A Validation Study of a Generative AI-Based Educational Process Model for Spatial Design
PLUS URBAN DESIGN ASSOCIATION · 30.04.2026 · 8 (1) · 93-106
DOI: 10.69913/udj.2026.04.8.1.93 ↗Kurz gefasst
Fachartikel in der Zeitschrift PLUS URBAN DESIGN ASSOCIATION, online erschienen am 30.04.2026. Verfasst von Young-Ju Lee.
Abstract (Originalsprache)
(Background and Purpose) As Generative Artificial Intelligence(GenAI) spreads across the practical field of spatial design, it is necessary to have a AI-integrated educational model within a design process framework. The researcher proposed a hypothetical GenAI-based educational process model for spatial design, using the Double Diamond model as a theoretical framework in the previous study, and then its educational feasibility was qualitatively examined through a pilot program in a university studio. As a follow-up research, the present study aimed to quantitatively verify the educational effectiveness of the model through learner surveys, in-depth interviews, and expert evaluations, and then empirically confirm its objective validity. (Method) A 15-week curriculum applying the model was implemented for 41 juniors and seniors(17 teams) enrolled in a Spatial Design studio course at a four-year university in Korea. The tools used were limited to text-based GenAI(ChatGPT) and image-based GenAI(Midjourney). To verify the educational effectiveness, 8 essential design competencies of NASAD were used as a framework, which were linked to the model's step-by-step process to develop and implement self-report metrics for learners and design competency evaluation metrics for experts. The collected data were analyzed using reliability analysis, frequency analysis, descriptive statistics, and one-sample t-tests, which was qualitatively supplemented by one-on-one in-depth interviews with learners. (Results) The learner survey results showed that all 8 competencies significantly exceeded the neutral midpoint, with an overall average score of M=4.306(SD=0.791), demonstrating a high level of awareness. By competency, Space Planning & Design scored highest(M=4.671), while Regulatory & System-Compliant Architectural Design scored lowest(M=3.939). In the expert evaluation, all items were rated significantly high, and the inter-expert reliability using ICC(2,k) yielded a value of 0.893, indicating good consistency. The analysis results showed that the most effective tools in each phase were as follows: text-based AI for information search and structuring in the Discover phase; image-based AI for visualization of colors and spaces in the Define phase; design refinement through iterative alternative generation in the Develop phase; and structuring design logic and enhancing representation quality in the Deliver phase. On the other hand, in areas requiring expert judgment reflecting actual spatial context, such as architectural and regulatory applications and ergonomics-based design, the effectiveness of AI utilization was found to be relatively limited. (Conclusions) This study empirically confirmed that the GenAI-based educational process model for spatial design serves as a valid educational framework for enhancing learners' awareness of design competency and the quality of deliverables in actual studio classes. GenAI is not a tool that replaces the design process, but rather a supplementary medium that supports learners' design thinking, demonstrating significant educational effectiveness in the areas of information search, visualization, alternative generation, and representation. Future research requires further refinement of AI utilization strategies in the legal and technical domains, and structural advancement of the educational model through in-depth analysis of curriculum implementation.
Zitieren
APA
Lee, Y. (2026). A Validation Study of a Generative AI-Based Educational Process Model for Spatial Design. PLUS URBAN DESIGN ASSOCIATION, 8(1), 93-106. https://doi.org/10.69913/udj.2026.04.8.1.93
Vancouver
Lee Y. A Validation Study of a Generative AI-Based Educational Process Model for Spatial Design. PLUS URBAN DESIGN ASSOCIATION. 2026;8(1):93-106. doi:10.69913/udj.2026.04.8.1.93
BibTeX
@article{Lee2026,
author = {Lee, Young-Ju},
title = {A Validation Study of a Generative AI-Based Educational Process Model for Spatial Design},
journal = {PLUS URBAN DESIGN ASSOCIATION},
year = {2026},
volume = {8},
number = {1},
pages = {93--106},
doi = {10.69913/udj.2026.04.8.1.93}
}
EUVIMED