Assurance by design: embedding the SAGE Defend step in AI-integrated higher education assessment
This paper conceptualises the SAGE Defend step, the sixth stage of the Structured AI-Guided Education framework, as a format-agnostic assurance checkpoint for AI-integrated higher education assessment. The study responds to a verification gap identified in earlier SAGE research, in which process documentation and AI interaction logs were found to support transparency but not, by themselves, to verify individual ownership of reasoning in group-based AI-integrated submissions. Adopting a design-informed conceptual approach grounded in design-based research principles, the paper integrates a multi-year programme of empirical SAGE studies, a structured synthesis of the assurance-task literature, and diagnostic observations from three Defend-proximate assessment implementations across undergraduate and postgraduate units at Central Queensland University. It distinguishes between assurance tasks that directly require students to demonstrate reasoning or performance, controlled assurance conditions that restrict the assessment environment, and corroborative assurance signals that provide corroborating but non-stand-alone evidence. On this basis the paper proposes a three-class assurance-task typology, an epistemic matching framework, and six design principles for embedding SAGE Defend within assessment sequences. It further argues that assurance should be distributed across the assessment sequence of a unit, so that each learning outcome is verified at a point and intensity proportionate to its stakes rather than concentrated in a single terminal examination. The paper frames this response as assurance by design, an approach that, echoing the established engineering principles of security by design and privacy by design, builds verification into the assessment sequence rather than appending it after the fact, and it names the compounding cost of the retrofitted alternative as assurance debt. Rather than presenting SAGE Defend as an oral examination model or claiming empirical validation of a single format, the paper positions Defend as a design principle through which educators can align verification tasks with the cognitive, professional, or technical competency being assessed. The contribution is therefore conceptual and practice-informed, offering a structured basis for the future empirical validation of specific Defend formats across disciplines, cohorts, and delivery modes.