Resolution-Based Education: A Decision-Relative Architecture for Capability Certification in the Generative-AI Era

Generative artificial intelligence has made a longstanding assessment problem harder to ignore: a successful performance does not always provide sufficient evidence for the capability claim an institution intends to certify. This paper develops Resolution-Based Education (RBE) as a complete educational operating architecture that separates performance attainment from certification resolution. RBE preserves existing outcomes, marks, programme structures and assessment practices, but adds an explicit evidential question before a capability-bearing decision is made: does the available evidence distinguish learner possibilities that would require different certification decisions? The architecture connects educational purpose, programme outcomes, Resolvable Capability Outcomes, curriculum, pedagogy, learning evidence, performance assessment, a Resolution Gate, decision-targeted evidence acquisition, finite stopping, certification states, course and programme evaluation, governance and continuous improvement. When existing evidence is already resolution-adequate, the process stops with no additional assessment burden. When it is not, additional evidence is selected only to resolve the decision-relevant ambiguity, subject to validity, reliability, fairness, accessibility, privacy and burden constraints; unresolved cases may be deferred rather than forced into a false binary decision. The paper specifies the implementation sequence, reporting mathematics, governance requirements and a prospective validation design. RBE therefore treats certification as an evidence-to-decision problem: performance remains important, but a capability claim is made only when the evidence is adequate for the distinction being certified.