From Plausibility to Verifiability: The PEARLS Framework for Developing Epistemic Agency in Generative AI-Mediated Higher Education

Synthesis: Wang (2026) addresses the plausibility–verifiability gap created by Generative AI: an AI-generated artifact can display the linguistic and structural markers of expertise while the evidence, provenance, and limitations needed to warrant reliance remain difficult to inspect. The paper introduces PEARLS, an artifact-level verification protocol organized around six interdependent dimensions — Process, Evidence, Access, Reproducibility, Legitimacy, and Source — that treats AI output as a provisional knowledge claim whose warrant must be assembled and examined. It advances verification-driven learning as a pedagogical mechanism through which learners develop expertise, and it demonstrates how the relative emphasis of the six dimensions varies across disciplines.

The Plausibility–Verifiability Gap

Generative AI has sharply reduced the cost of producing fluent explanations, syntheses, analyses, code, and recommendations, but it has not reduced the intellectual work required to establish whether those outputs deserve belief or use. This asymmetry — plausible surface, hard-to-inspect warrant — creates a distinctive challenge for learners. Existing AI literacy and evaluative-judgment frameworks specify broad competencies for critical and responsible engagement, yet students and novice users still need an actionable method for evaluating a particular AI-mediated artifact.

The PEARLS Verification Protocol

PEARLS is an artifact-level verification protocol organized around six interdependent dimensions:

1. Process — examining the reasoning and procedure behind the AI-generated output.
2. Evidence — assessing the grounds, data, and support for the claims made.
3. Access — determining the Accessibility and availability of sources and materials.
4. Reproducibility — evaluating whether the result can be reproduced or verified independently.
5. Legitimacy — judging the authority, standing, and trustworthiness of the source.
6. Source — tracing the provenance and origins of the information.

The framework integrates insights from epistemic cognition, epistemic vigilance, Cognitive Offloading, calibrated trust, evaluative judgment, and open-science principles. Its core move is to treat AI output as a provisional knowledge claim whose warrant must be assembled and examined rather than accepted on the basis of fluency.

Verification-Driven Learning

The paper advances verification-driven learning as a pedagogical mechanism: learners develop expertise by iteratively focusing consequential claims, tracing and testing their warrant, judging uncertainty and legitimacy, acting on the results, and reframing subsequent inquiry. This positions the act of verifying AI output not merely as a protective skill but as a genuine learning activity that builds domain expertise and Critical Thinking.

Five interdisciplinary cases — psychology theory, educational statistics, computer science, history, and health sciences — demonstrate how the relative emphasis of the six dimensions varies with disciplinary standards and the consequences of error. The paper concludes by deriving implications for Assessment design and proposing a research agenda encompassing construct validation, intervention studies, disciplinary calibration, equity, and human–AI interface design.

What this means for practice

• Educators. Add a compact verification record to assignments: have students identify one to three consequential AI contributions, show the evidence chain used to examine them, state what changed, and explain the final reliance decision — keeping the record selective, since requiring every prompt produces surveillance and administrative overload without improving judgment.
• Educators. Teach PEARLS as a disciplinary practice rather than a generic checklist: model a think-aloud audit of a deliberately flawed output, have students compare verified and unverified exemplars and audit in pairs, then fade the prompts as expertise develops.
• Educators. Apply a risk-based rule to decide what to verify — claims that are central, surprising, numerical, source-dependent, ethically sensitive, or difficult to reverse.
• Assessment designers. Build rubrics that judge the quality of reasoning rather than the quantity of documentation, and pair documentation with brief oral explanation, executable artifacts, source annotations, or in-class verification tasks so fabricated process logs lose their value. Weigh the load against resources, since verification consumes time, database access, disciplinary language, and sometimes paid tools, and learners with fewer resources are asked to bear a larger burden to prove responsible use.
• Researchers. Begin with construct clarification and content validation — expert panels testing whether the six dimensions are distinct, comprehensive, and interpreted consistently — before moving to controlled studies that compare PEARLS instruction with generic critical-thinking prompts and ordinary AI use.

Limitations

• PEARLS is a conceptual framework that has not yet been psychometrically validated or shown to improve learning; the paper reports no data collection, sample, or field implementation.
• The author acknowledges that the six dimensions overlap — Source is a form of Evidence, Access enables Reproducibility, and Legitimacy constrains the others — which is theoretically intentional but may complicate measurement, and the mnemonic may encourage superficial box-ticking when detached from disciplinary standards.
• The literature synthesis is integrative rather than exhaustive, and the five cases (psychology theory, educational statistics, computer science, history, and health sciences) are illustrative rather than empirically tested.
• The cases reflect selected Anglophone scholarly traditions and may not capture how knowledge, authority, access, and legitimacy are understood across cultures and institutions; the author states that co-design with students, educators, librarians, disability specialists, and disciplinary communities is necessary before broad implementation.

Citation

Wang, Z. (2026). From Plausibility to Verifiability: The PEARLS Framework for Developing Epistemic Agency in Generative AI-Mediated Higher Education. EdArXiv preprint.