Document Type : Original Article
Authors
1
Department of Architecture, Faculty of Architecture and Urbanism, Ferdowsi University of Mashhad, Mashhad, Iran .
2
Faculty of Architecture, Berlin University of Applied Sciences, Berlin, Germany
10.22034/rau.2026.2073574.1266
Abstract
The integration of artificial intelligence into architectural education, particularly in Iran, constitutes a significant paradigm shift within the discipline. While this transformation is advancing rapidly globally, its adoption remains uneven, contested, and highly context-dependent. In advanced economies, AI has transitioned from experimental tools to essential elements in design studios, research laboratories, and accreditation standards. In contrast, Iran faces unique challenges, including sanctions, infrastructure limitations, outdated curricula, and cultural concerns about preserving human creativity and national architectural identity. This study aims not only to document current developments but also to conduct a comprehensive multi-level analysis of opportunities, obstacles, and prevailing attitudes. Most importantly, it proposes a localized, phased, and ethically grounded conceptual model, “AI–Education–Architecture” (AI-EDU-ARCH), designed for implementation under resource constraints while maintaining a central focus on human creativity.
A mixed-methods concurrent triangulation design was employed, with qualitative inquiry serving as the primary approach and quantitative data providing confirmation and prioritization. The study sample included 95 purposefully selected participants from architecture faculties: 15 full-time faculty members, 10 PhD candidates, 25 master’s students, and 45 undergraduates, each with at least 6 months of documented experience with AI-related tools. Data collection occurred through three complementary methods: 73 semi-structured interviews (averaging 30 minutes each) conducted both in-person and online; a 20-item researcher-developed questionnaire combining 5-point Likert scales and open-ended questions; and a systematic review of 25 high-impact peer-reviewed papers published between 2015 and 2025. Qualitative data were coded and thematically analyzed in NVivo 14 using Braun and Clarke’s six-phase framework, resulting in 150 initial codes, which were consolidated into five main themes and 18 sub-themes. Quantitative data were analyzed in SPSS using descriptive statistics, chi-square tests for inter-group differences, one-way ANOVA, and the Friedman test to rank themes by participants’ perceived importance. Inter-coder reliability was established at Cohen’s κ = 0.85, and questionnaire reliability was confirmed with Cronbach’s α = 0.87.
The five main themes identified, ranked according to participants’ responses in the Friedman test (p < 0.001, χ²(4) = 41.68), were: (1) personalized learning, (2) enhanced communication and collaboration through AI-powered tools, (3) generation of innovative and previously unattainable design solutions, (4) significant acceleration of traditionally time-consuming design processes, and (5) a complex array of ethical, operational, cultural, and existential challenges.
Personalized learning was identified as the most significant theme, with support ranging from 71% among undergraduates to 93% among faculty and PhD students. Interviewees described future studios where AI tutors continuously monitor student progress, identify individual weaknesses (such as inadequate daylighting analysis or tectonic resolution), and generate tailored remedial exercises, reading lists, precedents, or mini-briefs aligned with each student’s learning pace and aesthetic preferences. PhD candidates expressed particular enthusiasm for AI as a “24/7 research co-pilot” capable of reviewing extensive literature in multiple languages and identifying knowledge gaps relevant to dissertation topics. Faculty members viewed personalization as a means to move beyond the traditional one-size-fits-all lecture model prevalent in Iranian architectural education.
Enhanced communication was ranked second and generated strong responses from participants. Faculty and PhD students highlighted the use of VR and AR walkthroughs that enable non-expert clients, such as government officials, neighborhood residents, or heritage-conservancy boards, to experience spatial outcomes in real time. This approach can reduce costly late-stage revisions and foster public consensus. Master’s students emphasized the value of interdisciplinary synchronization, where immersive models allow architects to assess the structural or HVAC implications of design decisions immediately. However, participants also noted significant practical challenges, including outdated computer lab equipment, unreliable internet connectivity that hinders cloud collaboration, and the high cost of VR headsets in a sanctioned economy.
The third theme, the creation of innovative designs, elicited both excitement and concern. Participants demonstrated that contemporary generative tools can rapidly produce numerous massing options that optimize for seismic performance, passive cooling, daylight autonomy, material efficiency, and culturally relevant geometric patterns inspired by Persian gardens or wind catchers. Faculty members regarded this capability as a significant opportunity to address pressing challenges in Iran, such as earthquake resilience, extreme heat, water scarcity, and rapid urbanization. Conversely, many undergraduates reported feeling “overwhelmed” or “deskilled,” expressing apprehension that their role was shifting from authoring architecture to curating machine-generated output.
Acceleration of the design process, while objectively indisputable, produced the most polarized reactions. Tasks that once consumed days or weeks—energy modeling with Ladybug/Honeybee, structural pre-sizing, zoning compliance checks, or iterative form-finding—now take minutes or even seconds. Experienced users celebrated the liberation of cognitive bandwidth for higher-order conceptual and ethical concerns. Less experienced students worried that extreme speed would encourage superficial “generate-and-pick” workflows that bypass the slow, reflective maturation traditionally considered the heart of architectural thinking.
The fifth theme, ethical and operational challenges, was ranked lowest in immediate perceived importance but generated the most in-depth and philosophically complex interview responses. Faculty members expressed concerns about the risk of “training selectors instead of authors” and cautioned that uncritical adoption could reduce architecture to mere prompting for engineering. Key issues included copyright and authorship of AI-generated images, cultural bias in predominantly Western training datasets, epistemic injustice toward local Iranian-Islamic knowledge systems, privacy concerns related to student work in cloud models, the ethics of using unauthorized software under sanctions, and broader questions about whether architecture can maintain its core values if machines begin to make judgments about beauty, dignity, or social justice. PhD students further critiqued the potential for long-term dependency on foreign black-box algorithms, raising concerns about the erosion of indigenous knowledge systems.
These five themes, along with their 18 sub-themes and statistically significant differences among stakeholder groups, provided the empirical basis for the three-layered conceptual model “AI–Education–Architecture.”
The first layer, systematic introductory familiarization, is designed for undergraduate students and aims to address fear, misconceptions, and lack of knowledge from the outset. This layer includes short, intensive bootcamps, free or low-cost MOOCs (many available in Persian), hands-on laboratories using open-source or sanction-bypassable tools (such as Leonardo.Ai with a VPN, Stable Diffusion on local GPUs, and ComfyUI workflows), and exercises that teach students to interact with text-to-image and text-to-3D models. The primary pedagogical objective is to demystify AI, positioning it as a collaborator rather than an infallible authority.
The second layer, deep practical integration within digital design studios, is intended for master’s students and serves as the operational core of the model. In this phase, AI tools become essential partners in all studio projects. Grasshopper and Dynamo facilitate parametric exploration; Maket.ai, Veras, and ARCHITECTURES support rapid plan and massing generation; Cove, Sefaira, Ladybug, and other tools provide real-time environmental feedback; Autodesk Forma is used for site analysis; and BIM platforms (Revit, Speckle, Hypar) enable live interdisciplinary coordination. Instructors transition from traditional critics to “AI-literate mentors,” guiding students in prompt writing, critical evaluation of algorithmic suggestions, model retraining for cultural or contextual relevance, and assuming full moral and aesthetic responsibility for project outcomes.
The third layer, ongoing ethical, social, and research governance, is directed at faculty members and PhD candidates and serves as the ethical foundation of the system. This layer advocates for the immediate formation of departmental AI Ethics Committees, the development of binding codes for responsible use, the creation of locally fine-tuned models based on Persian architectural precedents, systematic research on mitigating cultural bias, longitudinal studies on creativity metrics before and after AI adoption, and policy advocacy for domestic GPU clusters, legal open-source repositories, and negotiated exceptions to sanctions for educational software. This governance ensures that technological integration is consistently accompanied by critical reflection.
The principal strength of the “AI–Education–Architecture” model is its three-layered integration—foundational training, professional application, and ethical-research oversight—tailored to Iran’s specific constraints, including sanctions and limited resources. This structure establishes a sustainable balance among technological innovation, human creativity, and practical implementation, while supporting ongoing evaluation and improvement.
The model integrates AI with situated learning, aims to address infrastructural gaps in Iran, and operates through three core components: educational, applied, and ethical.
Future research should examine AI’s impact on sustainable Iranian architecture, cultural challenges and resistance to change, level-specific curricula, and the long-term effects on the profession and society. The proposed model offers a clear framework for educating architects in the digital era while preserving human and cultural values.
Keywords
Subjects
