Journal of Information Technology in Construction
Journal of Information Technology in Construction
ITcon Vol. 31, pg. 1336-1356, http://www.itcon.org/2026/55
From automation to authorship: Computational reasoning and design agency in architectural AI tools
| DOI: | 10.36680/j.itcon.2026.055 | |
| submitted: | September 2025 | |
| published: | November 2026 | |
| editor(s): | Bosché F | |
| authors: | Seeja Sudhakaran
University of West London https://orcid.org/0009-0002-8575-3505 seeja.sudhakaran@uwl.ac.uk Changfeng Fu, Professor University of West London https://orcid.org/0000-0002-2019-5445 Charlie.Fu@uwl.ac.uk Efcharis Balodimou, Dr University of West London https://orcid.org/0000-0003-1249-3941 Efcharis.Balodimou@uwl.ac.uk Mukhtar Maigari, Dr University of West London mukhtarmaigari@gmail.com Prapooja KC University of West London prapooja.kc@research.uwl.ac.uk Paul Coates, Dr University of Salford https://orcid.org/0000-0002-5437-1777 S.P.Coates@salford.ac.uk | |
| summary: | The increasing presence of AI in architectural practice is renegotiating who, or what, holds creative authority over a project. This paper introduces a novel two-construct framework, AI Logic (AL) and Architectural Value (AV), to operationalise this shift. AL captures a tool's computational reasoning across automation, optimisation, and predictive capability, while AV captures design agency support, contextual sensitivity, and exploratory capacity. Positioning fifty AI-driven tools within this AL-AV construct via tension map analysis identifies stage-specific patterns within the selected dataset. It shows that design agency is redistributed unevenly across the project lifecycle, expanding in early design stages where tools act as creative collaborators, and contracting under efficiency-driven logics in later stages where automation prioritises monitoring and compliance over qualitative judgement. Cross-referencing against secondary parameters such as building typology, user group, tool maturity, and regulatory context surfaces fundamental gaps in current AI provision. The selected tools predominantly support residential and commercial applications. Documented regional orientations include North American and European standards, while coverage of other regions and typologies is less evident within the sample. To assess scoring consistency, ten AEC professionals and the author independently evaluated a subset of five tools, with one tool representing each of RIBA Stages 1–5. Intraclass correlation analysis demonstrated strong panel-level inter-rater reliability for both constructs, while comparison of the author's scores with the expert consensus indicated close alignment. Following this calibration, the author independently applied the same evaluation framework to the remaining forty-five tools. Beyond its analytical application, the framework offers a practical means of informing future AI development and adoption. It gives toolmakers a vocabulary for embedding stakeholder values in development decisions, and architects a means of recognising embedded biases before adoption, so that increasing automation extends rather than erodes the disciplinary and cultural values of architecture as a whole. | |
| keywords: | Artificial Intelligence, architectural design, design agency, computational design, project lifecycle | |
| full text: | (PDF file, 1.254 MB) | |
| citation: | Sudhakaran, S., Fu, C., Balodimou, E., Maigari, M., KC, P., & Coates, P. (2026). From automation to authorship: Computational reasoning and design agency in architectural AI tools. Journal of Information Technology in Construction (ITcon), 31, 1336-1356. https://doi.org/10.36680/j.itcon.2026.055 | |
| statistics: |