Architecture Science: A Research Program for Architecture Formation and Transformation

Abstract

AS-02 converts the bounded foundations of Architecture Science into an externally actionable research program for architecture formation and transformation. It specifies research workstreams, major research questions, empirical hypothesis families, minimum experimental units, comparative-study requirements, negative controls, simulation roles, dataset and benchmark architectures, counterexample programs, replication levels, maturity criteria, and explicit program-level strengthening, narrowing, absorption, restructuring, partial-rejection, and reconstruction/discontinuation dispositions. The paper is deliberately evidence-bounded. Architecture Science-specific empirical propositions remain at CS-2 — Research Hypothesis or lower. The release does not claim that Architecture Science is an established independent discipline, that universal architecture laws have been demonstrated, that generalized architecture reconstruction has been validated, that transferable identity criteria or the transformation taxonomy are established, that hostile-domain transfer or independent generativity has been demonstrated, or that architecture-level analysis has shown added value over the strongest appropriate native methods. The initial research portfolio prioritizes architecture-versus-aggregate discrimination through unstructured and randomized-relational negative controls, generalized reconstruction against strong native inverse methods, identity and transformation classification, hostile-domain transfer, comparative added-value testing, independent unassisted use, and benchmark/data specification. The program is designed so that negative results, counterexamples, predecessor-discipline capture, specialization collapse, native-method superiority, and replication failure can narrow, restructure, absorb, or reject Architecture Science propositions. This Zenodo record is designed as a single co-release object containing the AS-02 manuscript and its Supporting Constitutional & Scientific Record. The Supporting Record provides bounded traceability for claim states, literature and comparator burdens, falsification conditions, cross-domain limitations, adversarial objections, anti-immunization rules, evidence and replication state, maturity constraints, and research provenance. The Canonical Release Baseline Manifest records the Stage-0/Stage-1 lineage used for release preparation.