A recent conceptual analysis published in the journal Human-Centric Intelligent Systems proposes a shift in educational philosophy to address the risks of intellectual dependence in the age of artificial intelligence. Researchers Hugo-Henrik Hachem, Franziska Primus, Katerina Linden, Fredrik Heintz, and Mattias Wiggberg argue that common approaches to AI literacy, which rely on the concepts of critical thinking and critical consciousness, may inadvertently reinforce the very hierarchies of knowledge they seek to dismantle. By framing students as naïve and teachers as authoritative critical thinkers, these educational models risk consecrating a reliance on human and AI-based instructors that inhibits genuine intellectual independence. The paper suggests that rather than teaching students to evaluate AI through a lens of criticality prescribed by experts, education should instead foster intellectual emancipation based on the principle that intelligence is universal and indivisible.
The study highlights a paradoxical triple bind wherein current AI literacy efforts aim to mitigate intellectual dependence on AI tools by subjecting students to an intellectual dependence on AI-savvy teachers. The authors contend that the definitions and empirical evaluations of critical thinking and critical consciousness in the context of AI literacy are often ambiguous or tautological, making it difficult to measure progress without relying on subjective intellectual hierarchies. When these educational goals are employed, they assume that a teacher must guide a learner from an uncritical to a critical state. The authors apply the educational philosophy of Jacques Rancière to argue that intellectual independence should not be treated as a future goal to be achieved through pedagogical guidance, but as a starting point and a condition of teaching and learning. This perspective shifts the focus from assessing levels of thought to verifying the learner’s will and commitment to engage with a specific body of knowledge.
To put this into practice, the researchers propose a thing-centred pedagogy, which centres the curriculum on a specific object of study, such as a book, a dataset, or a Massive Open Online Course, rather than on the learner’s perceived intellectual level. In this framework, the teacher adopts the role of an ignorant teacher, who does not seek to explain the material or guide the student to a pre-determined interpretation, but instead facilitates engagement and verifies the student’s invested effort and attention. This approach is intended to decentralize the teacher’s authority and place it within the thing itself. By treating the thing as an island of knowledge that students can navigate using their own intelligence, educators can model freedom and encourage students to find their own path to understanding without the crutch of redundant explications.
The paper also extends this philosophy to AI-mediated learning by introducing the concept of the ignorant AI-tutor. Contrary to recent initiatives that aim to program AI to act as a Socratic dialogue partner—a method the authors criticize for being inherently hierarchical, adversarial, and prone to guiding students toward instructor-favored conclusions—an ignorant AI-tutor is designed to ask open-ended, non-guiding questions directly linked to specific passages or locations within the study material. By leveraging an epistemology of ignorance, such a system avoids generating explanations or grading student responses, which often involve factual inaccuracies, hallucinations, or biased interpretations. Instead, the tutor prompts the learner to explore the text and report what they see, think, or make of it, thereby maintaining the learner’s intellectual autonomy and sustaining their curiosity through direct engagement with the material.
The implementation of an ignorant AI-tutor requires significant design considerations to ensure it remains a helpful tool rather than a directive authority. The study suggests that such systems should include a refusal mode for direct answers and interpretive summaries, effectively blocking attempts by the user to circumvent the learning process by asking for the correct answer. Instead, the AI should deflect and redirect the user back to the source material. Furthermore, to address the challenge of hallucinations, the authors argue that these should be treated as non-catastrophic and potentially generative opportunities for the student to verify the content and engage creatively. By making the source material searchable and the references traceable, learners can play an active role in verifying the AI’s prompts, turning potentially problematic interactions into active exercises in critical verification.
The authors acknowledge that their proposed solutions face significant challenges in implementation, particularly within educational institutions that rely on traditional assessment regimes and intellectual hierarchies. Moving from a model that evaluates intelligence to one that verifies the will to learn requires a major shift in institutional culture. They also note that their approach could be seen as underwhelming to proponents of AI-in-education who are focused on quantitative measures of thinking and standardized learning outcomes. However, the study posits that maintaining these hierarchies in the face of widespread AI adoption risks creating a passive, dependent learner who is prone to cognitive debt and metacognitive laziness. For the researchers, the goal is not to eliminate AI from the classroom, but to reconfigure the relationship between human intelligence and machine technology.
Ultimately, the paper argues for the institutionalization of intellectual independence as a foundational principle in both education and the workplace. This approach, which the authors characterize as a public pedagogy of and with AI, aims to free lifelong learning from the constraints of deficit models, where the learner is viewed as lacking knowledge that must be supplied by an external authority. The authors call for future research to explore how this thing-centred approach translates across diverse disciplines and how human-computer interaction can be further developed to ensure that AI systems act as extensions of human intelligence rather than as its substitutes.
In conclusion, the research presents a call for a fundamental reconsideration of the current trajectory of AI literacy and AI-mediated learning. By shifting focus from the binary of critical and naïve thought to the universal capacity for intelligence, educators can better prepare individuals to navigate the complexities of an AI-augmented world. The study emphasizes that while the proposed solutions are one of several possible emancipatory paths, they provide a necessary critique of the assumptions currently guiding educational policy. By prioritizing the engagement with a thing and the cultivation of an individual’s will, the proposed framework offers a pathway to ensure that learning remains a transformative and independent human exercise in an age where the risk of intellectual conformity is greater than ever before.
To illustrate how learning can occur without explanatory hierarchies, the analysis draws on the historical case of Joseph Jacotot, a nineteenth-century lecturer at the University of Louvain who tasked Flemish-speaking students with learning French using a bilingual edition of Le Télémaque. Because Jacotot spoke no Flemish and provided no textual explanations, the students acquired reading fluency solely by comparing the two texts, repeating passages, and verifying their observations independently. This historical precedent informs the authors’ pointed critique of traditional educational frameworks such as Bloom’s Taxonomy, which divides human cognition into hierarchical tiers ranging from lower-order memorization and comprehension to higher-order analysis and evaluation. The researchers argue that distributing learning outcomes across these rigid cognitive strata reinforces an artificial division of intelligence, wrongly presuming that foundational knowledge lacks creative synthesis or that lower-tier learners inherently require authoritative explicators to ascend toward critical consciousness.
The authors ground these philosophical ideas in a concrete classroom scenario featuring a university course built around the six-module Elements of AI open online curriculum. In this instructional model, the instructor explicitly relinquishes the role of an explicator and instead functions as an ignorant teacher who shares an equal distance from the course materials alongside the students. Rather than evaluating whether student interpretations match conventional expert conclusions, the instructor guides the group through perception-based provocations, such as asking learners what surprised them most about how machine models learn. Students are then asked to pinpoint exact locations within the instructional modules to support their reflections. Institutional evaluation requirements are satisfied through reflexive self-assessment and collaborative grading rubrics that measure invested study time, sustained attention, and engagement behaviors rather than testing intellectual conformity against standardized answer keys. Under this arrangement, teaching centers entirely on motivating sustained student effort, demonstrating that learners can acquire complex technical understanding independently.
Beyond classroom instruction, the paper specifies concrete technical and interface design principles for implementing thing-centred AI-mediated tutoring systems, illustrated through a graduate study group examining algorithmic bias. The proposed architecture requires language models to be tuned with mild temperature settings to generate unexpected provocations while strictly constraining outputs to verifiable locations within an uploaded digital artifact. To prevent cognitive passivity, the user interface features split-screen navigation, full-text search capabilities, and a strict call-and-response pacing mechanism that delivers one provocation at a time without issuing automated nudges or corrective evaluations. Furthermore, the system incorporates private note-taking spaces and transparent logging mechanisms that track document queries, navigational paths, and gaze as verifiable proxies for student effort. Crucially, these interaction logs remain in the possession of the learners rather than serving as administrative surveillance, allowing individuals to track their own intellectual development over time.
Subject of Research: Educational philosophy and AI literacy
Article Title: The Dizzy Spells of AI: Enacting Intellectual Independence in AI Literacy and AI-Mediated Learning
Article References: Hachem, H.-H., Primus, F., Linden, K., Heintz, F., & Wiggberg, M. (2026). The Dizzy Spells of AI: Enacting Intellectual Independence in AI Literacy and AI-Mediated Learning. Human-Centric Intelligent Systems, 6(3), 390-411. https://doi.org/10.1007/s44230-026-00168-z
Image Credits: AI Generated
DOI: 10.1007/s44230-026-00168-z
Keywords: AI literacy, intellectual independence, Rancière, thing-centred pedagogy, AI-mediated learning, educational philosophy, ignorant teacher, postcritical imagination, Bloom's Taxonomy critique, Socratic method, intellectual emancipation, AI-tutors, epistemology of ignorance, algorithmic bias, human-centric intelligent systems, metacognition, lifelong learning, educational hierarchies, pedagogy of AI, digital assessment
Cite Scienmag News
Blake Davidson. (November 17, 2026). Challenging Intellectual Dependency Through Thing-Centred AI Literacy. Scienmag. https://scienmag.com/challenging-intellectual-dependency-through-thing-centred-ai-literacy/
Blake Davidson. "Challenging Intellectual Dependency Through Thing-Centred AI Literacy." Scienmag, 17 November 2026, https://scienmag.com/challenging-intellectual-dependency-through-thing-centred-ai-literacy/. Accessed 20 November 2026.
Blake Davidson. "Challenging Intellectual Dependency Through Thing-Centred AI Literacy." Scienmag. November 17, 2026. https://scienmag.com/challenging-intellectual-dependency-through-thing-centred-ai-literacy/








