Representation-Aware Reasoning: From Representational Vigilance to Candidate Re-Representations

Representation-Aware Reasoning (RAR) addresses a conditional methodological problem: once an active representation becomes an explicit object of scrutiny, how should alternative representations be formulated and compared without turning an anomaly into an automatic requirement for representational change? RAR identifies the active representation, fixes the question and relevant constraints, preserves established achievements, examines local explanations and corrections, and formulates candidate re-representations with discriminating consequences. Its legitimate outcomes include PRESERVE, LOCAL CORRECTION, CANDIDATE SET, and DEFER / NON-DETERMINED. Proposed revision 3 explicitly distinguishes failure on an unchanged question from extension to a new question, clarifies the boundary between local correction and representational change, and aligns RAR with the seven-stage architecture of Vigilance Explicative et Représentationnelle (VER v5). RAR remains modular with respect to neighboring frameworks. TBER addresses emergence, transition, selection, stabilization, genealogy, and closure when these become the scientific question. RAWM addresses the organizational consequences of sufficiently specified or tested candidate representations within a functioning world model. A candidate generated or supported by RAR does not, by itself, establish an internal representational transition or justify integration into a world model. The article is conceptual and methodological. It includes a worked analytical example contrasting observation aliasing with calibration error and comparing present-state enrichment, history-based reconstruction, and probabilistic representation. It reports no new biological measurements, human-subject study, trained-system experiment, or empirical benchmark result. A prospective comparative evaluation program is proposed but has not yet been performed.