The emergence of powerful artificial intelligence systems in educational contexts has precipitated what might be characterized as an ontological crisis in higher education. As Anthropic's recent analysis of one million student conversations with it's leading edge AI model, Claude reveals, university students are engaging with AI across four distinct patterns—Direct Problem Solving, Direct Output Creation, Collaborative Problem Solving, and Collaborative Output Creation—each representing fundamentally different relationships to knowledge production. Yet institutional responses have largely centered on questions of prohibition versus permission, academic integrity versus cheating, and human versus artificial intelligence—binary framings that fundamentally misapprehend the onto-epistemological and changing socio-cultural and socio-economic nature of these emerging phenomena.
This essay draws upon the heterdox feminist physicits Karen Barad's ideas of agential realism, entanglement and intra-acton to develop a more adequate theoretical framework for understanding our changing educational system and this new human-AI educational assemblages' needs. Drawing on here quantum physics background Barad's theory proposes that the universe comprises phenomena which are "the ontological inseparability of intra-acting agencies," challenging individualist metaphysics through the concept of intra-action—a neologism that signals how objects emerge through particular intra-actions rather than preceding their interaction. Applied to educational contexts, this framework suggests that student-AI engagements should not be understood as interactions between discrete entities, but as intra-active phenomena that mutually constitute both "student" and "AI" as emergent agencies within specific learning assemblages and generating wider environments.
The stakes of this reframing extend beyond academic theory. As the Anthropic study reveals, students are not unsurprisingly primarily using AI for higher-order cognitive functions, with Creating (39.8%) and Analyzing (30.2%) being the most common operations, leading to almost hysterical concerns about an "inverted pyramid" where students rely on AI for complex thinking without developing foundational skills. The ideas from this range from banning AI from elementary school to banning AI altogether from more radical adherants. This pattern, interpreted through humanist frameworks, appears as cognitive outsourcing that threatens educational integrity and finds roots in previous bannings of technologies ranging from the calculator to the internet to Wikipedia for education. Through an agential realist lens, however, these new AI human relationships may signal the emergence of posthuman learning practices that better recognize the always-already distributed nature of cognition.
Agential Realism and the Material-Discursive Nature of Learning
To understand why current educational framings of AI may also prove inadequate, we must first examine Barad's critique of representationalism and its implications for learning theory. For Barad, agential realism serves simultaneously as epistemology (theory of knowing), ontology (theory of being), and ethics, captured in the neologism "onto-epistemology"—because specific practices of mattering have ethical consequences that exclude other kinds of mattering. This onto-epistemological understanding dissolves the nature/culture divide that has long structured educational thought, revealing learning as always-already material-discursive practice and to other extents, techno-ideological.
In agential realist terms, "the smallest units of analysis are phenomena: 'A phenomenon is a specific intra-action of an 'object' and the 'measuring agencies'; the object and the measuring agencies emerge from, rather than precede, the intra-action that produces them'". Applied to educational contexts, this means that "student," "AI," and "knowledge" do not exist as discrete entities that subsequently interact, but rather emerge through specific intra-active phenomena within material-discursive learning assemblages. The information or data is turned into knowledge through the student intra-acting with AI to produce higher order answers to questions that previously did not exist but appeared through this relationship specifically within the space of this encounter.
Consider the difference between asking "How is the student using AI?" versus "What student-AI phenomena are being enacted through specific learning practices?" The former question assumes bounded subjects and objects engaging in representational exchanges. The latter recognizes that studenthood, artificial intelligence, and knowledge itself are performed through relational becomings within particular apparatus. This shift has profound implications for how we understand learning, assessment, and educational ethics.
Beyond Individual Cognition: Apparatus and Distributed Agency
Barad's concept of apparatus challenges actor-network theory by proposing that apparatus are not assemblages of humans and nonhumans, but rather "the condition of possibility of 'humans' and 'non-humans,' not merely as ideational concepts, but in their materiality". Educational apparatus—comprising technologies, assessment practices, institutional policies, physical spaces, and embodied subjects—do not simply mediate pre-existing learning processes but actively participate in the ongoing materialization of learners, knowledge, and educational relations.
Recent scholarship has begun exploring how "agential realism offers a useful perspective for exploring how generative AI matters in literacy practices, not as a unilaterally destructive force, but as a set of phenomena that intra-actively reconfigures literacy practices". This perspective proves crucial for understanding why student-AI collaborations cannot be adequately assessed through frameworks premised on individual cognitive processing.
As Barad argues, "agency is a relationship and not something that one 'has'". In student-AI assemblages, agency emerges relationally through specific material-discursive practices rather than being possessed by either human or artificial entities. When students engage in what the Anthropic report categorizes as "Collaborative Problem Solving," they participate in the ongoing reconfiguration of agentic boundaries within learning phenomena. The question is not whether the student or the AI is the true agent of learning, but how particular configurations of student-AI-curriculum-assessment assemblages enact specific forms of distributed cognition.
Intra-acting Learning: Beyond Human-AI Interaction
The concept of intra-action proves particularly generative for understanding educational AI phenomena. Barad's formulation challenges "the usual notion of interaction which assumes a metaphysics of independent entities," proposing instead that "through and within intra-actions there is a differentiating-entangling so that an agential cut is enacted that cuts things together-apart (one move) such that differences exist not as absolute separations but in their inseparability".
Applied to student-AI learning assemblages, this means that students and AI systems do not exist as bounded entities that subsequently interact, but rather become-with each other through specific learning practices. When a student engages with Claude to analyze philosophical concepts or debug code, student and AI are mutually constituted through their relational becoming. The boundaries between "student thinking" and "AI processing" are not pre-given but enacted through particular cuts within the learning phenomenon.
This perspective radically reframes debates about academic integrity. Rather than asking whether students are "cheating" by using AI, we might ask: What kinds of learning subjects and objects are being materialized through specific student-AI-curriculum assemblages? How do different apparatus configurations enable or constrain particular forms of intellectual becoming? What ethical relationships are being enacted through these learning partnerships?
The Ethics of Entangled Learning
For Barad, ethics cannot be separated from ontology and epistemology because "specific practices of mattering have ethical consequences, excluding other kinds of mattering". In educational contexts, this means that decisions about AI integration are never merely technical or pedagogical—they are onto-ethico-epistemological choices that shape what kinds of subjects, objects, and relations become possible within learning assemblages.
As recent scholarship notes, "within agential realism, researchers are not only responsible for the kind of knowledge that they seek 'but, in part, for what exists'". Educators and institutions bear similar responsibility for the learning realities they help materialize through their responses to AI. Prohibiting AI use does not preserve some pure domain of human learning—it enacts particular cuts that may foreclose posthuman learning possibilities while reinforcing humanist fantasies of bounded cognition.
Conversely, unreflective AI integration risks materializing learning assemblages that reproduce existing inequalities or diminish student agency. The ethical question is not whether to allow AI in education, but how to configure student-AI-curriculum assemblages that enhance rather than diminish possibilities for flourishing.
Case Study: Reframing the Anthropic Findings
The Anthropic study's finding that "students are primarily using Claude to create and improve educational content across disciplines (39.3% of conversations)" while also "frequently using Claude to provide technical explanations or solutions for academic assignments (33.5%)" takes on new significance when interpreted through agential realist frameworks.
Rather than evidence of problematic cognitive outsourcing, these patterns might indicate students' intuitive recognition of learning as distributed phenomenon. When students engage AI for "creating coding projects or analyzing law concepts," they participate in material-discursive practices that have always characterized scholarly work—consultation with texts, collaboration with others, use of conceptual and technological tools. The difference is that AI makes visible the entangled nature of cognition that humanist frameworks have consistently obscured.
Disciplinary Differences and Material Specificities
The study's finding that "Computer Science students are particularly overrepresented (accounting for 36.8% of students' conversations while comprising only 5.4% of U.S. degrees)" while "Business, Health, and Humanities students show lower adoption rates" reveals how different disciplinary apparatus configure student-AI relations differently.
From an agential realist perspective, these disciplinary variations reflect the material-discursive specificities of different knowledge practices rather than simply different adoption rates. Computer science assemblages may more readily accommodate student-AI intra-actions because coding practices have long recognized the distributed nature of programming cognition—reliance on libraries, frameworks, documentation, and collaborative problem-solving. Humanities disciplines, conversely, may maintain stronger attachments to individualized authorial agency that render student-AI collaborations more problematic within existing apparatus.
The report's observation that "Natural Sciences & Mathematics conversations tended toward Problem Solving" while "Education showed the strongest preference for Output Creation" further illustrates how disciplinary apparatus shape the kinds of student-AI phenomena that emerge. These differences are not merely methodological preferences but reflect the material-discursive constitution of different knowledge domains.
Beyond the "Inverted Bloom's Taxonomy Pyramid" Metaphor
The study's concern about an "inverted Bloom's Taxonomy pyramid" where students rely on AI for higher-order cognitive functions while potentially neglecting foundational skills exemplifies the limitations of humanist learning theory. This metaphor assumes that learning proceeds hierarchically from simple to complex operations, with individuals accumulating cognitive capacities through linear progression.
Agential realism suggests a different understanding. Rather than discrete cognitive levels within individual minds, learning involves ongoing reconfigurations of distributed cognitive assemblages. The question is not whether students are developing proper cognitive hierarchies, but how student-AI-curriculum assemblages enable or constrain possibilities for intellectual becoming. While a student may not have generated an essay that they later claim they have authored, they have surely orchestrated the 'event' horizon' for this materialization and there is higher probability that they have even read the essay, tried to contribute to it adding a human flourish and their attempt at understanding, often the point of the exercise in the first place.
Students engaging AI for analysis and creation may be pioneering more sophisticated forms of distributed cognition that we have yet to learn to judge —learning to think-with AI in ways that exceed the capacities of either human or artificial intelligence alone. Rather than cognitive dependency, this might represent cognitive evolution toward posthuman learning practices better suited to contemporary knowledge conditions and to the previous Bloom's taxonomy, an upside-down pyramid. The upside down pyramid may also be viewed as the inverted telescope out to the wider cognitive universe through the student's eyes.
Implications for Educational Practice and Policy
Rethinking Assessment and Academic Integrity
If student-AI engagements are understood as intra-active phenomena rather than interactions between discrete entities, traditional approaches to academic integrity require fundamental reconceptualization that may involve large debates and overturning previous hierarchies. These hierarchies previously existed say in the dark ages where a priestly cast had access to literacy in Greek/Latin and the larger populace was reliant on them for the veracity of words. In the Renaissance this shifted so that the entire populace eventually could read and write. AI now enables the entire populace to produce knowlede and synthesize on higher levels to many extents opening wider doors perhaps also similarly too wide for a professorial elite. Rather than policing boundaries between authentic student work and AI assistance, educators might focus on configuring learning assemblages that enable ethical student-AI collaborations.
This might involve developing assessment practices that explicitly acknowledge the distributed nature of contemporary knowledge work and production while maintaining educational value. Rather than asking "Did the student or the AI produce this work?" we might ask "What kinds of learning relationships are being enacted through this student-AI collaboration and what has been learned?" and "How do these collaborations enhance or diminish possibilities for intellectual growth?"
Toward Posthuman Pedagogical Practices
Recent scholarship has begun exploring "how agential realism makes a difference ethically and politically" for educational practices, illustrating "how the philosophy works in all phases of education in terms of pedagogy and research". This work suggests possibilities for pedagogical practices that explicitly embrace the posthuman condition rather than defending humanist assumptions.
Posthuman pedagogies might involve:
- Collaborative Problem-Solving Assemblages: Rather than prohibiting AI use, educators could design assignments that explicitly require student-AI collaboration while building critical reflection on these partnerships and deepening how they operate and what the students have learned from working this way with AI.
- Diffractive Learning Practices: Following Barad's methodology of diffraction, students might learn to read ideas "through one another" by engaging multiple AI systems, comparing their responses, and developing critical perspectives on AI capabilities and limitations for a higher level discriminating use of AI say for STEM discipline, humanaities and visual (image) and audio-visual (video) and multi-modal purposes.
- Ethical Response-ability: Drawing on Barad's ethics, students could develop response-ability for the learning assemblages they participate in—recognizing their role in configuring the boundaries and possibilities of student-AI phenomena.
Institutional and Policy Considerations
The Anthropic study notes that "institutional policies regarding AI use in education vary widely, and might significantly impact the patterns we observe". From an agential realist perspective, these policies are not neutral regulations but material-discursive practices that actively participate in configuring educational realities.
Institutions might consider policies that:
- Acknowledge Distributed Cognition: Rather than policies premised on individual cognitive ownership, institutions could develop frameworks that recognize the always-already collaborative nature of knowledge production.
- Support Ethical AI Collaboration: Instead of blanket prohibitions or permissions, policies could provide guidance for ethical student-AI partnerships that enhance rather than diminish learning.
- Enable Experimentation: Recognizing the emergent nature of these phenomena, policies could create spaces for pedagogical experimentation with different student-AI configurations.
Challenges and Limitations
The Risk of Technological Determinism
While agential realism offers valuable frameworks for understanding student-AI assemblages, it risks obscuring power relations and structural inequalities that shape access to and uses of AI technologies. Not all students have equal access to AI systems, and the material conditions of their educational environments significantly influence the kinds of student-AI phenomena that become possible.
Furthermore, AI systems embody particular values, biases, and limitations that cannot be dissolved through theoretical frameworks emphasizing entanglement. As recent research notes, "ChatGPT and AI more broadly generates text in language that fails to reflect the diversity of students served by the education system or capture authentic voice of diverse populations". These material realities require critical engagement alongside posthuman theoretical frameworks.
Institutional Resistance and Practical Implementation
The shift from humanist to posthuman educational frameworks faces significant institutional resistance. Research suggests that "the failure to transform education through AI stems from a lack of consideration of Gardner's proposed ninth intelligence type—pedagogical intelligence". Educational institutions, assessment practices, and professional identities remain largely structured around humanist assumptions that may actively resist posthuman pedagogical innovations let alone the possibilities of Einstein's quantum physics and these entangled implications.
Practical implementation of agential realist approaches requires substantial changes to curriculum design, assessment practices, faculty development, and institutional policies—changes that may prove difficult within existing educational structures.
The Question of Human Flourishing
Critics might argue that emphasizing student-AI collaboration risks diminishing distinctively human capacities for critical thinking, creativity, and ethical reflection. While agential realism troubles the human-AI binary, this does not necessarily mean that all forms of student-AI assemblage enhance human flourishing.
The challenge is developing criteria for assessing which student-AI configurations enable flourishing without falling back into humanist assumptions about bounded individual development. This requires ongoing research into the effects of different learning assemblages on student capacities for critical engagement, creative expression, and ethical responsibility.
Future Directions and Research Implications
Empirical Studies of Learning Assemblages
This theoretical framework suggests the need for empirical research that examines student-AI learning assemblages as phenomena rather than investigating how students use AI tools. Such research might employ ethnographic methods, discourse analysis, and material-semiotic approaches to understand how different configurations of student-AI-curriculum assemblages enable or constrain particular forms of learning.
Recent work has begun exploring "posthumanist approach to AI literacy" through case studies that reveal "productive tension between students' experiments with posthumanist literacy and their entrenched humanistic assumptions". This research direction could be expanded to examine how students navigate between humanist institutional expectations and posthuman learning possibilities. We do really need injections of Dewey, Piaget, Montessori, Steiner or other forward thinking educators in times of change to think more deeply regarding 'education', testing and learning with our current tools
Comparative Analysis of Disciplinary Configurations
The disciplinary variations revealed in the Anthropic study suggest the need for comparative research examining how different knowledge domains configure student-AI relations. Such research might explore:
- How disciplinary epistemologies shape possibilities for student-AI collaboration
- What institutional changes might enable posthuman learning practices across different fields
- How disciplinary assessment practices might evolve to accommodate distributed cognition
Longitudinal Studies of Learning Trajectories
Rather than examining isolated student-AI interactions, research might investigate how student participation in AI assemblages affects learning trajectories over time. This could include studies of:
- How student-AI collaborations influence development of critical thinking capacities
- Whether early AI integration affects subsequent learning relationships
- How students develop ethical frameworks for AI collaboration
- How needs and use cases change from K-12, undergraduate, graduate and professional disciplines
Critical Studies of Power and Access
Agential realist approaches to educational AI require critical examination of how power relations, structural inequalities, and access issues shape the kinds of student-AI assemblages that become possible. Research might investigate:
- How socioeconomic factors influence student-AI collaboration opportunities
- How socio-cultural, peer group and familial environments influence student AI collaboration
- What cultural and linguistic biases are embedded in AI systems and how these affect different student populations
- How institutional resources and policies create differential access to posthuman learning possibilities
Conclusion: Toward Response-able Learning Assemblages
This essay has argued that Karen Barad's ideas regarding entanglement and intra-action offers beginning other metaphors and resources for understanding human-AI learning assemblages beyond the limitations of humanist educational frameworks. Rather than treating AI as an external tool that students either use responsibly or misuse through cheating, agential realism reveals student-AI engagements as intra-active phenomena that challenge foundational assumptions about individual cognition, knowledge ownership, and learning processes. These should also not be separated from workforce imperatives but rather prepare students for the AI dominated world they will be entering, hopefully better equipped than their parents or current generation not having these opportunities through their formal educational possibilities
The empirical evidence from Anthropic's study of student-AI interactions, when interpreted more positively through agential realist frameworks, suggests that students may be pioneering posthuman pedagogical practices that better acknowledge the always-already distributed nature of cognition and evolutionary drive of the students with nature and AI in alignment. Rather than cognitive outsourcing that threatens educational integrity, student-AI collaborations might represent learning evolution toward more adequate engagement with contemporary knowledge conditions and the global technocentric society they will be entering.
However, this reframing does not dissolve ethical questions about AI in education—it reconfigures them. Instead of asking whether AI use is permissible, we must ask how to configure student-AI-curriculum assemblages that enhance possibilities for flourishing while acknowledging the material constraints and power relations that shape these configurations.
In educational contexts, this means developing response-ability for the learning assemblages we participate in materializing. Educators, students, and institutions share responsibility for configuring educational apparatus that enable rather than constrain possibilities for intellectual and ethical development within the society students will be entering.
The challenges ahead are significant. Educational frameworks requires substantial changes to institutional structures, assessment practices, and professional identities. It demands critical engagement with power relations and inequalities that shape access to educational AI. And it necessitates ongoing research into how different learning assemblage configurations affect student development and social flourishing.
Yet the potential rewards are equally significant. New educational approaches offer possibilities for learning practices better suited to our entangled world—practices that acknowledge the distributed nature of cognition while maintaining commitments to critical engagement, creative expression, and ethical responsibility. As AI technologies become increasingly integrated into educational contexts, new frameworks provide essential resources for navigating these changes in ways that enhance rather than diminish possibilities for human and more-than-human flourishing.
The students in Anthropic's study who engaged AI for collaborative problem-solving and output creation may be ahead of their institutions in recognizing the posthuman condition of contemporary learning. The question is whether educational structures can evolve to support these emergent practices while maintaining commitments to equity, critical thinking, and ethical development. Agential realism suggests that this evolution is not only possible but necessary—and that it requires new forms of response-ability for the learning assemblages we collectively bring into being.
References
Anthropic. (2025). Education Report: How University Students Use Claude. https://www.anthropic.com/news/anthropic-education-report-how-university-students-use-claude
Barad, K. (2003). Posthumanist performativity: Toward an understanding of how matter comes to matter. Signs: Journal of Women in Culture and Society, 28(3), 801-831.
Barad, K. (2007). Meeting the Universe Halfway: Quantum Physics and the Entanglement of Matter and Meaning. Duke University Press.
Murris, K. (2022). Karen Barad as Educator: Agential Realism and Education. SpringerBriefs in Education.
Scholz, G. (2024). Agential realism as an alternative philosophy of science perspective for quantitative psychology. Frontiers in Psychology, 15, 1410047.
Stanford Institute for Human-Centered AI. (2024). AI will transform teaching and learning. Let's get it right. HAI News.
#KarenBarad #AIEducation #IntraAction #DistributedCognition #AcademicIntegrity ##AILiteracy #EducationalPhilosophy #LearningAssemblages #AICollaboration #CriticalAIStudies #EducationalTechnology #PhilosophyOfEducation #AIEthics #CollaborativeLearning #DisruptiveEducation #EdTechResearch #RelationalLearning #EducationalInnovation #AIPolicy