Next Article in Journal
Enhancing High-Strength Lightweight Cement Composites with Hollow Glass Microspheres for Advanced Construction Applications
Next Article in Special Issue
Wi-FAB: An Applied Educational Workflow for Prototyping Discrete Components with Planar-Joint Assemblies Through Creative Robotics
Previous Article in Journal
Option-C Verified Semantic Digital Twins for Decarbonized, Pressure-Reliable Central Business District Hospitals
Previous Article in Special Issue
Coupling of Pawnshop Building Distribution and Urban Spatial Structure in Macau via GIS and Space Syntax Analysis
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Buildings
Volume 16
Issue 6
10.3390/buildings16061097
buildings-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Human–AI Collaboration in Architectural Design Education: Towards a Conceptual Framework

by
Şerife Hikmet
* and
Nazife Ozay
Faculty of Architecture, Eastern Mediterranean University, North Cyprus, Famagusta 99628, Türkiye
*
Author to whom correspondence should be addressed.
Buildings 2026, 16(6), 1097; https://doi.org/10.3390/buildings16061097
Submission received: 2 February 2026 / Revised: 28 February 2026 / Accepted: 5 March 2026 / Published: 10 March 2026
(This article belongs to the Special Issue Emerging Trends in Architecture, Urbanization, and Design)
Browse Figures
Versions Notes

Abstract

Rapid developments in artificial intelligence (AI) have prompted increasing attention to human–AI collaboration across various fields. This study focuses on the architectural design process and examines collaboration with generative AI (GenAI) within the context of architectural education. Creative cognition in design-based learning and the process of collaboration with AI are crucial. Insights on AI usage and design process perceptions are gathered from semi-structured interviews of architecture students. The data were analyzed using primarily inductive thematic analysis to understand their experiences in architectural design education. It aims to construct a conceptual framework to interpret the creative cognition during human–AI collaboration in the architectural design process through algorithmic thinking strategies and existing theories. The literature review acted as the foundation for the theoretical background, adopting existing models and theoretical perspectives to support the conceptual framework generation. The study contributes to human–AI collaboration in architectural design contexts. Additionally, the conceptual framework proposal derived from the empirical insights and relevant literature can serve as the basis for conducting further explorations of a potential model in architectural design education.

1. Introduction

Rapid technological developments have reshaped professional practices and educational approaches. While traditional teaching methods remain valuable, evolving professional demands require new forms of adaptation. Artificial intelligence (AI) has broadened many subjects due to its foreseen applications as indicated by various “academic papers and practical applications” [1] (p. 2).
Guo et al. [2] mentioned the beneficial effects of AI on creativity; however, educational contexts still experience difficulties in terms of the uniform state of creative output, insufficient innovations, and overdependence of students on AI. Therefore, the human–AI collaboration in the educational context becomes critical. The notion of collaboration should be treated as a pedagogical approach [3]. AI adoption has been addressed in studies alongside other disciplines, such as “Embracing Artificial Intelligence (AI) in Architectural Education: A Step Towards Sustainable Practice?” by Komatina et al. [4] which examines AI in architecture from the viewpoint of sustainability. The role of AI in architectural design constitutes another distinct area, in which both the potential challenges and opportunities have been discussed by Zhang et al. [5]. Co-design is considered a form of co-creation involving both professional designers and others without formal [6] training. This understanding is parallel with the logic of collaboration between AI and the designer, where AI as a tool can be the co-design partner.
Alexander et al. [7], in their seminal work titled “A Pattern Language”, reflects that design can be broken into parts, units that are repeatable as patterns. This notion is parallel with the architectural design solution generation process and the way AI identifies and forms structures in data. Algorithmic thinking (AT) is a method where the problem can be divided into smaller segments [8,9], which presents the cognitive problem-solving methods by identification of subcomponents [10]. Integrating new approaches to sustain the relationship between the basic learning attributes and emerging paradigms becomes significant.
Therefore, the students’ AI usage perceptions during an architectural design process are analyzed through a set of questions. Qualitative research was adopted by semi-structured interviews with 40 students to gather relevant insights to benefit the framework generation, in addition to the related existing models. The “reflection-in-action” and “reflection-on-action” concepts from Schön’s [11] model were adapted into a new pedagogical framework that aims to elaborate the integrated stages of co-design with prompt-based GenAI tools. A perception-informed conceptual framework for structuring GenAI-supported co-creative design processes in architectural design education grounded in creative cognition and algorithmic thinking is the contribution.
Prior studies have focused on creativity, creative problem solving, AI tool usage and their shared dimensions across these domains. This study seeks to interpret how Gen AI tools can be positioned within creative design processes in educational contexts. Accordingly, it addresses a main research question: “How can a framework conceptualize the integration of GenAI tools in the creative design process?” By integrating relevant theoretical perspectives into a unified structure, the framework offers a more holistic interpretation of AI integration in architectural design education.

2. Literature Review

2.1. An Overview of Design Process and Creativity from an Architectural Perspective

Cross [12] suggests that the design process unfolds through actions that occur between the solutions and problems. Generally, problem solution is the basis of the design process [13,14]. Moreover, an interesting aspect of design protocols is the mutual relation of verbal–conceptual and visual–graphics information [13].
Designers form a new space configuration by mental image and analogies [15]. This reveals that architectural design is not only comprising technical knowledge, but also creative cognitive actions. The architectural design process is a thinking activity with a series of stages [16]. Creative production occurs through mental processes and is initiated with problem definition (preparation), unconscious information understanding (incubation), idea generation (illumination) and idea testing (verification) [17,18]. Creativity is a complex field; Guilford [19] highlights the difficulties of studying creativity, as it involves an exchange between conscious and unconscious processes. Smith and Beda [20] refer to the unconscious work as an assembly way to find a creative idea; however, if the action was purely random or unconscious, the idea generated by the student would not leave a space for the successful design development [21].

2.2. Current Role of AI in Architecture and Architectural Education

AI systems rely on problem-solving methods and logical algorithms [22]. In architecture, tasks are undertaken by architects to develop functional and aesthetic design solutions [23]. In this respect, AI systems can generate alternatives to assist in evaluating specific parameters, reflecting a rational orientation at a computational level. Within the design field, AI emerged. One perspective anticipates AI functioning in ways similar to human intelligence, particularly when integrated with tasks traditionally associated with human cognitive abilities. Another perspective frames AI as a parallel entity, while acknowledging its distinct limitations in comparison to human intelligence [24]. The current role of AI is increasingly visible across various domains, and its continued development is likely to influence professional practices. A comparable technological shift occurred in the mid-1980s, when graphics and CAD programs became accessible to architects and engineers. By the early 1990s, these tools were no longer confined to heavy and expensive mainframes but were available on mini-computer predecessors. This shift ignited the start of the digital age in design [25].
Architectural education is envisioned to follow the “present and future challenges”; the profession will be transformed as a consequence of technological developments [24] (pp. 1261–1264). Artificial intelligence is becoming an essential part of architectural education through exceeding the limits and shifting the traditional design methodologies [5].
The rise in AI technology highlights to reconsider the instructors’ role and pedagogy. AI is rapidly advancing and affecting higher education services [26]. Studies on AI and architecture reflect the challenges of students and the ethical sides of human–AI collaboration [27,28]. The skills acquired by the students become insufficient when they graduate, as they were once exceptional, but are turning out to be ordinary because of the proficiency in computing, automation, and generative skills in software and hardware. The students who were born after the 1980s are categorized as “digital natives” and having digital acquaintance with the technology had an impact on their “learning preferences” [29] (pp. 162–166), which underlines the need for necessary adaptations.

2.3. Design Thinking vs. Algorithmic Thinking: Cognitive Aspects of Human–AI Collaboration in Architectural Design Education

In the 1960s, the design framework gradually started to be established [30,31]. Certain methods were applied, and among those, they were organized as input and output. The input serves as an acknowledgement, while the output reflects the will to know. The system is supported by brainstorming, analogy, and attributes in order to open a path and overcome mental blockages. Those act as the basis of design thinking (DT), which is crucial for the generation of ideas and identification of the ill-defined problems [32]. Problem-solving is a part of design thinking skill [33]. It is not only part of DT, but AT also encompasses problem-solving.
Alexander et al. [7] brought attention to the algorithmic nature of design patterns. An algorithm is the process of having determinate stages that refer to a problem. Algorithms are fundamental to structure, illustrating a step-by-step method to solve a problem or to carry out tasks [23,34]. AT has computational thinking skills such as abstraction, decomposition, pattern recognition, and problem generalization [35]. Terzidis [23] broadened the subject by mentioning AT and creativity. It created paths for designers to discover the systematic and generative approaches. Ürey [36] studied the effects of the heuristic and algorithmic educational methods in basic design education to observe the impact on the development of the students’ creative cognition. AT methods are implemented in architectural design studios, computational methods, including algorithms also adopted AT [36,37]. Studies highlight algorithmic practices and their significance for design students in terms of computational competency [38,39,40]. AI changes the manner of thinking, acting and interacting [41]. Accordingly, this study reflects approaches to systematizing the architectural design process with AT methods.
AI is recommended as a support for the process rather than having it for the finalization of the project. There is a potential for AI to weaken human creativity. On the other hand, AI can be taken into consideration as a means of “cognitive augmentation” [42] (p. 59). As a GenAI tool, Midjourney’s main function is to generate visual products from text input. It has been found that the visuals are benefiting the conceptual design stage as plan schemes, there are architectural collages and presentations [43]. Prompts in text format are excessively dependent on text-to-image generative AI tools [44].
During the prompting process, cognitive actions take place; it is complex since it requires written language. Written language production involves cognitively planning processes [45]. Design patterns allow us to perceive the nature of design solution from the level of abstraction [46]. Consequently, it can be mentioned as DT and AT during the prompting.

3. Materials and Methods

A qualitative research design was adopted, employing semi-structured interviews with architecture students from different design studios. The data were analyzed using a primarily inductive thematic analysis to explore students’ perceptions of AI use and related processes. The emphasis of the analysis remained on interpretive meaning rather than numerical distribution. The framework was developed based on the themes and was contextualized with reference to the relevant literature in the field. Model of Schön [11] and Geneplore Model [15] formed the basis of the theoretical background of the conceptual framework formation. The interpretation of cognitive processes can be framed through the protocol analysis method by Ericsson and Simon [47]. The Geneplore model by Finke et al. [15] acts as a part of the proposed model, which sheds light on the creative cognition aspect of the architectural design process in collaboration with GenAI.
The co-design phase of this process consists of prompt generation and output evaluation, which aligns with the “generate” and “explore” parts of the model. Additionally, protocol analysis [47] helps to understand the ideation stage during the design process. Protocol analysis allows verbal feedback to understand the cognitive aspect of human–AI collaboration in architectural design.
The interview was conducted among the architecture students in the 3rd and 4th year design studios at the Eastern Mediterranean University, Department of Architecture. The reason behind the selection of those studios is that they are at the intermediate phase, starting from Arch391. As students progress into advanced design studios, they approach becoming prospective professional architects. Thus, they are expected to have an architectural understanding and should have experience with computational tools, showing digital competency. The studios that are involved in the study are Arch391 (Architectural Design Studio—III), Arch392 (Architectural Design Studio—IV), Arch491 (Architectural Design Studio—V) and Arch492 (Architecture Graduation Project). Figure 1 presents the research methodology.

3.1. Data Collection

Four focus groups were formed to enhance variation across participants and reduce the risk of group-specific bias. Ten students from each design studio were selected, resulting in a total of 40 architecture students participating in the study (20 female and 20 male). The study’s ethics statement was approved by Eastern Mediterranean University’s Architecture, Planning and Design Ethics Sub-Committee (No: ETK00-2024-0047, 11 March 2024). Data were collected through semi-structured, face-to-face interviews conducted within the focus group setting. Prior to the interviews, written informed consent was obtained from all participants. The interviews occurred at the Department of Architecture, Eastern Mediterranean University (EMU), and each interview lasted approximately 10–15 min. Participants’ responses were documented in written form during the interviews. To enhance data fidelity, participants clarified and confirmed their responses in real time. The interview protocol consisted of 14 questions divided into two sections. The first section included demographic questions. The second focused on participants’ preferences regarding the use of artificial intelligence in architectural design, including areas of application and future expectations (see Appendix A). The first set of questions was closed-ended, whereas the remaining questions were open-ended. The questions were not directly targeting GenAI employment, enabling the students to openly reflect on their experiences without any limitation.

3.2. Data Analysis

Thematic analysis was employed to explore patterns of meaning across the data. Following the approach outlined [48,49], themes were actively developed through an iterative and interpretive process rather than merely summarizing responses. Codes were generated to capture the underlying meaning of participants’ answers; therefore, some codes do not replicate participants’ exact wording but reflect the essence of their statements.
The analysis was conducted using MAXQDA, a qualitative data analysis tool. A primarily inductive strategy was adopted. Initial open coding was applied to the responses to identify meaningful units, which were then grouped into broader categories through selective coding. These categories were refined into sub-themes and overarching themes through iterative comparison across the data. Coding was carried out focusing on participants’ explicit expressions rather than latent interpretation. Initial codes were generated directly from the participant-validated textual records and subsequently organized into sub-themes and broader themes through iterative review and comparison. Traceability was maintained by systematically aligning data extracts with their corresponding codes and thematic categories. Selected excerpts are presented in the findings section to demonstrate the progression from raw data to thematic structure. Table 1 includes examples of the coding progression.
The first section of the interview (see Appendix A) served as a warm-up to prepare participants for the subsequent questions. Beginning with the fifth question, responses were systematically analyzed to construct main themes, related sub-themes and associated codes. In total, ten main themes were identified (see Appendix A, Table A2). Coding was conducted by the first author. Emerging interpretations were discussed with the co-author to strengthen analytical rigor.

4. Results

The first set of questions included nationality and age; the related patterns with the demographic information were not found. Participant counts (n) refer to the number of students expressing each theme and are presented for contextual clarity.
The second part of the survey began with a question on computer program usage, positioning students within a broader technological context before introducing AI-related topics. Responses were coded under the theme of computer programs used by students. AutoCAD appeared as the most frequently mentioned tool (n = 37). In addition, “site analysis” and “synthesis” were identified as stages in Arch391 and Arch491 design studios, where computational tools were particularly emphasized.
The seventh question introduced AI by asking whether participants had used it for architectural design purposes. While some reported prior use, this was not always within the context of their studio projects. The theme of AI usage revealed that applications extended beyond text-to-image generation. For example, one interviewee noted, “Yes, ChatGPT for some design ideas.” (P 35), indicating conceptual support.
At the same time, opposing perspectives were evident. Several participants stated that they had not used AI; “No, I didn’t use it.” (P34), “No, because like to do my research and do sketches.” (P38) emphasizing a preference for independent research and sketching. Others mentioned contextual limitations or low accuracy: “I’ve tried it and I didn’t use it as it does not take into the consideration the context of the site.” (P40). “I know the tool but due to its low accuracy I prefer not to use”. (P10). Conversely, AI was described as useful for “analysis and rendering.” (P22).
Participants generally reported learning about AI tools through social media or peers, with Instagram (n = 20) being the most frequently mentioned source. Broader reflections on AI usage highlighted its role in conceptual development, proposal preparation, and visualization. As noted by P15, “It was helpful especially in the visualization process of the project”, while P36 referred to its use as this: “For starting the proposal part and rendering of the plan. It’s easy for me and clear.” (P36).
The theme of advantages and disadvantages included the sub-theme “Observed benefits”, where time efficiency emerged as a remarkable advantage. A participant stated AI’s advantage as, “… is a great tool to assist and fasten up the work, and helps to expand broader the creativity.” (P17). Others acknowledge aesthetic enhancement but questioned its comprehensiveness in addressing contextual and analytical aspects of design: “I think the only advantage is the aesthetic look other than that AI cannot do the whole process that architect can do including site visit, site analysis then solving the problems and designing accordingly.” (P29). The primary concerned disadvantage mainly observed in output accuracy, reinforcing the perceived need for guidance in effective AI integration.
With the subsequent question (see Appendix A), the participants were asked to identify the design phases in which they used AI. The coded responses clustered around the sub-themes “Conceptual phase”, “Development”, “Presentation” and “Analysis”. These were associated with rendering, visual presentation and early-stage idea generation. For instance, one participant stated, “For me, I feel it is going to be beneficial only in the 3D rendering or even the 2D it’s just for finishing the project.” (P29). Another explained its usefulness as “Concept generation process and post production as it consumes less time and is able to generate ideas.” (P33).
The theme concerning AI’s supportive role received predominantly positive responses (n = 29). The interviewees emphasized speed, efficiency, and competitive advantage. For instance, “It will support our designer role with AI, we can create fast, and we can get fast solutions.” (P7). P17 suggested that architects using AI as an assistant tool may be “… one step further from others ...”. However, a small group expressed reservations (n = 5), and responses reflected uncertainty. P3 noted, “It has a mind of its own so it can overcome the project and take away the sense of accomplishment in design.”, indicating concerns.
The theme “Usage of AI for the future projects” comprised three sub-themes: “AI will be used”, “AI will not be used”, and “Not decided yet.” The majority (n = 31) indicated an intention to continue using AI, with some expressing strong commitment, such as planning further specialization in AI: “Yes, absolutely. I am even planning to take an AI master due to its beneficial effects.” (P16). On the other hand, a minority rejected future use, citing concerns about creative autonomy and reliance on database-driven systems (P24).
Regarding instructional expectations, the theme “Expectations from instructors for the use of AI tools” revealed a predominantly supportive stance toward guided integration. A small number of students (n = 7) favored restriction, especially in foundational design courses; most participants advocated structured guidance rather than prohibition. A participant noted, “I believe the instructors could find the right way in which it could be used and guide us through that.” (P3).
Another theme addressed perceptions of the architect’s future role. Although many participants anticipated continuity, some suggested potential shifts toward supervisory or technologically integrated roles. For instance, “Architects will still be needed but may be fewer; it is also possible they may move to taking charge of such computers (help program them)” (P3). Others emphasized the enduring importance of human character, contextual awareness and creative agency in design (P1, P9).
Finally, participants were asked whether they felt concerned about becoming architects in the age of AI. Responses were grouped under “Feeling worried” and “Not being anxious.” The majority (n = 30) reported no significant concern, generally framing AI as a supportive tool. The majority do not worry, since most of them think of AI as a tool. A smaller group expressed anxiety, often linked to perceived gaps in formal AI education. As one student stated, “Yes, I don’t think the university equipped us with enough knowledge about AI and technology so yes I am worried but I have to learn it on my own.” (P40).
Collectively, these findings suggest that students predominantly perceive AI as a collaborative tool rather than a replacement, a perspective that informed the development of the proposed framework.

5. Discussion

Interpretation of the Findings

This section interprets the findings in relation to existing literature and broader discussions on architectural education and professional practice. The question addressing expectations of students from their instructors points to a perceived need for adjustments in pedagogical approaches. Participants reported adopting AI tools primarily during the synthesis phase of the design process, particularly for ideation purposes. In this context, ideation was associated with generating design concepts [50]. AI tools were considered beneficial in supporting this process, especially in terms of time management. Although many students expressed positive attitudes toward AI, several indicated that the outputs did not fully meet their expectations. A recurring issue concerned limited knowledge about how to use these effectively, suggesting that guidance and structured instruction may enhance their application. This issue can be tried to resolve by adapting the Geneplore model via its generate and explore features.
Most participants did not perceive AI as a direct threat to their future professional roles, regardless of studio year. This openness may be discussed in relation to arguments in the literature suggesting that architects engage constructively with computational tools rather than resisting them [25]. While participants generally believed that the role of the architect would remain stable, broader discussions in the field suggest that architectural practice continues to evolve in response to technological, technical, and socio-economic developments. Increasing complexity in design processes has often been associated with the concept of “wicked problems”. The participants did not explicitly frame their views in these terms. Their openness to AI-assisted processes may be discussed in relation to ongoing debates about how emerging technologies can support decision-making in complex design environments. However, the present findings reflect students’ perceptions rather than empirically demonstrating such professional transformation. In this respect, AI may be considered as a supplementary tool that supports decision-making processes rather than replacing professional judgment. Recent studies have also begun to explore the integration of tools such as ChatGPT within the Architecture, Engineering and Construction (AEC) field, including areas such as construction management, indicating ongoing experimentation with emerging technologies [51]. These interpretations are limited to students’ perceptions and do not constitute proof of structural–professional transformation.
Although the participants usually revealed the advantageous views, these disadvantageous opinions highlighted concerns related to the inappropriate use of AI tools. Participants referred to the need for instructor support and the issue of inaccurate outputs. These findings suggest the relevance of reconsidering how AI integration is addressed within architectural design education, particularly in terms of providing clearer guidance for a more structured framework. A conceptual framework was developed based on the foundational model of Donald A. Schön [11]. The students are expected to input effective prompts at a GenAI by aiming for a design solution that is aimed to be explored. Integrating protocol analysis [47] and the Geneplore model for reflecting the thoughts by verbalizing them to organize the cognitive aspect [15].

6. Conceptual Framework Formation

Başarır [24] mentioned that the integration of AI into the architectural design curriculum has the potential to make the designers more aware of architectural design matters. Architecture as a profession will be transformed as a consequence of technological developments. Moreover, the virtual and physical mediums in the design education field should be in a manner to enhance the metacognitive skills of the students [52]. Empirical findings revealed areas of difficulty and expectation among the students. These observations were interpreted through selected theoretical lenses, which informed the structure of the proposed framework. Thus, the framework should be understood as theoretically led and empirically contextualized rather than purely inductively derived.
The framework includes architecture students who are required to find a solution to an architectural design problem, with a focus on text-to-image GenAI. Prompt writing is related to the cognitive dimension, as writing activity includes cognitive aspects [45]. Prompt engineering strategies are an effective way to communicate with large language models (LLM). There are a few techniques for those, including zero-shot, few-shot and chain-of-thought (CoT) prompting. Those strategies are specific to LLMs [53,54]. However, those methods can be applied to text-image-generation models in an indirect manner. The discussed issues underscore the necessity of an educational formation for AI that students and instructors can benefit from at the same time. The proposed framework is primarily explanatory in nature, yet it also introduces a staged structure that may serve as a conceptual roadmap. It does not prescribe fixed implementation steps; it suggests possible progressions and alignments among key dimensions of AI integration. Figure 2 presents a module of the framework consisting of an internal structural layer. Particular phases might show iteration due to the nature of the progress. By structuring theoretical constructs in relation to observed student challenges, the framework moves beyond a purely descriptive account and offers directionally informed guidance for pedagogical reflection and future inquiry. The module occurs within the framework, which includes stages such as “Design Plan”, “Implementation with GenAI” and “Evaluation” as shown in Figure 3. “Implementation with GenAI” consists of the internal structural layer within itself.
The proposed framework is theoretically grounded in established models of creative cognition and education. While the conceptual foundation derives from prior theory, the students’ reported challenges, their AI usage informed how these theoretical components were selected, emphasized, and structured. In particular, students’ perceptions in the early stages of design directed analytical attention toward the monitoring of initial design phases and creative cognition processes. Due to these aspects, the Geneplore model was adopted, which allows interpreting the generative and exploratory phases during the architectural design. It constitutes a theoretically anchored structure that was refined and contextualized through empirical insight. The Geneplore Model by Finke et al. [15] is a framework of cognition that presents the creative process consisting of two stages: generation and exploration. The generative phase can be stated as raw solutions, undetermined thoughts and mental images. On the other hand, the exploratory part is evaluated, developed and functional for the solution generation. In the Geneplore model, mental synthesis and transformation are regarded as key generative mechanisms in the creative process. In contrast, image scanning is categorized as an attribute-discovery activity and is understood as an exploratory process, playing a significant role in identifying unforeseen or emergent features within a visual representation [15]. From this perspective, the AT strategy supported by prompting and conceptual generation can be situated within the Geneplore model for the interpretation of creative cognition. Moreover, the protocol analysis is employed to trace and examine the creative design processes as it will transfer their idea by thinking out loud [47] (see Figure 4). The overall structure of the proposed approach is grounded in Schön’s model, as summarized in Table 2. The proposal adopted its key features from Schön’s [11] reflection-in-action and reflection-on-action model, which served as the basis for forming the general outline, delineating “reflection-in” as occurring during and “reflection-on” as occurring at a later stage. Table 2 shows the degree of relation of the concepts with existing models.
Novel outcomes result from small persistent steps, which demystify sudden insights [55]. From this understanding, it has AT characteristics; problem identification and generation of small steps towards a big problem emphasize the decomposition part of it. The structured progression of stages is “Define Problem”, “Decompose”, “Step Plan Provided” and “Prompt”, taking their basis from AT. AT elaborated as a strategy of having defined steps to decompose a complicated problem [35]. Human–AI problem solution process is a complex task [56], the AT has the potential to conceptualize the process of human–AI collaboration.
For the student to be able to segregate the design solution into stages, which has the potential of a smooth transition to the verbal part, is noted as “Prompt”. GenAI provides a basis for potential testing and output generation. The AT approaches were employed at different phases to interpret the text input as a part of the architectural design process, which involves a co-creative medium with GenAI. In addition, the output evaluation has the potential to support the ideation stage. They may adopt the idea for the design problem solution. Furthermore, the consensual assessment technique (CAT) [57] can be considered for the creativity measure of the output.
At first, according to the design problem, the students can generate their ideas together with GenAI. Figure 3 presents the framework, in which each phase and step is articulated through concise descriptions. Figure 5 illustrates the structure, which is organized around three phases “Input”, “Process”, “Output”. The process phase integrates generation and exploration, leading to the production of outputs. The bidirectional arrows illustrated in the figure represent iterative feedback loops between phases, indicating that the process allows revision and refinement rather than following a strictly linear progression. In this sense, the framework reflects a dynamic interaction among phases.

7. Limitations and Suggestions for Future Studies

The empirical part of the research was conducted by a limited number of people. Although theoretically grounded, the framework reflects insights derived from a single institutional context. Therefore, its applicability to other educational environments should be approached with caution and may require contextual adaptation. Therefore, this issue can limit the generalization of the findings. However, as the study aims to propose a conceptual framework partially grounded in students’ experiences, this limitation primarily may affect transferability rather than the depth of contextual insight.
The proposed framework, developed and supported through the relevant literature and by the responses of the semi-structured interview, provided an insight into architectural design education, especially from the perspective of the design studios. In spite of this, the framework can be improved into a pedagogical model; however, this needs to be tested on the students to have further data regarding the topic.
In the future, the proposed framework can be applied to focus groups. This has the possibility to contribute, in terms of the potential model’s effectiveness and applicability. Moreover, the framework proposal can be applied in other design project-based fields to assess its effectiveness.

8. Conclusions

Interpreting creative cognition with AT strategies during the architectural design process, and involving GenAI collaboration with a holistic approach, is the main focus of this study. The analysis of the semi-structured interviews indicated that most of the students are familiar with AI technologies. In addition, the challenges they encountered in using GenAI, together with the aspects they perceived as beneficial, informed considerations for the development of the proposed framework. The relevant theoretical background provided conceptual support for this development. The information collected provided insights that can inform the development of a conceptual proposal, which may serve as a foundation for a future pedagogical model supporting the collaborative use of GenAI tools. AT helps systematize the process with text-to-image GenAI in order to resolve the creative cognitive awareness of students during the human–AI collaborative process. The AI-generated content may create a space for reflection in which students can assess the potential of the produced outputs, which provides a co-creative design context. The study encourages further research to develop the framework towards a model for human–AI collaboration.

Author Contributions

Conceptualization, Ş.H. and N.O.; methodology, Ş.H. and N.O.; validation, Ş.H. and N.O.; formal analysis, Ş.H. and N.O.; data curation, Ş.H. and N.O.; writing—original draft preparation, Ş.H.; writing—review and editing, Ş.H. and N.O.; visualization, Ş.H.; supervision, N.O.; project administration, Ş.H. and N.O. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

The study’s ethics statement approved by Eastern Mediterranean University’s Architecture, Planning and Design Ethics Sub-Committee (No: ETK00-2024-0047, 11 March 2024).

Informed Consent Statement

Informed consent was obtained from all participants prior to their inclusion in the study.

Data Availability Statement

Due to ethical considerations and participant privacy, the data generated in this study cannot be shared.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
AIArtificial Intelligence
GenAIGenerative Artificial Intelligence
DTDesign Thinking
ATAlgorithmic Thinking

Appendix A

Table A1. Interview questions used in the study.
Table A1. Interview questions used in the study.
Interview Questions
First Section
(1)
What is your gender?
■
Male
■
Female
■
Do not prefer to specify
(2)
What is your age?
(3)
Where are you from?
(4)
Which architectural design studio are you studying?
■
Arch391—Architectural Design Studio—III
■
Arch392—Architectural Design Studio—IV
■
Arch491—Architectural Design Studio—V
■
Arch492—Architecture Graduation Project
Second Section
(5)
What are the computer programs that you use for the architectural design?
(6)
Generally, at which stage of the design process do you prefer to use the computational design tools?
(7)
(a) Have you used AI tools/technology before for your architectural design process? If yes, please specify.
(b) If yes, which platforms made you aware about the presence of such AI tools?
(8)
What stages of design do you think AI would be beneficial for your design project? If you used it before, please specify according to your experience.
(9)
Are there any advantages or disadvantages of using AI in your design projects?
(10)
(a) What stages of design do you think AI would be specifically beneficial for your design project?
(b) How do you feel about using AI to support your designer role?
(11)
Are you planning to use AI for your future projects?
(12)
What is your expectation from the instructors related to the use of AI?
(13)
What will be the role of architect in the future?
(14)
Are you worried of becoming an architect in the era of AI?
Table A2. The content derived from the semi-structured interviews.
Table A2. The content derived from the semi-structured interviews.
Main ThemeSub-ThemeCode
Computer program usagePrograms for rendering, drawing and modelingAutoCAD
Lumion
Revit
SketchUp
Rhino
Photoshop
Twinmotion
Other computer programs
Computational tool usage in designStage of design problem having computational tool adoptionSite analysis
Proposal/Initial idea
Concept
Site synthesis
Usage of AISpecifying if ever used AI tools in projectEncountered, tried
Haven’t tried, no
Mediums that enabled the AI tool to be knownThe platforms which give informationSocial media platforms
By personFriends
Architectural magazinesBlogs, architectural websites
Out of context(Only mentioned in text)
Design stages that have benefit of AI usageStages having beneficial usageConcept
Proposal/Initial idea
Site analysis
Advantages and disadvantages of AI usage in design projectsObserved drawbacksFeeling of dependence
Inhibition of creativity
Not accurate
Observed benefitsTime management, helpful
Design stages to have the usage of AIConceptual phaseIdeation, concept
Development3D model, visuals/façade
PresentationRender, visual presentation
AnalysisSite analysis, Case study
Support for the designer rolePositive supportSupport
Negative supportWill not support
Out of context(Only mentioned in text)
Usage of AI for the future projectsAI will be usedFuture projects will benefit from AI
AI will not be usedFuture projects will not benefit from AI
Not decided yetNot sure and do not know whether to use AI
Expectations from instructors for the use of AI toolsRestrictive mannerShould be restricted, Rejected
Supportive mannerAcceptance, Openness
Extra ViewsFeeling of guidance and not knowing (Only mentioned in text)
Role of architect in the futureRemain similarSame, in need
AlterationChange, differ, evolve,
UncertainDo not know, not sure
The status of being worried as an architect in the era of AIFeeling worriedFelt afraid, worried
Not being anxiousThinking AI as tool, no worries

References

  1. Hunde, B.R.; Woldeyohannes, A.D. Future prospects of computer-aided design (CAD)—A review from the perspective of artificial intelligence (AI), extended reality, and 3D printing. Results Eng. 2022, 14, 100478. [Google Scholar] [CrossRef] [Scilit]
  2. Guo, W.; Liang, Z.; Wang, C.; Li, X.; Hu, H.; Chen, S.; Yu, Q.; Zhao, Q. Student-AI collaborative creative Problem-Solving: The role of human agency. Comput. Educ. 2025, 239, 105433. [Google Scholar] [CrossRef] [Scilit]
  3. Joseph, A.; Kroll, D.; Srivastava, A.; Sharifi, E.; Datey, A. Collaboration in architectural design studio pedagogy: A literature review. CoDesign 2025, 1–27. [Google Scholar] [CrossRef] [Scilit]
  4. Komatina, D.; Miletić, M.; Ružičić, M.M. Embracing Artificial Intelligence (AI) in Architectural Education: A Step towards Sustainable Practice? Buildings 2024, 14, 2578. [Google Scholar] [CrossRef] [Scilit]
  5. Zhang, Z.; Fort, J.M.; Mateu, L.G. Exploring the Potential of Artificial intelligence as a tool for Architectural Design: A Perception Study using Gaudí’sWorks. Buildings 2023, 13, 1863. [Google Scholar] [CrossRef] [Scilit]
  6. Sanders, E.B.N.; Stappers, P.J. Co-creation and the new landscapes of design. CoDesign 2008, 4, 5–18. [Google Scholar] [CrossRef] [Scilit]
  7. Alexander, C.; Ishikawa, S.; Silverstein, M.; Jacobson, M.; Fiksdahl-King, I.; Angel, S. A Pattern Language: Towns, Buildings, Construction; Oxford University Press: New Yok, NU, USA, 1977. [Google Scholar]
  8. Sarı, U.; Pektaş, H.M.; Şen, Ö.F.; Çelik, H. Algorithmic thinking development through physical computing activities with Arduino in STEM education. Educ. Inf. Technol. 2022, 27, 6669–6689. [Google Scholar] [CrossRef] [Scilit]
  9. Katai, Z. The challenge of promoting algorithmic thinking of both sciences- and humanities-oriented learners. J. Comput. Assist. Learn. 2014, 31, 287–299. [Google Scholar] [CrossRef] [Scilit]
  10. Larkin, J.; Simon, H. Why a Diagram is (Sometimes) Worth Ten Thousand Words. Cogn. Sci. 1987, 11, 65–100. [Google Scholar] [CrossRef]
  11. Schön, D.A. The Reflective Practitioner: How Professionals Think in Action; Basic Books: New York, NY, USA, 1983. [Google Scholar]
  12. Cross, N. Descriptive models of creative design: Application to an example. Des. Stud. 1997, 18, 427–440. [Google Scholar] [CrossRef] [Scilit]
  13. Akin, Ö.; Lin, C. Design protocol data and novel design decisions. Des. Stud. 1995, 16, 211–236. [Google Scholar] [CrossRef] [Scilit]
  14. Lawson, B. How Designers Think; Routledge: London, UK, 2005. [Google Scholar] [CrossRef] [Scilit]
  15. Finke, R.A.; Ward, T.B.; Smith, S.M. Creative Cognition: Theory, Research, and Applications; MIT Press: Cambridge, MA, USA, 1992. [Google Scholar]
  16. Cikis, S.; Ek, F.I. Conceptualization by visual and verbal representations: An experience in an architectural design studio. Des. J. 2010, 13, 329–354. [Google Scholar] [CrossRef] [Scilit]
  17. Kristensen, T. The physical context of creativity. Creat. Innov. Manag. 2004, 13, 89–96. [Google Scholar] [CrossRef] [Scilit]
  18. Wallas, G. The Art of Thought; Harcourt, Brace: Orlando, FL, USA, 1926. [Google Scholar]
  19. Guilford, J.P. Creativity. Am. Psychol. 1950, 5, 444–454. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  20. Smith, S.M.; Beda, Z. Unconscious work doesn’t work. Creat. Res. J. 2023, 35, 369–379. [Google Scholar] [CrossRef] [Scilit]
  21. Weisberg, R.W. Creativity: Beyond the Myth of Genius; W.H. Freeman & Co: New York, NY, USA, 1993. [Google Scholar]
  22. Russell, S.J.; Norvig, P.; Davis, E. Artificial intelligence: A Modern Approach; Prentice Hall: Saddle River, NJ, USA, 1995. [Google Scholar]
  23. Terzidis, K. Algorithmic Architecture; Routledge: London, UK, 2006. [Google Scholar]
  24. Başarır, L. Modelling AI in architectural education. Gazi Univ. J. Sci. 2022, 35, 1260–1278. [Google Scholar] [CrossRef] [Scilit]
  25. Bernstein, P.G. Architecture, Design, Data: Practice Competency in the Era of Computation; Birkhaüser: Basel, Switzerland, 2018. [Google Scholar]
  26. Popenici, S.A.D.; Kerr, S. Exploring the impact of artificial intelligence on teaching and learning in higher education. Res. Pract. Technol. Enhanc. Learn. 2017, 12, 22. [Google Scholar] [CrossRef] [Scilit]
  27. Jin, S.; Tu, H.; Li, J.; Fang, Y.; Qu, Z.; Xu, F.; Liu, K.; Lin, Y. Enhancing Architectural Education through Artificial Intelligence: A Case Study of an AI-Assisted Architectural Programming and Design Course. Buildings 2024, 14, 1613. [Google Scholar] [CrossRef] [Scilit]
  28. Wang, J.; Shi, Y.; Chen, X.; Lan, Y.; Liu, S. Teaching with Artificial Intelligence in Architecture: Embedding Technical Skills and Ethical Reflection in a Core Design Studio. Buildings 2025, 15, 3069. [Google Scholar] [CrossRef] [Scilit]
  29. Uncu, E.Ö.; Çağdaş, G. The pedagogical alignment of computational thinking to architecture education for the 21st century learners. J. Comput. Des. 2022, 3, 159–172. [Google Scholar] [CrossRef] [Scilit]
  30. Cross, N. Forty years of design research. Des. Stud. 2007, 28, 1–4. [Google Scholar] [CrossRef] [Scilit]
  31. Ball, L.J.; Christensen, B.T. Advancing an understanding of design cognition and design metacognition: Progress and prospects. Des. Stud. 2019, 65, 35–59. [Google Scholar] [CrossRef] [Scilit]
  32. Cross, N. Designerly Ways of Knowing; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2006. [Google Scholar]
  33. Razzouk, R.; Shute, V. What is design thinking and why is it important? Rev. Educ. Res. 2012, 82, 330–348. [Google Scholar] [CrossRef] [Scilit]
  34. Knuth, D.E. Algorithms. Sci. Am. 1977, 236, 63–81. [Google Scholar] [CrossRef] [Scilit]
  35. Yusuf, A.; Noor, N.M. Modeling students’ algorithmic thinking growth trajectories in different programming environments: An experimental test of the Matthew and compensatory hypothesis. Smart Learn. Environ. 2024, 11, 38. [Google Scholar] [CrossRef] [Scilit]
  36. Ürey, Z.Ç.U. Fostering Creative Cognition in Design Education: A Comparative Analysis of Algorithmic and Heuristic Educational Methods in Basic Design Education. METU J. Fac. Archit. 2021, 38, 53–80. [Google Scholar] [CrossRef] [Scilit]
  37. Ostrowska-Wawryniuk, K.; Strzała, M.; Słyk, J. Form follows parameter: Algorithmic-Thinking-Oriented course for early-stage architectural education. Nexus Netw. J. 2022, 24, 503–522. [Google Scholar] [CrossRef] [Scilit]
  38. Vazquez, E. Teaching parametric design: Fostering algorithmic thinking through incomplete recipes. Open House Int. 2023, 49, 736–751. [Google Scholar] [CrossRef] [Scilit]
  39. Austin, M.; Qattan, W. I’m a visual thinker: Rethinking algorithmic education for architectural design. In Living Systems and Micro-Utopias: Towards Continuous Designing, Proceedings of the International Conference on Computer-Aided Architectural Design Research in Asia; The Association for Computer-Aided Architectural Design Research in Asia (CAADRIA): Hong Kong, 2016; pp. 829–838. [Google Scholar] [CrossRef] [Scilit]
  40. Tan, F. Translation of Form: A Model for Teaching Design Coding to Undergraduate Students. Nexus Netw. J. 2024, 26, 897–924. [Google Scholar] [CrossRef] [Scilit]
  41. Chen, X.; Xie, H.; Zou, D.; Hwang, G.-J. Application and theory gaps during the rise of Artificial Intelligence in Education. Comput. Educ. Artif. Intell. 2020, 1, 100002. [Google Scholar] [CrossRef] [Scilit]
  42. Crespo, S.; McCormick, F. Augmenting Digital Nature Generative Art as a Constructive Feedback Loop. In Machine Hallucinations: Architecture and Artificial Intelligence; Del Campo, M., Leach, N., Eds.; John Wiley & Sons: Oxford, UK, 2022; Volume 3, pp. 54–59. [Google Scholar]
  43. Çelik, T. Generative design experiments with artificial intelligence: Reinterpretation of shape grammar. Open House Int. 2023, 49, 822–842. [Google Scholar] [CrossRef] [Scilit]
  44. Tan, L.; Luhrs, M. Using Generative AI Midjourney to enhance divergent and convergent thinking in an architect’s creative design process. Des. J. 2024, 27, 677–699. [Google Scholar] [CrossRef] [Scilit]
  45. Flower, L.; John, R.H. A Cognitive Process Theory of Writing. Coll. Compos. Commun. 1981, 32, 365–387. [Google Scholar] [CrossRef] [Scilit]
  46. Kelly, N.; Gero, J.S. Design thinking and computational thinking: A dual process model for addressing design problems. Des. Sci. 2021, 7, e8. [Google Scholar] [CrossRef] [Scilit]
  47. Ericsson, K.A.; Simon, H.A. Protocol Analysis: Verbal Reports as Data; MIT Press: Cambridge, MA, USA, 1993. [Google Scholar]
  48. Braun, V.; Clarke, V. Reflecting on reflexive thematic analysis. Qual. Res. Sport Exerc. Health 2019, 11, 589–597. [Google Scholar] [CrossRef] [Scilit]
  49. Braun, V.; Clarke, V.; Rance, N. How to use thematic analysis with interview data (process research). In The Counselling and Psychotherapy Research Handbook; UWE Research Repository (UWE Bristol): Bristol, UK, 2014. [Google Scholar]
  50. Lee, J.H.; Ostwald, M.J. The relationship between divergent thinking and ideation in the conceptual design process. Des. Stud. 2022, 79, 101089. [Google Scholar] [CrossRef] [Scilit]
  51. Liang, C.J.; Le, T.H.; Ham, Y.; Mantha, B.R.K.; Cheng, M.H.; Lin, J.J. Ethics of artificial intelligence and robotics in the architecture, engineering, and construction industry. Autom. Constr. 2024, 162, 105369. [Google Scholar] [CrossRef] [Scilit]
  52. Yazici, G.; Dogan, F. Fostering metacognition in the design studio: The effect of minimal interventions on architectural students’ metacognitive awareness. Think. Ski. Creat. 2024, 54, 101651. [Google Scholar] [CrossRef] [Scilit]
  53. Brown, T.B.; Mann, B.; Ryder, N.; Subbiah, M.; Kaplan, J.; Dhariwal, P.; Neelakantan, A.; Shyam, P.; Sastry, G.; Askell, A.; et al. Language Models are Few-Shot Learners. arXiv 2020, arXiv:2005.14165. [Google Scholar] [CrossRef] [Scilit]
  54. Wang, L.; Xu, W.; Lan, Y.; Hu, Z.; Lan, Y.; Lee, R.K.-W.; Lim, E.-P. Plan-and-Solve prompting: Improving Zero-Shot Chain-of-Thought reasoning by large language models. arXiv 2023, arXiv:2305.04091. [Google Scholar] [CrossRef] [Scilit]
  55. Perkins, D.N. The Mind’s Best Work; Harvard University Press: Cambridge, MA, USA, 1981. [Google Scholar]
  56. Hou, C.; Zhu, G.; Sudarshan, V.; Lim, F.S.; Ong, Y.S. Measuring undergraduate students’ reliance on Generative AI during problem-solving: Scale development and validation. Comput. Educ. 2025, 234, 105329. [Google Scholar] [CrossRef] [Scilit]
  57. Amabile, T.M. Social psychology of creativity: A consensual assessment technique. J. Personal. Soc. Psychol. 1982, 43, 997–1013. [Google Scholar] [CrossRef]
Buildings 16 01097 g001
Figure 1. Research methodology. (Source: authors).
Figure 1. Research methodology. (Source: authors).
Buildings 16 01097 g001
Buildings 16 01097 g002
Figure 2. Module layout. (Source: authors).
Figure 2. Module layout. (Source: authors).
Buildings 16 01097 g002
Buildings 16 01097 g003
Figure 3. Illustration of the conceptual framework. (Source: authors).
Figure 3. Illustration of the conceptual framework. (Source: authors).
Buildings 16 01097 g003
Buildings 16 01097 g004
Figure 4. General layout for the interpretation of the process. (Source: authors).
Figure 4. General layout for the interpretation of the process. (Source: authors).
Buildings 16 01097 g004
Buildings 16 01097 g005
Figure 5. The basic structure of the framework. (Source: authors).
Figure 5. The basic structure of the framework. (Source: authors).
Buildings 16 01097 g005
Table 1. Example of coding progression. (Source: Authors).
Table 1. Example of coding progression. (Source: Authors).
Data ExtractInitial CodeSub-ThemeTheme
“For starting the proposal part and rendering of the plan. It’s easy for me and clear.” (P36)ConceptStages having
beneficial usage		Design stages that have benefit of AI usage
Proposal/Initial idea
Site analysis
“I won’t use it in the future because the
designer should be free.” (P24)		Future projects will not benefit from AIAI will not be usedUsage of AI for the future projects
“I believe the instructors could find the right way in which it could be used and guide us through that.” (P3)Openness, AcceptanceSupportive mannerExpectations from instructors regarding AI integration
Table 2. The relation of Schön’s [11] “reflection-in-action” and “reflection-on-action” concepts with the proposed conceptual framework.
Table 2. The relation of Schön’s [11] “reflection-in-action” and “reflection-on-action” concepts with the proposed conceptual framework.
Relation with the ModelCorrelation Type with Schön’s TheoryReflection Type
Reflection with protocol analysisDirect relationReflection-on-action
Implementation of the processDirect relationReflection-in-action
Geneplore partSemi relationReflection-on-action & Reflection-in-action
PromptingIndirect relationReflection-in-action
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Hikmet, Ş.; Ozay, N. Human–AI Collaboration in Architectural Design Education: Towards a Conceptual Framework. Buildings 2026, 16, 1097. https://doi.org/10.3390/buildings16061097

AMA Style

Hikmet Ş, Ozay N. Human–AI Collaboration in Architectural Design Education: Towards a Conceptual Framework. Buildings. 2026; 16(6):1097. https://doi.org/10.3390/buildings16061097

Chicago/Turabian Style

Hikmet, Şerife, and Nazife Ozay. 2026. "Human–AI Collaboration in Architectural Design Education: Towards a Conceptual Framework" Buildings 16, no. 6: 1097. https://doi.org/10.3390/buildings16061097

APA Style

Hikmet, Ş., & Ozay, N. (2026). Human–AI Collaboration in Architectural Design Education: Towards a Conceptual Framework. Buildings, 16(6), 1097. https://doi.org/10.3390/buildings16061097

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop