Digital Assessment of Metacognition Across the Psychosis Continuum: Measures, Validity, and Clinical Integration-A Scoping Review.

Abstract

<i>Background and Objectives:</i> Metacognition-related processes (e.g., confidence calibration, self-evaluation and the use of feedback) have been linked to cognitive insight, self-evaluation, and daily functioning in psychosis. However, clinic-based assessments only provide limited information. Digital methods may capture state-like variations and contextual factors, but it is unclear to what extent they operationalise core metacognitive monitoring constructs versus adjacent self-evaluative/insight-related constructs. We mapped digital approaches used to assess metacognition-related constructs across the psychosis spectrum, summarising the associated feasibility and validity. <i>Materials and Methods:</i> We conducted a scoping review (PRISMA-ScR) of psychosis-spectrum studies that used digital tools to assess metacognition-related targets. These included ecological momentary assessment/experience sampling (EMA/ESM), task-based paradigms with confidence ratings, and hybrid approaches. Searches covered MEDLINE (via PubMed), Scopus, and IEEE Xplore, with the final search run on 15 December 2025. We charted constructs, operationalisations, feasibility/engagement indices and reported links with clinical or functional measures. <i>Results:</i> The empirical evidence map comprised 13 studies directly assessing metacognition-related constructs; eight additional implementation/methodological sources were synthesised separately to contextualise feasibility, reporting, ethics, and governance. EMA studies more often assessed adjacent self-evaluative constructs, including context-linked self-appraisal bias, conviction, and self-report-context mismatch in daily life, whereas task-based studies more directly assessed confidence-accuracy calibration and feedback updating. Across EMA studies, greater momentary symptom severity and more restricted contexts were often associated with inflated self-evaluations and divergence from observer-rated functioning. Task-based studies indicated that confidence calibration and feedback utilisation may diverge from objective performance; in performance-controlled paradigms, some studies reported comparable metacognitive sensitivity/efficiency, but the overall evidence remains uncertain. Passive sensing was common in psychosis research but was rarely explicitly tied to metacognitive constructs. <i>Conclusions:</i> Current digital work spans both core metacognitive monitoring constructs and adjacent self-evaluative/insight-related constructs, rather than a single unitary construct. Clinical translation remains hypothesis-generating: interpretability may be improved by combining clinical anchors, low-burden EMA, and optional contextual streams, but thresholds, workflows, and signal-action rules require prospective validation.