Accountable Architecture for Thinking Under Uncertainty

Existing theories of consciousness lack a measurable criterion for accountability: the ability of a system to identify and isolate the source of its own errors. Without this criterion, it is impossible to formalise self-correction or assess cognitive processes. This paper introduces the Rozum Framework (RF), which derives from a single logical foundation: the Open World Theorem, which establishes that no system can achieve a complete and certain description of external reality. From this theorem emerge three consequences – the Existence, Accountability, and Innovation descriptive axioms (strategic imperatives). To describe information processing, the Reliability of Output formula is proposed RO = CI × CS × CA, representing the certainty of input, internal statements, and internal algorithms respectively. The constraint CI ∈ [0,1) is not a modeling assumption but a logical necessity. RF defines consciousness as accountable self-recognition and rozum as consciousness that thinks – a substrate-independent conscious entity enabled by and uses language. The framework identifies an inevitable architectural bias toward false certainty. By revealing the necessary logical connections among existing concepts RF provides the unified, substrate-independent architecture in which accountability and self-correction are measurable functional properties.