prompts judged anti-human — while the standards by which a prompt counts as human are themselves left unexamined.

The philosophical root is Enlightenment humanism's rational, autonomous subject, whose emancipatory rhetoric the authors (following Simon) treat as complicit with colonialism, patriarchy and capitalism. Biesta's critique is quoted directly: humanism posits a norm of humanness and excludes those who cannot live up to it. Bayne's point that humanist epistemology privileges detachment and control carries this into Pedagogies and Teaching Strategies, where universal rationality produces homogenizing models of the ideal student.

Post-humanism, and why transhumanism does not correct anthropocentrism

The paper separates two responses to humanism's limits. Transhumanism, it argues, keeps the core humanist values and pursues them technologically; technology becomes a means to make humans 'more human', which presupposes a definition of the human that the paper regards as unavailable and undesirable. Post-humanism instead decentres the human subject: Haraway's cyborg, Calarco's challenge to the human–animal divide, and Latour's Actor-Network Theory of human and non-human actors jointly dissolve the boundary that positions humans as sole agents of learning and meaning-making. Taylor's 'edu-crafting' is offered as the corresponding research practice, treating AI as a participant in the co-creation of educational realities rather than a tool serving human ends.

Learnification in two registers: learning about AI and learning through AI

Biesta's learnification thesis supplies the paper's second critical vocabulary. Because learning is treated as natural and inherently good, a particular kind of learning can be prescribed as though it were learning as such; the authors trace this in public exhortations that those who fail to learn about AI will be 'dinosaurs'. The implication they draw is that failure to learn AI literacy becomes a personal fault, and that legitimate reasons for resisting AI in some educational settings — or for keeping AI literacy out of young children's curricula — become hard to voice.

The learner-centered register targets personalized learning, Adaptive Learning and AI tutoring directly. If education's task is to facilitate, guide and support learning, then ambiguity about what counts as enhancement is papered over; 'enhancing the learning experience' functions as a general phrase for making learning 'better', and the authors ask whether a language of pleasure and comfort belongs to a language of education at all. Quoting Pring, they argue that individualization purchases personalization at the cost of the common learning experience on which shared understanding depends. Learner Agency and Learner Identity are at stake: on this account, an adaptive tutor that is adjusted to the student and does not reveal its own subject-ness cannot resist the student's egological world-view. The authors are explicit that this is an argument about the dominant discourse's limits, not a claim that AI can never serve educational purposes — they concede it might even challenge anthropocentrism.

What this means for practice

• Instructors. Treat the human-centered framing of a tool as a claim to interrogate, not a guarantee of educational value; ask what the tool adapts to, and what it makes it harder for a student to refuse.
• Keep something to resist. Design at least some educational encounters that are not adjusted to the individual learner, so that encountering a world that does not mirror the learner's will remains part of the experience of education.
• Curriculum designers and institutions. Personalization is not self-justifying; weigh individualized pathways against the shared learning experience that makes common understanding and public discussion possible.
• Policymakers and procurement. Note that 'human-centered' in AIED policy documents often means user-centered and learner-centered, quietly shifting the aim of education from teaching to the efficient production of learning — and recognize that the human-centered story is also a market story, in which the authors describe schools as bearing the compliance burden and, at times, as first test subjects for corporate products.
• Researchers. Read AIED policy documents as objects of analysis: the vocabulary of leverage, exploitation and enhancement does normative work that a critique of implementation effectiveness will not surface.

Limitations

• This is a conceptual and critical argument paper, not an empirical study: it reports no dataset, no participants and no measured outcomes, and its claims are philosophical positions rather than findings.
• The analysis reads a purposive selection of high-level policy documents and scholarly sources; the authors do not systematically sample AIED policy or demonstrate how local implementations align with or diverge from those documents, and they explicitly leave that policy-to-practice gap for other work.
• The paper's treatment of transhumanism and post-humanism is a compressed reading of complex literatures, and the authors themselves flag that disputes over the human/non-human boundary are set aside rather than resolved.
• The critique of learner-centered education extends beyond AI: the authors state that it applies to learner-centered teaching by humans as well, so the paper is not a test of AI's specific effects.
• The authors frame their conclusions as provisional and call for prudence rather than for any determinate policy; the argument does not specify which educational aims AI should serve instead.
• No claim here about how students in real classrooms experience AI tutoring should be attributed to this paper.

Citation

Wang, S., & Jin, H. (2026). Anthropocentrism and learnification: resisting the dominant discourse of human-centred Artificial Intelligence in Education. Journal of Philosophy of Education.