AI-Enabled Evolution of Material Intelligence: From Passive Properties to Autonomous Behaviour in Adaptive Built Environments

Terms such as smart, adaptive, intelligent and autonomous are applied to materially responsive and AI-enabled building systems despite major differences in demonstrated capability. This study develops and preliminarily evaluates an evidence-centred framework for classifying material intelligence in adaptive built environments. A design-science-supported, theory-directed synthesis used 11 OpenAlex search families, 4,102 deduplicated candidates, 66 verified priority scholarly sources and ten official sources. Source credibility, access status, evidential directness and implementation mode were recorded via a traceable source-to-criterion audit chain. The resulting framework separates functional capability (A0-A5), governance assurance (G0-G3), contextual performance and evidence maturity. Under the proposed definition, material intelligence begins at A3, where evaluated outcomes modify subsequent physical action. A4 distinguishes predictive anticipation (A4-P) from learning-enabled control (A4-L), while A5 requires sustained, bounded task or sub-goal reconfiguration under uncertainty. Evidence is coded E/P/A/NR and qualified as operational, testbed, modelled or proposed. Application to six purposively selected cases produced A1/G0, A2/G0, provisional A2/G1, A0 as tested/G0, modelled A4-P/G0 and modelled A4-L/G0. No case verified operational A3-A5 or G2-G3. The framework therefore offers falsifiable preliminary discrimination, not a validated maturity scale, and identifies the field evidence needed for independent validation. Digital Twin Technology in Engineering Practice: Insights from Academic and Industry Professionals in an African Developing Country Emmanuel Olorunyomi Aremu1*, Jesuferanmi Adetayo1, Ogagaoghene Odokuma1,