Engineering Determinacy: Structuring Established Knowledge So That It Need Not Be Reinterpreted

Engineering work continuously produces information that becomes established: what a task covers, which behaviors are permitted, which artifact a requirement refers to, where a decision came from, and what state must survive a handoff. Yet established information often remains expressed in the same semantic form as information that is still unresolved. As long as it exists only in natural language, a later consumer must read, understand, and interpret it before use. What engineering settled previously can therefore become subject to interpretation again. This paper introduces Engineering Determinacy as a principle concerning the form in which established engineering information exists. Rather than remaining accessible only through semantic reconstruction, established information can be given an explicit engineering representation. Two conceptual operations are distinguished. Extraction makes explicit information already contained in semantic material, such as turning an implied scope into an enumerated set. Attachment assigns engineering information that the semantics did not originally contain, such as an identifier, anchor, or trace. Because explicit structure can still be incomplete, stale, or wrong, the principle also requires bounded engineering treatment sufficient to maintain the representation for its intended purpose. A further consequence is model independence. When an established engineering fact remains available only as semantic content, its operational interpretation may vary with the model, context, or interaction in which it is reconstructed. Once it has an explicit engineering representation, preserving that fact no longer has to depend on the semantic capability of the model consuming it. The principle was recognized retrospectively across a series of previously published mechanisms addressing different engineering problems. Scope sets, behavior states, identifiers, trace links, specification structures, decision classifications, and continuation fields were developed independently, yet repeatedly performed the same function: they gave established engineering information an existence independent of the semantics through which it was originally expressed. This paper names that shared property and positions it in relation to requirements traceability, formal specification, context engineering, knowledge representation, and AI governance. Keywords: Engineering Determinacy; AI-Assisted Software Engineering; Established Engineering Knowledge; Semantic Interpretation; Structured Representation; Requirements Traceability; Model Independence; AI Governance; Large Language Models; Software Engineering Practice