Framing human-AI dynamics: An epistemological perspective on generative AI practices

Synthesis: Strydom (2026) introduces seven human–Generative AI (GAI) engagement paradigms grounded in personal epistemological beliefs, addressing what the author calls the "theory deficit" in educational technology. The conceptual, theory-building paper differentiates five epistemological dimensions (source, certainty, organization, control, and speed of knowledge acquisition) and uses them to theorize seven enacted paradigms of human-GAI engagement: guarded, possibility-focused, augmented, pioneering, symbiotic, values-based, and equity. Rather than fixed traits, the paradigms are enacted patterns of engagement that emerge across disciplinary, socio-technical, and institutional contexts — with direct implications for teaching, assessment, professional development, and AI governance. This is a significant theory-building contribution to the knowledge base's AI in Education foundational strand.

Key Findings

1. A theory-building response to the "theory deficit." The paper argues that educational-technology and GAI research suffers from limited theory development, and responds with a conceptual framework — grounding human-GAI engagement in Schommer's multidimensional model of personal epistemological beliefs rather than in tool-evaluation or common-sense deduction.
2. Seven enacted paradigms. The framework theorizes seven human-GAI engagement paradigms — possibility-focused (Optimist), pioneering (Dreamer), augmented (Enhancer), symbiotic (Collaborator), guarded (Guardian), values-based (Defender), and equity (Advocate) — each examined against the five epistemological dimensions.
3. Paradigms are enacted, not possessed. Individuals are not fixed in a single paradigm; they move between them across contexts or adopt hybrid positions. This has direct consequences for practice — shifting engagement requires changing the socio-technical environment, not just beliefs.
4. A continuum from optimism to caution. The paradigms span from optimistic (possibility-focused, pioneering) and capability-enhancing (augmented, symbiotic) positions to cautious (guarded), ethical (values-based), and equity-critical (equity) orientations — reflecting the full range of the field's discourse on GAI.
5. Exploratory and falsifiable by design. The framework is explicitly untested — it cannot yet establish paradigm prevalence, stability, movement over time, or whether implied interventions work. It is offered as a heuristic and a stimulus for empirical research rather than a complete account.

The epistemological grounding

The paper builds on Schommer's multidimensional model of personal epistemological beliefs, differentiating among five dimensions:

• Source of knowledge — whether knowledge originates externally or is constructed by the knower
• Certainty of knowledge — whether knowledge is fixed or evolving
• Organization of knowledge — whether knowledge is compartmentalised or integrated
• Control of knowledge acquisition — whether learning is controlled by the learner or external forces
• Speed of knowledge acquisition — whether learning is quick/all-or-nothing or gradual

Strydom contends these dimensions offer a productive lens on how individuals differently position themselves relative to GAI — explaining why the same tool is engaged very differently across people and contexts. The framework thus extends the knowledge base's Learning Theories strand into the human-GAI interaction space.

The seven paradigms in detail

• Possibility-focused (Optimist): An optimistic epistemological stance acknowledging GAI's power to synthesize, organize, and identify patterns across large literatures. Uncertainties are viewed as sites of productivity, provided the source of knowledge is appropriately distributed between human and machine, with humans retaining final interpretive judgment.
• Pioneering (Dreamer): Views GAI as a participant in knowledge creation rather than a means. Grapples with the philosophical implication that if innovation can be attributed to human-machine co-production, the origin of knowledge becomes an open matter. Associated with accelerating Peer Assessment, but raises speed-vs-rigor trade-offs.
• Augmented (Enhancer): Views GAI as an embedded tool that becomes part of individual reasoning by changing the organization of knowledge. Extends the extended-mind thesis toward an "amplified mind" — individuals actively contribute to a dynamic cognitive space while retaining interpretive Learner Agency.
• Symbiotic (Collaborator): A relational, posthumanist orientation viewing human and machine cognition as permeable and entangled; knowledge is co-created via the relationship itself, with Learner Agency distributed across the human-machine dyad rather than retained individually.
• Guarded (Guardian): Foregrounds epistemic vigilance — interrogation of who makes knowledge claims, on what grounds, and their reliability before acceptance. Reflects concerns about detection accuracy, academic integrity, authorship, and techno-solutionism; not resistance but deliberate, slower verification.
• Values-based (Defender): Prioritizes ethical AI use and the ethics of knowledge creation — fairness, transparency, accountability, Privacy. Asks not just whether claims are true but how they were arrived at responsibly.
• Equity (Advocate): Foregrounds power, ideology, and equity. Recognizes GAI's democratizing potential while attending to digital inequality and bias — GAI trained on English-language, Western academic sources risks reproducing hierarchies of whose knowledge counts; technology is not neutral.

What this means for practice

• Instructors. Diagnose the enacted paradigm before designing the task: because the paradigms are enacted rather than possessed, changing how students engage with GAI means changing the socio-technical environment — task design, disciplinary norms, assessment rules — not persuading students to hold different beliefs about AI.
• Instructors. Let pioneering and possibility-focused learners experiment with alternative Assessment formats — process evidence, co-produced work, exploratory drafts — instead of requiring polished-product-only submissions that suit only the guarded.
• Faculty developers. Treat a guarded orientation as a coherent epistemological position rather than resistance, and use the seven paradigms as a diagnostic vocabulary for differentiated staff development aimed at the values-based and equity orientations.
• Administrators. Apply the five epistemological dimensions when writing institutional AI policy: the same tool is engaged in different ways across disciplines and departments, so a single institutional stance on GAI will not fit all of them.
• Researchers. Treat the framework as a testable heuristic rather than a settled account: it extends the knowledge base's Human AI Collaboration, Student-AI Interaction, and philosophy of AI threads, but paradigm prevalence, stability, and movement between paradigms remain unmeasured.

Limitations

• The paper is conceptual and theory-building, not empirical: the seven paradigms were derived through literature synthesis and conceptual abstraction, so nothing here establishes that they correspond to stable, observable patterns.
• It cannot yet show how prevalent each paradigm is, how individuals move between them, or whether hybrid positions hold — the questions on which its central claim, that paradigms are enacted rather than possessed, depends.
• The implied interventions (differentiated staff development, alternative assessment formats, paradigm-informed governance) are untested; the author positions the framework as a stimulus for research and dialogue rather than a complete account.
• The grounding is a single theoretical lineage — Schommer's multidimensional model of personal epistemological beliefs — refined against source literature by one author, with no independent or inter-coder validation of the paradigm set.

Citation

Strydom, S. (2026). Framing human-AI dynamics: An epistemological perspective on generative AI practices. Journal of Applied Learning & Teaching, 9(2).