I Know How I Know: Mapping the Declarative Topography of Confidence in Higher-Order Assessment

Abstract Assessment outcomes coupled with confidence ratings can identify confidence–accuracy calibration, but they do not explain whether confidence is grounded in explicit evidence, reasoned justification, or intuitive judgement. This distinction is important in health-professions and higher-order assessment contexts, where learners must monitor the reliability, completeness and evidential basis of their own knowledge. This paper examines validity evidence for the Declarative Domain Metacognitive Instrument (DDMI), a four-category instrument where learners report their metacognitive articulation state: a guess with no statable rationale (DDMI 1); felt sense of the answer with no articulable basis (DDMI 2); partial reasoning stated but remaining incomplete (DDMI 3); or an explicit reason, rule, or mechanism the learner can state (DDMI 4). The DDMI is grounded in Koriat's distinction between experience-based and information-based monitoring and dual-process metacognitive theory. The instrument was embedded in pharmacology assessments in one nursing cohort (76 students, 4,258 responses), DDMI categories showed an ordered relationship with accuracy, rising from 26.4% to 53.9% (gamma = 0.268); the lowest category performed at chance, the next decisively above it. The gradient persisted after adjustment for item difficulty (OR = 1.403, 95% CI 1.310–1.502), within individual students and under clustering, and was not reducible to conventional confidence magnitude. Reported articulation was fallible: even DDMI 4 responses were inaccurate almost half the time, identifying cases in which learners reported an explicit basis for knowing but selected an incorrect response. The DDMI offers formative value for identifying misplaced certainty in health professions and higher-order assessment; response-process validation is the immediate next step.