Generative AI as a Didactic-Pedagogical Mediator: Rethinking Human Roles and Pedagogical Design in Higher Education

Synthesis: Moganadas and colleagues (2026) argue that generative AI in higher education cannot be understood through the traditional dyadic instructor–student model, because GenAI now participates directly in explanation, Feedback, and knowledge construction rather than serving as an external supplement or threat. They propose a nested instructor–student–GenAI triadic model that positions GenAI as a bounded didactic-pedagogical mediator operating within a shared didactic mediation space, governed by institutions and wider stakeholders. The framework translates into five researchable propositions covering learning mediation, instructor role transformation, developmental AI literacy and learner Learner Agency, AI-transparent process-oriented Assessment, and institutional AI Governance.

Key Findings

The dyadic model is structurally inadequate. Existing pedagogical frameworks interpret learning primarily as an interaction between instructor and student, with GenAI positioned either as an external tool or as a threat to academic integrity. Because students already use GenAI independently, this framing fails to capture how it reconfigures pedagogical agency, responsibility, and knowledge work.
GenAI is best framed as a bounded didactic-pedagogical mediator. The paper distinguishes this role from tool, tutor, assistant, scaffold, or co-educator: GenAI's outputs (explanations, examples, feedback, simulations, prompts) support representation and learning only when students interpret, question, verify, revise, and integrate them under instructor-designed tasks and disciplinary criteria. Its mediating function therefore depends on human judgment and institutional governance.
Learner agency and AI literacy are not automatic. Agency is a moral and cognitive capacity that must be deliberately cultivated through task design, transparent assessment, disciplinary guidance, institutional support, and equitable access. It develops progressively across a program of study rather than from tool access or basic prompting.
Assessment is a pedagogical design condition, not just an evaluation mechanism. Assessment design shapes whether students use GenAI to bypass cognitive effort or to support drafting, verification, comparison, and visible reasoning. AI-transparent, process-oriented design that requires documentation, verification, and reflective justification is a more educationally defensible response than prohibition or detection.
Instructor work is redistributed rather than reduced. The instructor's role shifts from the exclusive source of information toward orchestrator, curator, critic, ethical steward, and pedagogical innovator — a conditional, institutionally supported transformation grounded in Technological Pedagogical Content Knowledge (TPACK), professional agency, and AI literacy rather than an automatic consequence of GenAI availability.
Five unresolved tensions remain. The paper identifies epistemic (fluency versus factual reliability), ethical (the accountability gap for AI-generated work), structural (platform dependency and commercial control), equity (access versus meaningful, critical use), and affective-relational (human care and social presence) tensions that pedagogical design alone cannot resolve.

Study Design & Method

This is a conceptual hypothesis-and-theory paper rather than an empirical study. Its model is derived through a focused integrative interdisciplinary synthesis of literature from higher education, educational technology, learning sciences, instructional design, cognitive psychology, policy, ethics, and institutional governance, analyzed inductively and deductively rather than as an exhaustive systematic review. The result is a nested, multi-level model with three interdependent layers. The student, instructor, and GenAI are arranged around a central shared didactic mediation space — the arena where task design, prompting, GenAI outputs, feedback, verification, revision, reflection, and assessment are enacted — in which interaction, negotiation, co-regulation, and production operate as interconnected processes, and knowledge construction emerges from students' active meaning-making across human participants, AI outputs, disciplinary criteria, and institutional expectations. Higher education institutions form a governance layer that enables, constrains, legitimizes, and regulates the triad through curriculum and assessment structures, policies, infrastructure, professional development, procurement, data protection, and accountability; they are expected to act adaptively toward emergent risks while staying anchored in institutional aims, especially because student GenAI use often happens beyond managed platforms. Around them, an outer ecosystem of accreditation bodies, employers, policymakers, regulators, technology providers, and platform companies shapes institutional governance indirectly through professional expectations, regulatory frameworks, labor-market demands, and procurement dependencies.

The model integrates four theoretical perspectives: sociotechnical and ecosystem (sociotechnical theory, activity theory), instructor-design (Technological Pedagogical Content Knowledge (TPACK)), learning-mediation (sociocultural theory, distributed cognition, Scaffolding, cognitive load theory, ICAP), and inclusive-relational (UDL, Community of Inquiry). Five researchable propositions operationalize it: (1) structured, instructor-designed tasks that require verification and reflection strengthen GenAI-mediated learning and transfer; (2) instructor role transformation depends on AI literacy, TPACK, professional agency, and institutional support; (3) students become AI-literate co-constructors of knowledge when they are progressively supported across a program of study; (4) AI-transparent, process-oriented assessment design aligned with learning outcomes improves learning quality, learner agency, and academic integrity; and (5) institutional governance and infrastructure enable responsible triadic pedagogy when they provide clear policy, equitable access, data protection, professional development, human oversight, and accountability. The model is discipline-sensitive, illustrated across medicine, law, creative arts, business and management, engineering, and teacher education.

What this means for practice

Instructors. Put the paper's epistemic questions to work on every GenAI-assisted task: ask students what kind of knowledge claim a response makes, what evidence would justify it, what assumptions it embeds, which disciplinary standards apply, what has been omitted or simplified, and what lies beyond the system's capacity to judge. That questioning is what turns an AI output from an answer into the object of inquiry the triadic model requires.
Instructors. State for each task component whether GenAI is prohibited, permitted, required, or restricted, and require documentation of prompts plus the accept, reject, and revision decisions students made, with rubrics that reward reasoning, verification, source use, revision quality and ethical disclosure rather than the final product alone.
Instructors. Do not build your integrity response on AI-detection tools — detectors remain contested and may produce false positives, particularly for students writing in a non-native language — and treat discussion, the modeling of disciplinary judgment, and feedback conversations as non-delegable contact time, since the affective-relational tension between human care and social presence is the one condition of learning the paper says pedagogical design alone cannot resolve.
Faculty developers. Embed AI literacy progressively as a graduate competency across programs — foundational (acceptable use, privacy, integrity, hallucinations, bias, authorship), intermediate (inquiry, feedback interpretation, comparison, revision, reflective learning), and advanced (epistemological and methodological judgment) — rather than delivering one-off technical workshops.
Administrators. Support instructors through professional development embedded in peer co-design and communities of practice, recognize assessment redesign, output verification and AI-literacy support as legitimate academic labor in workload models and promotion criteria, protect instructor professional judgment from institutional or commercial pressure, and back this with principle-based adaptive policy, equitable access, data protection and periodic stakeholder review, calibrating strictness to discipline so that high-stakes fields such as medicine and law require stricter verification and oversight than open-ended creative work.

Limitations

The study presents no primary empirical data; the model is conceptual, synthesizing existing literature and offering researchable propositions rather than empirically validated findings; the focused integrative synthesis is not a systematic review (e.g., PRISMA), so the literature base may not capture every relevant empirical study or academic discourse.
The disciplinary anchors (medicine, law, creative arts, business, engineering, teacher education) are illustrative and do not constitute evidence that the model operates as proposed in those contexts.
The model remains abstract and may operate differently across institutional contexts — research-intensive, teaching-oriented, open and distance, transnational, private, and resource-constrained institutions.
Identified mediators and moderators remain unquantified: GenAI-supported feedback may improve revision in one context and foster dependency in another, and AI-transparent assessment may strengthen integrity while increasing workload elsewhere.

Connected Concepts

Connected Articles

Reclaiming Epistemic Agency: A Critical Framework for Human-Generative AI Co-Agency in Education — Critical framework for human-GenAI co-agency in education
From Plausibility to Verifiability: The PEARLS Framework for Developing Epistemic Agency in Generative AI-Mediated Higher Education — Framework for developing epistemic agency in GenAI-mediated higher education
Beyond Detection: Redesigning Authentic Assessment in an AI-Mediated World — Moving beyond detection toward authentic assessment with AI
Generative AI across the disciplines: an activity theory perspective on undergraduate students' AI use and disclosure practices — Activity-theoretic analysis of GenAI across disciplines
Reconceptualizing Community of Inquiry in the Age of Generative Artificial Intelligence — Reconceptualizing Community of Inquiry for generative AI
Same Question, Different Answer? Measuring and Mitigating Prompt Privilege for Equitable AI Access — Equitable AI access and the prompt-privilege gap

Citation

Moganadas, S. R., Marín-González, F., Nun, S. H., & Gan, C. L. (2026). Generative AI as a Didactic-Pedagogical Mediator: Rethinking Human Roles and Pedagogical Design in Higher Education. Frontiers in Education, 11, 1856839.

Metadata

Study design	Theoretical Analysis
Level	Higher Ed
Intended audience	Instructors
Learning and instruction	Pedagogies and Teaching Strategies
Technologies and techniques	Generative AI
Assessment and measurement	Assessment
Institutions and policy	AI Governance
Foundations of AI in education	Learner Agency, AI Literacy, Learning Design, Teaching
Page kind	Framework

Source History Edit Chat