AI-Assisted Research Competency in Secondary Education: A Framework for Epistemic Agency, Authorship and Responsible Knowledge Production

Synthesis: Beau, Flaquière, and Lazar (2026) propose AI-Assisted Research Competency (AARC), an integrated capacity to conduct inquiry with generative AI without surrendering authorship, judgment, verification, or intellectual responsibility. Grounded in virtue epistemology, AI literacy, and AI intuition, AARC comprises seven teachable dimensions that develop from guided foundation through cross-domain transfer to autonomous inquiry. The framework translates the construct into an analytic rubric, a five-step decision routine, and three recurring commitments — verify, cite, and reflect — and is illustrated through a design case in upper-school courses. It offers secondary educators a concrete scaffold for responsible knowledge production in an AI-augmented era.

The Problem

Generative AI can support student research while obscuring whether learners actually performed the intellectual work an inquiry was meant to develop. This obscures the link between submitted work and student capability, and raises new questions about responsible knowledge production in education.

The AARC Framework

AARC defines the capacity to conduct inquiry with AI without relinquishing epistemic agency. It is grounded in virtue epistemology, AI literacy, and AI intuition, and comprises seven teachable dimensions that progress from a guided foundation, through cross-domain transfer, to autonomous inquiry. It is operationalized as an analytic rubric, a five-step decision routine, and three recurring commitments: verify, cite, and reflect.

What this means for practice

• Instructors. Grade the research process, not only the product: the analytic rubric scores observable behaviors — tool selection, interaction design, verification, synthesis and argument, authorship and disclosure — because a polished submission cannot show whether the student did the cognitive work.
• Instructors. Teach verify, cite, and reflect as three recurring commitments that apply at every phase of inquiry, and require students to defend their choices in reflective defense rather than merely declaring AI use.
• Curriculum designers. Sequence the competency deliberately, from a guided foundation through cross-domain transfer to autonomous inquiry, so students face escalating epistemic demands instead of a single AI policy at intake.
• Instructors. Hold learners to the rubric's top band in assessment — comparing AI and non-AI alternatives, detecting fabrication and bias, and defending an argument against objections — so that competent AI-assisted research means exercising judgment.
• Curriculum designers. Anchor the progression to tool-independent criteria (the seven dimensions and the verify–cite–reflect commitments) so the curriculum does not need rewriting with each model release.

Limitations

• The paper is explicitly conceptual and design-oriented, not an empirical study: the authors state that the rubric's reliability, the progression's effects on student reasoning, and the comparative value of scaffolded AI-assisted research against traditional or unscaffolded approaches remain to be tested.
• The framework originates in a single, well-resourced international school, and the authors state that transfer to differently resourced contexts cannot be assumed; cross-site implementation is listed as future work.
• The specific tools and failure modes named in the paper will date quickly given how fast AI capabilities change, as the authors acknowledge; only the seven dimensions, the AI literacy and AI intuition framing, and the verify–cite–reflect commitments are claimed to be tool-independent.

Citation

Beau, A., Flaquière, D., & Lazar, G. (2026). AI-Assisted Research Competency in Secondary Education: A Framework for Epistemic Agency, Authorship and Responsible Knowledge Production. EdArXiv preprint.