Oshidero, D. F. O. and Coley, D. A. (2026) Talking carbon: A lexical approach to predictive embodied carbon analysis via machine learning. Architectural Engineering and Design Management, 22(1), pp. 235-268. ISSN 1745-2007
Abstract
Sustainable architecture faces significant challenges, particularly during the early design stages where critical decisions often lack sufficient detail for traditional environmental analysis. This paper presents the first development and use of an artificial intelligence-based tool designed to predict embodied carbon emissions from high-level natural human language descriptions of buildings. The new approach combines a Histogram-based Gradient Boosting regression model with a multi-step system of Natural Language Processing techniques to convert complex, unstructured text into structured features and quantities suitable for predictive modelling. The work rests on a foundation of 150,000 new synthetic training samples, generated by systematically randomising building specifications. Evaluation of the method's performance was based on four strands: extraction sensitivity, relative accuracy, linguistic robustness, and usability. In tests of extraction sensitivity, the method successfully identified core structural and external elements over 80% of the time. Relative accuracy assessments with seven real-world buildings revealed a Spearman's rank correlation of 0.71, confirming the system's ability to identify differences in carbon-intensity. Linguistic robustness was proven by describing identical buildings in multiple ways, with predicted values differing by only 10%. A user study of 43 industry professionals produced a System Usability Scale score of 84.74. This reflects a strong acceptance of the method and the potential for its integration into existing workflows, emphasising its promise as an approach for advancing architectural practice. Collectively, these outcomes highlight the success of this new approach for embodied-carbon assessment and underscore the potential of AI-enabled insight in sustainable design.
| Item Type: | Article |
|---|---|
| Uncontrolled Keywords: | AI; architecture; artificial intelligence; early-stage design; embodied carbon; machine learning |
| Index terms: | accuracy, machine learning, usability, artificial intelligence, workflow, sustainable architecture, embodied carbon, sustainable design, architectural practice, environmental analysis, regression model, building specification, predictive modelling, integration, early design stage, face |
| Subjects: | organizational analysis, management, user-centered design, prediction and forecasting, design process, artificial intelligence, business, statistical analysis, professional development, environmental impact, psychology, professional practice |
| Topics: | Business Strategy, Sustainability, Design Practice, Organizational Design, Information Management, Digital Applications, Research Practice |
| Descriptive scope: | 3 PCA |
N.B. Descriptive scope is a count of how many of the five facets of empirical research are indicated by the words used in title, abstract and keywords. It is not intended as a judgement on the research; merely a count of the kind of word we would expect to indicate Phenomenon, Concepts, Theoretical framing, Empirical techniques, Analytical techniques. If all five are present, then a code of “5 PCTEA” will indicate this. If you feel the coding for this record is questionable, we welcome discussion around the terms we matched or the way we categorized them. The facet you would expect may not be coded, or a facet may be coded inappropriately. This can also bear on a larger question, of which facets should be treated as defining in construction management research. Please get in touch, and we will look at it. More details here