Artificial Intelligence and Communities of Inquiry: Reimagining Educational Experiences

Synthesis: This conceptual article by the creators of the Community of Inquiry (CoI) framework argues that generative AI adoption in education requires a coherent theoretical framework. Drawing on the CoI framework and its construct of shared metacognition, it shows how collaborative inquiry can integrate AI in ways that preserve human Learner Agency and sustain deep, meaningful learning. It warns that without critical, reflective inquiry, learners risk passivity, diminished authenticity, and overdependence on AI outputs. Shared Metacognition — collective monitoring and management of thinking — offers a responsible pathway for engaging critically with AI-generated content.

Key Findings

• A coherent theoretical framework is essential: The authors argue AI adoption must move beyond ad hoc implementation; the CoI framework, a collaborative-constructivist process model grounded in critical reflection and discourse, is well suited to guide responsible integration of generative AI (Stenbom & Garrison, 2026).
• AI is a sociotechnical actor, not a neutral tool: Within a community of inquiry, AI intersects with Teaching, Social, and Cognitive Presence in different ways, and may assume multiple, coexisting roles depending on design and positioning.
• Five roles for AI in a community of inquiry: (i) support for learning design, (ii) independent resource for learners, (iii) support function and guide for instructors via analytics, (iv) AI agent as a member of the community, and (v) sustained dialogical relationship of inquiry between human and AI agent.
• Risks of uncritical adoption: If reflective inquiry is bypassed, learners may become passive recipients of AI-generated information, accept outputs uncritically, and forfeit the cognitive engagement and authenticity essential to deep learning.
• Shared Metacognition is the responsible pathway: Comprising self- and co-regulation (each with monitoring and management functions), shared metacognition enables learners to audit and verify AI results and maintain reflective responsibility, supported by the Shared Metacognition instrument (Garrison & Akyol, 2015a, 2015b).
• Learning analytics and human oversight: AI-driven analytics can reveal learning processes and support monitoring of presences, but must be balanced against risks of bias and oversimplification; the human instructor remains in charge, with AI serving as adviser.
• Meaning requires human purpose: Regardless of AI's apparent reasoning, meaning becomes educational only when humans interpret, question, and integrate outputs — requiring educators to slow inquiry for reflection, verification, and skeptical engagement.

What this means for practice

• Instructors. Treat AI as a catalyst for inquiry rather than an answer source: structure complexity, visualize relationships, and prompt deeper questioning instead of letting a generated summary end the negotiation of meaning.
• Instructors. Slow inquiry deliberately so learners audit AI output, because verification, reflection, and skeptical engagement are what shared Metacognition adds, and bypassing reflective inquiry is what produces passive acceptance of generated content.
• Instructional designers. Position AI against the presences instead of above them: decide which of the five roles (learning design support, independent learner resource, instructor support through analytics, AI agent as community member, sustained dialogical inquiry) each activity needs, since AI intersects Teaching, Social, and Cognitive Presence differently.
• Administrators. Keep the instructor in charge and model critical use from instructional leadership: learning analytics can scale diagnostic insight into Cognitive Presence, but the framework puts the human instructor as the decision maker with AI as adviser.

Limitations

• This is a conceptual argument rather than an empirical study: it answers how a collaborative-constructivist perspective can inform understanding of generative AI, contributing no course data, participants, or outcome measures.
• Its empirical anchors are cited rather than generated: the metacognitive-support evidence it leans on (Martha et al., 2023) used the Shared Metacognition questionnaire in collaborative inquiry without generative AI, so the AI-specific claims remain untested.
• The framework and the instrument it recommends come from the same tradition as its authors — the Community of Inquiry framework and the Shared Metacognition instrument (Garrison & Akyol, 2015) — so the article extends its own theory rather than testing it against rival accounts.
• The five-role taxonomy is tied to the tools current at writing (ChatGPT, Copilot, Gemini, Claude); the authors themselves cite the warning that generative AI change is outpacing our capacity to understand and regulate it.

Citation

Stenbom, S., & Garrison, D.R. (2026). Artificial Intelligence and Communities of Inquiry: Reimagining Educational Experiences. International Review of Research in Open and Distributed Learning, 27(2), 114-131.