Towards a philosophy of ensemble cognition: Reconceptualising agency and mind in AI-mediated educational environments

Synthesis: Jho, Park & Ahn (2026) introduce Ensemble Cognition (EC) as a philosophical framework reconceptualizing thinking as emerging from dynamic interactions between human and artificial agents, rather than residing solely within individual minds. EC distinguishes functional agency (AI's capacity to influence outcomes without consciousness) from moral responsibility (which remains exclusively human), and articulates five features—distributed agency, dynamic centrality, cognitive orchestration, multi-representational integration, and context-sensitive switching—to characterize learning environments where cognitive leadership shifts dynamically between humans and AI. It offers educational philosophy conceptual resources for critically engaging with AI-mediated learning while preserving commitments to human flourishing and educational justice.

Key Findings

• EC as a post-consciousness-paradigm framework. The paper argues that contemporary AI-mediated environments strain three foundational assumptions of the "consciousness paradigm" in educational philosophy: the autonomy assumption (authentic learning from individual reflection), the consciousness assumption (cognition requires first-person experience), and the stability assumption (cognitive roles are fixed across contexts). EC reconceptualizes thinking as a collaborative achievement among human and artificial agents.

• Critique of 4E cognition, distributed cognition, and actor-network theory. Each framework raises specific conceptual questions in AI-mediated education: 4E cognition (embodied, embedded, enacted, extended) was built for biological cognition and struggles with agents that lack sensorimotor embodiment yet inherit embodied structure linguistically; distributed cognition (DCog) excels descriptively but provides insufficient normative resources for evaluating whether cognitive distributions serve educational goals; actor-network theory (ANT) attributes symmetrical agency to humans and non-humans, which conflicts with education's fundamentally asymmetrical Ethics and obscures distinctions between functional participation and moral responsibility.

• Functional agency without consciousness. AI systems exhibit adaptive responsiveness, contextual sensitivity, and generative capacity that place them in an intermediate category between passive instruments and conscious agents. This is functional agency—genuine causal efficacy in cognitive processes that does not require self-reflection, intentionality, or phenomenal consciousness.

• Functional agency ≠ moral responsibility. The framework's central distinction addresses anthropomorphization concerns: AI can influence learning outcomes without bearing moral responsibility for them. Responsibility remains with human agents who design, deploy, and oversee AI systems in educational contexts.

• Five features of ensemble cognition. (1) Distributed agency—no single actor monopolises cognition; (2) dynamic centrality—cognitive leadership shifts by task demands, shaped even by how users craft prompts; (3) cognitive orchestration—coordinated integration of diverse cognitive resources toward coherent outcomes; (4) multi-representational integration—integration of linguistic, visual, and mathematical resources (distinct from neuroscientific multimodality); (5) context-sensitive switching—cognitive systems adapt their organization to situational demands, and different AI architectures generate qualitatively distinct collaborative dynamics.

• Distributed metacognitive awareness. Thinking about thinking expands from individual self-monitoring to collaborative cognitive management: students must understand how different agents contribute, when to rely on AI versus human judgment, and how to orchestrate collaborative processes. This raises questions about whether it enhances or outsources distinctively human capacities for reflection and self-direction.

• Distributed pedagogical arrangements and orchestrated authorship. Teachers become cognitive orchestrators managing interactions between students, AI, and resources; student authorship is reconceptualized as "orchestrated authorship" where responsibility lies in transparently managing collaborative processes and exercising critical judgment over AI contributions.

• Critical boundaries and concerns. The framework acknowledges the transparency problem (opaque AI decision-making), cultural embeddedness (AI privileging particular values like efficiency over care, caution, and contextual sensitivity, per Jackson 2025), human cognitive development and intellectual autonomy, power and democratic participation, and its own cultural situatedness within Western cognitive science and analytic philosophy of mind.

Educational Significance

EC's value lies in offering conceptual vocabulary for asking better questions about human-AI collaboration in learning, navigating between uncritical anthropomorphization of AI and reductive instrumentalism that treats it as a mere tool. It directly engages the Cognitive Offloading debate, questions about Learner Agency and intellectual autonomy, and the reconceptualization of Metacognition as collaborative. The framework positions Human AI Collaboration and Pedagogical Agent roles within Higher Education and broader AI in Education contexts, while its engagement with Embodied Learning and Constructivism traditions grounds it in existing learning theory.

What this means for practice

• Instructors. Make the distribution of cognitive work explicit in every task—state what the human decides, what the AI contributes, and where responsibility sits—so students see that an AI's functional agency never carries moral responsibility with it.
• Instructors. Teach Metacognition as a collaborative rather than a solo skill: have students articulate how different agents contribute, when to rely on AI output versus human judgment, and how the collaborative process is to be orchestrated.
• Designers. Treat dynamic centrality and context-sensitive switching as design variables rather than background conditions, since different AI architectures generate qualitatively distinct collaborative dynamics and shift who leads a task.
• Researchers. Measure whether distributed arrangements enhance or outsource reflection and self-direction instead of assuming either, because the framework states the empirical adequacy of ensemble cognition as an open question.
• Faculty developers. Audit whose values the tools encode—efficiency over care, caution, and contextual sensitivity—and make that audit part of how staff exercise critical judgment over AI contributions.

Limitations

• This is a conceptual, philosophical analysis with no empirical data: the five features, functional agency, and ensemble cognition itself are argued rather than tested, and the authors list the framework's empirical adequacy as unresolved.
• The framework is culturally situated in Western cognitive science, analytic philosophy of mind, and Anglo-American educational philosophy; its individual–AI emphasis may fit collectivist educational contexts poorly, and the authors call for engagement with Indigenous and Eastern traditions before claiming global applicability.
• Granting AI functional agency risks the anthropomorphization the authors set out to avoid, and the collaboration emphasis may mask power relationships and inequalities embedded in AI systems rather than address them.
• The transparency problem is identified, not solved: the opacity of AI decision-making is argued to be potentially incompatible with educational ideals of transparency and comprehensibility.

Citation

Jho, H., Park, C., & Ahn, D. (2026). Towards a philosophy of ensemble cognition: Reconceptualising agency and mind in AI-mediated educational environments. Educational Philosophy and Theory. doi:10.1080/00131857.2026.2654678.