Preserving Human Judgment in AI-Enabled Education: The SAM Model for AI Readiness, Assessment, and Human-in-the-Loop Decision Making

Generative artificial intelligence (AI) shifts educational focus from access to information toward judging machine-generated claims and regulating the division of epistemic work. This conceptual article proposes the SAM model—self-regulation (S), ambiguity tolerance (A), and methodological competence (M)—as a bounded person-level resource model within broader AI readiness, with educators’ AI readiness as a central application. S denotes regulatory capacity, A a disposition-like willingness to tolerate unresolved epistemic ambiguity, and M the capacity to evaluate claims, differentiated into cross-domain M1 and domain-specific M2. SAM’s distinctive claim is configurational: M should predict independently assessed judgment more strongly at higher S and, under high ambiguity, when A is higher. These relations, the M1/M2 distinction, limited compensability, and incremental prediction beyond shared and neighboring accounts remain hypothetical. The model separates resources from enacted processes, contextual affordances, and outcomes and requires independent performance criteria. For AI-based assessment and learning analytics, educators’ interpretations of AI-derived evidence and construct inferences from behavioral traces require validated mappings addressing alternative explanations, comparability, fairness, and use. SAM thus offers a research agenda predicting when AI support preserves human judgment rather than merely improving supported performance.