2.1. Overview of key design research frameworks

Archer has identified academic research in three main fields: science, the humanities and design (Archer Reference Archer1981). The claim that design is an independent discipline (Cross Reference Cross1982) marked a distinct point in the history of design research. In the 1970s, many prominent academic architects promoted this view in the design community.

Theoretical attempts to categorize design research have traditionally aimed at defining its boundaries, articulating its aims and legitimizing it as a scholarly endeavor. Three of the most widely referenced models – Frayling’s (Reference Frayling1993) triad of research into, through and for design; Cross’ (Reference Cross2006) domains of designerly ways of knowing; and Buchanan’s (Reference Buchanan2001) typology of basic, applied and clinical research – have significantly shaped how design research is taught and discussed in academic settings (Table 1).

Table 1.

Key frameworks in design research

Three modalities of design research are presented by Frayling’s model: research through design, which incorporates practice-led inquiry; research into design, which concentrates on historical, theoretical or cultural knowledge; and research for design, which aims to improve tools or techniques. However, Cross groups research into three categories based on their sources: people (design epistemology), processes (design praxeology) and products (design phenomenology). By separating basic research for fundamental knowledge, applied research for focused problem-solving and clinical research based on actual professional contexts, Buchanan’s method presents a pragmatic typology.

These frameworks have been instrumental in shaping how design research is discussed and taught, particularly in architecture schools. However, they rarely engage with questions of when in the design process research is conducted or under what research attitude it is carried out. The temporal placement of research, whether in early conceptual stages, during or after design, is often a tacit element of the research process and is not explicitly articulated or discussed. Similarly, the attitude guiding the research, be it that of a practitioner, academic researcher or hybrid, remains implicit in most models. This leaves a conceptual gap in understanding how time approach and attitude shape the kinds of knowledge design research produced.

2.2. Gaps and conceptual tensions

While the frameworks listed in Table 1 are important and widely referenced, they initially operated in isolation and were rarely seen to be synthesized or critically compared. Each offered a particular lens on how knowledge in design is structured or generated, but they lacked integration when applied to architectural contexts. Within architectural education, they often remain abstract and somewhat disconnected from the realities of design practice (Findeli et al. Reference Findeli, Brouillet, Martin, Moineau and Tarrago2008). They rarely specify when in the design process research takes place or whose perspective shapes the inquiry. For instance, a research project conducted by a practitioner during conceptual development might follow entirely different assumptions and methods compared to one led by an academic in a retrospective study (Groat & Wang Reference Groat and Wang2013). These distinctions – though subtle – have significant implications for how design knowledge is generated and applied (Jonas Reference Jonas2007).

Attempts have been made to bridge these silos. Frankel & Racine (Reference Frankel, Racine, Durling, Bousbaci, Chen, Gauthier, Poldma, Roworth-Stokes and Stolterman2010), for example, mapped Frayling’s triad (Frayling Reference Frayling1993; Archer Reference Archer1995; Friedman Reference Friedman2008) onto Buchanan’s structure (Frayling Reference Frayling1993; Archer Reference Archer1995; Findeli Reference Findeli1999; Downton Reference Downton2003; Friedman Reference Friedman2003; Cross Reference Cross2007), establishing correspondences between research about design (basic), research through design (applied) and research for design (clinical). Clemente, Tschimmel & Pombo (Reference Clemente, Tschimmel and Pombo2017) proposed a fourth category – research from design – derived from doctoral case studies, extending the typology to reflect emerging practices better. They also integrate authors’ contributions among them Frayling (Reference Frayling1993), Cross (Reference Cross2007), Friedman (Reference Friedman2008) and Findeli et al. (Reference Findeli, Brouillet, Martin, Moineau and Tarrago2008). In the current study, there are instances which have not been widely interpreted due to the manuscript’s coherency. For example, Wang’s arguments in book Architectural Research Methods (Groat & Wang Reference Groat and Wang2013) provide a comprehensive overview of discussion around research by design alongside presenting several paradigms in design research. As it has been claimed that research by design grew out of the tradition of design research, Wang’s results are worthwhile. Other research on this topic is Lee & Lee (Reference Lee and Lee2019), which is in accordance with Frankel & Racine (Reference Frankel, Racine, Durling, Bousbaci, Chen, Gauthier, Poldma, Roworth-Stokes and Stolterman2010), but variations can be observed.

These efforts are outlined in Table 2. Still, most of these efforts overlook the role of (1) Timing (the point of research integration in design) and (2) Attitude.

Table 2.

Other frameworks in design research and their similarities or differences with key frameworks

Gap No. 1: Schön’s theory of reflective practice (Schön Reference Schön1983, Reference Schön1992; Reich Reference Reich2017; Busciantella-Ricci & Scataglini Reference Busciantella-Ricci and Scataglini2024) introduces time as a central variable, distinguishing between reflection-in-action (unconsciously – through experience – during design) and reflection-on-action (self-consciously – retrospective – after design). According to Schön (Reference Schön1983, Reference Schön1992), the timing of research integration not only affects the nature of the data collected but also influences the type of knowledge produced. Findings from this study revealed that some research categorized at first glance as “through design” in fact aligned more closely with reflection-on-action, since their documentation was compiled after the completion of the design process. This theoretical and empirical overlap underscores the value of integrating Schön’s temporal framework with existing design research classifications to capture such variations better. His work demonstrates that while research and reflection are conceptually distinct, an awareness of the integration point of research into the design can influence the structure and organization of various forms of design research (Figure 1).

Figure 1.

The role of timing in design research (Gap No. 1).

Gap No. 2: Sanders, meanwhile, adds the dimension of attitude by contrasting expert-driven approaches with participatory, user-centered ones (Sanders Reference Sanders2006). Also, Findeli and colleagues consider three end users of design research are interested in its outputs: the design research community (researchers), the design practice community (practitioners) and the design education community (educators) (Findeli et al. Reference Findeli, Brouillet, Martin, Moineau and Tarrago2008). These insights point toward the need for a more flexible and responsive model. Beyond design methods, a research attitude is crucial (Figure 2). Rowe was one of the first pioneers in the study of design thinking in architecture, a concept central to the design research community since the 1980s (Rowe Reference Rowe1994). The Delft Protocols Workshop (Cross et al. Reference Cross, Christiaans and Dorst1996) further explored design thinking and cognitive psychology through protocol analysis of individual and team processes. Both Rowe’s approach to “design thinking” and the Delft Protocols’ cognitive analysis provide evidence of how research attitudes and within-process interactions shape the generation of ideas and design decisions. In the dataset analyzed for this study, patterns emerged that resonate with those identified by Rowe and the Delft team – particularly in studies adopting a “practitioner + user” attitude, where joint sessions and direct feedback formed integral parts of the design process. The consistency between our observational findings and previous studies bolsters the soundness of the suggested attitude-based structure and shows its usability in different design contexts.

Figure 2.

The role of attitude in design research (Gap No. 2).

In response to this appeal, the current study suggests a conceptual framework that superimposes two important elements on top of the current classifications: (1) when research is conducted during the design process and (2) who performs it. This framework provides a matrix that aids in putting studies within the broader field of architectural design research, rather than taking the place of previous models. By doing this, it seeks to offer more conceptual clarity as well as useful advice for negotiating the challenges of doing research in design-driven domains.

2.3. Conceptual framework

To address the conceptual gaps identified in previous models, this study introduces an integrative framework that incorporates two underexplored but highly influential dimensions in design research: time-based approach and research attitude. Rather than proposing a new classification system, the proposed approach first seeks to enrich existing frameworks – particularly those of A, B and C as outlined in Table 1 – by adopting a time-based perspective that allows for more nuanced interpretation and application. The time-based approach refers to the point of research integration in design. Drawing from literature in architectural design pedagogy and practice, it seems possible to differentiate between three temporal phases: reflections on pre-design, design or post-design. This dimension recognizes that the relevance, purpose and methodology of research may shift depending on when it is situated within the design process.

The research attitude, in parallel and in the subsequent stages of this study, reflects the dominant orientation and guiding role of the research process. Also, as mentioned earlier, this study has attempted to focus specifically on architecture from the broad disciplines of design to conduct a more detailed investigation. The precedent for this can be found in the studies of Till and Lawson. Where Till thought of a model that goes beyond both the science/art and quantitative/qualitative divisions (Till Reference Till2008), to enable thematic and interdisciplinary research across the three stages and to allow contributions from very different kinds of expertise – scientists, historians and practitioners – to contribute to the research.

This study does not claim these to be exhaustive or mutually exclusive, but they serve as useful heuristics to understand better how attitudes and research goals vary across contexts. By crossing these two dimensions, the conceptual framework facilitates a transparent analysis of how different studies locate themselves within the design research landscape – not only in terms of what kind of research is conducted but also when it happens and under what research attitude. This enables both more precise categorization and a deeper appreciation of the dynamics shaping knowledge production in architectural design.

Beyond the two primary dimensions discussed, namely the timing of research and the research attitude, this study also takes into account the kinds of knowledge that design research tends to produce. Building on the work of Cross and Schön, three modes of knowing can be distinguished that often operate in parallel. First, epistemological knowledge emerges from analytical or theoretical inquiry, often conducted by researchers working outside the act of designing itself. Then there is praxeological knowledge, grounded in the hands-on process of designing – what Schön famously called “reflection-in-action.” Finally, phenomenological knowledge arises from how design is experienced, whether by designers, users or communities. These modes of knowing are not always categorized in taxonomies, but they shape how design research in architecture is understood. Figure 3 offers a way to map this understanding, showing how timing and attitude intersect across different approaches drawn from the literature.

Figure 3.

Conceptual framework.

Figure 3 presents the conceptual framework guiding this study, combining two key dimensions identified from the literature: the point of research integration in design (before, during or after) and the attitude of the research (for example, practitioner, researcher or user). Each intersection in the framework is informed by established theoretical typologies. For instance, research about design led by researchers before or after the design process corresponds to Frayling’s classification and Buchanan’s research categories. Practitioner-driven inquiries during design align with Schön’s reflective practice and Cross’ praxeological approach. User-centered approaches during or after design relate to participatory design or post-occupancy evaluations, as discussed by Groat & Wang. By integrating these models, the framework not only maps existing design research typologies but also surfaces conceptual blind spots, particularly where time and research attitude have not been jointly considered. This framework sets the stage for the analytical categorization of architectural design research studies in the following sections.

3.1. Inclusion and exclusion criteria

Also, Table 4 presents a list of the inclusion and exclusion criteria used in the process of study selection. As it has already been noted, for the purpose of screening and collecting studies, the updated guidelines of PRISMA 2020 were followed by this paper (Page et al. Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann, Mulrow, Shamseer, Tetzlaff, Akl, Brennan, Chou, Glanville, Grimshaw, Hróbjartsson, Lalu, Li, Loder, Mayo-Wilson, McDonald and Moher2021). Figure 4 includes the details of screening studies. The collection of selected studies has been screened in three separate stages – title, abstract and full text.

Table 4.

Inclusion and exclusion criteria of selected studies

3.2. Quality appraisal (CASP)

All 94 identified studies underwent full-text screening and critical appraisal using the Critical Appraisal Skills Programme (CASP) qualitative checklist by independent reviewers (P.A. & N.S.) (CASP Qualitative Checklist 2018). Studies were categorized as high (9–10), moderate (7.5–9) or low (≤7.5) based on quality scores (0–10) assigned using a predefined checklist. The CASP criteria included evaluation of research aims, methodological clarity, ethical considerations, rigor of data collection, transparency of analysis and the validity of findings. In this context, several systematic reviews can be referenced. For instance, DeSa et al. have applied the same ranges for the three intervals (high, moderate and low) for overall quality (DeSa et al. Reference DeSa, Gebremeskel, Omonaiye and Yaya2022). Additionally, some studies, such as Tao et al. (Reference Tao, Coates and Maycock2012), have adopted a more flexible approach by considering scores of 75% or higher (7.5–10) as indicative of high quality. Therefore, it can be concluded from these similar studies that scores of 7.5 and below may be excluded from further analysis.

Following resolution of any disagreements regarding study quality (When one reviewer had given a study an 8 and another a 5, the average being 6.5, the study would be excluded from the review since the score was less than 7.5. When this happened, a third reviewer (M.Z.) would be invited for a meeting to ascertain whether he or she felt that the study was worth a higher quality rating than 7.5. The objective was to agree on the two existing scores to provide the study with an opportunity to not exclude from further review), only studies categorized as high (n = 24) or moderate (n = 46) quality (scores 7.5–10) were included for analysis of the connection between design research and design process in the context of architecture, and the theoretical synthesis of the research was done based on them. This procedure ensured that decisions regarding inclusion or exclusion were not based on single-reviewer judgments, but were collaboratively discussed to enhance fairness and minimize bias.

The results that come from the scores taken from evaluators are used in determining whether the evaluated resources can conclude by the percentage of agreement between evaluators, and Cohen’s kappa coefficients (Warrens Reference Warrens2011). Evaluator scores, P.A. & N.S., have been adopted for computing quality ratings, low, medium and high, for every study on its own in the research. Inter-rater agreement assessments were performed using percent agreement and Cohen’s kappa coefficient. Kappa interpretation followed established guidelines slight, fair, moderate, substantial and perfect (Landis & Koch Reference Landis and Koch1977; Conger Reference Conger2016). As can be seen in Table 5, in the current study, the inter-rater agreement was 89.36%, while the kappa coefficient was 0.835, indicating almost perfect agreement between the evaluators. Disagreements were resolved through discussion, and when necessary, with the involvement of a third reviewer. This process contributed to the reliability and transparency of the quality appraisal.

Table 5.

Details of kappa coefficient and percentage of agreement between evaluators

3.3. Coding and categorization

Following the checklist application, a comprehensive list of records was reviewed. Title and abstract screening identified some preliminary thematic areas. The coding process followed a structured consensus approach: each author independently coded the studies for two analytical dimensions, (1) time of research integration (pre-, during or post-design) and (2) research attitude (practitioner, practitioner + user, practitioner + AI, researcher or user), before meeting in structured consensus sessions to resolve disagreements. A structured procedure was used to determine author agreement, starting with the categorization of each paper across three research timings and research attitudes. The timing dimension showed scant inter-author discord. An attitude was accepted if it was endorsed by most of the authors. In instances of disagreement, a structured group meeting was held with the three reviewers involved in the checklist procedure (P.A., N.S. and M.Z.). The goal of these sessions was to either gain a majority consensus on an existing attitude or to suggest and name a novel attitude founded on theoretical structures. It is vital to mention that the practitioner attitude with AI, and the user attitude were derived from these meetings. Following the conclusion of each study’s attitude, the next phase involved the raw coding of ideas. This stage followed the same method of naming and improvement, enabling the adjustment of existing names or the formation of new ones, as shown in Figure 8.

The thematic analysis followed that of Braun and Clarke. The steps involved in conducting a thematic analysis are identified as (Braun & Clarke Reference Braun and Clarke2006; Jones, Coviello & Tang Reference Jones, Coviello and Tang2011): familiarization with data; generating initial codes; searching for themes; reviewing themes; defining and naming themes; and producing the report. Thematic areas are the key concepts that represent the holistic view of the selected studies. During the second phase of the analysis, themes provided relevant information on search terms and recommendations of subthemes. Accordingly, the authors comprehensively labeled the thematic areas, while sub-themes were derived from the analyzed literature. The results are presented in the ensuing sections.

The coding of the studies subsequently began with the initial coding of each study against the listed attitudes in reading the literature. The authors thereafter independently open-coded the studies according to each attitude, and the results are presented as raw concepts in Section 4.2 and Figure 8. This was followed by discussion of how to render these concepts into indicators. Inter-rater reliability was assessed using Cohen’s kappa coefficient at two key stages: the determination of attitudes and the conceptual coding phase. At the attitude-determination stage, the level of agreement was found to be in the “almost perfect” range. In the subsequent conceptual coding stage, the agreement was in the “significant” range. Both results were deemed sufficient to proceed with the research.

Through extensive authors’ discussion, and consideration of the included studies, the findings were first categorized into sub-factors. During the next phase, collated sub-factors were consolidated into more general factors. The iterative nature of coding and discussion allowed refinement of categories through readings of selected studies introduced earlier. Further details about this process can be obtained in Section 5.2 and Figure 11. Also, this section achieves validity and reliability through triangulation, peer debriefing and an audit trail.

4.1. Based on time approach

In the literature, one of the main classifications is related to Christopher Frayling’s study in 1993 (Frayling Reference Frayling1993), which has been partially followed by other researchers in the naming of various types (Friedman Reference Friedman2008; Forlizzi, Zimmerman & Stolterman Reference Forlizzi, Zimmerman and Stolterman2009; Frankel & Racine Reference Frankel, Racine, Durling, Bousbaci, Chen, Gauthier, Poldma, Roworth-Stokes and Stolterman2010). Research into/about/on design, research through/by design and research for design are the three categories into which design research is separated. Based on this classification, 70 selected studies were initially labeled after reading the full text and then analyzed more clearly. Several selected studies aligned with this highly cited framework. However, in other cases a conflict was recorded between the intent of the research and the attempt to apply this model. Some studies, while aligned with the spirit of research through design, did not incorporate research integration during the design process. Conversely, other studies were conducted after the design phase, but they lacked any connection to research about design. These mismatches were observed in relation to Frayling’s typology, which despite its influence did not fully accommodate the temporal diversity found in contemporary architectural research. As a result, a group of studies has been identified and is temporarily referred to in this section as a fourth type of research X design. Consequently, the first potential gap in research emerged, referring to the point of research integration in design (time-based approach). Further discussion of this section can be found in Section 5.1.

After the classification of the 70 selected studies based on design research approaches, reading and labeling them were conducted based on different subjects. The following chart shows the timeline of the studies within these four categories (Figure 5).

Figure 5.

Labeling of studies based on design research approaches by year.

Also, referring to the book Research Methods in Architecture authored by Linda Groat and David Wang (Groat & Wang Reference Groat and Wang2013), it was determined which strategy or strategies were used in each study (Figure 6). This classification for studies made it possible to identify the variety of methodological approaches used in architecture – from interpretive-historical to experimental-applied – and to examine how these strategies intersected with the time-based and attitude-based dimensions analyzed in this research. This methodological linkage revealed that certain temporal categories (such as post-design) showed a stronger association with particular strategies and tactics, a relationship explored in more depth in the Discussion (Section 5). As shown in the figure, 18 studies in research about design were identified as utilizing the interpretive-historical research strategy. On the other hand, research through design combines various qualitative research strategies such as case study, ethnography or simulation research. The other two types of design research (For and X types), as per their labels, were recorded as relatively more recent, with an increase observed since the 2010s. In the following, the tactics used and the duration of the research were analyzed and categorized (Figure 6).

Figure 6.

Strategies and tactics employed in the studies by year and design research approaches.

Analysis and reading of studies conducted as research about design shows that this type is about the cognition of design. Also, they generally do not use qualitative strategies, and only four studies had case studies in their research. They have often used historical-interpretive research with archival documents tactics from the researcher’s point of view. Some of these researches are about reflective knowledge (Jacques Reference Jacques1993; Koskinen et al. Reference Koskinen, Zimmerman, Binder, Redström and Wensveen2011; Robertson & Simonsen Reference Robertson and Simonsen2012; Robertson & Wagner Reference Robertson and Wagner2012; Luck Reference Luck2018; Herr Reference Herr2019). Others have dealt with design creativity (Purcell & Gero Reference Purcell and Gero1998; Biskjaer & Christensen Reference Biskjaer and Christensen2021), and more specifically, they have examined topics such as the role of drawings and the primary generator. In this cluster of studies, an analytical distance between researcher and process was observed, with retrospective insights reported more frequently than real-time ones. Another part of the studies in this field briefly focuses on evidence-based design (Lawson Reference Lawson2010; Pati & Barach Reference Pati and Barach2010; Shepley & Song Reference Shepley and Song2014; Halawa et al. Reference Halawa, Chalil Madathil, Gittler and Khasawneh2020), and the relationship between computers and design (Rosenman, Gero & Maher Reference Rosenman, Gero and Maher1994; Hight Reference Hight2013; Joachim & Aiolova Reference Joachim, Aiolova, Gerald, Carayannis and Campbell2015; Kızılcan Reference Kızılcan2021) and universal design (Heylighen Reference Heylighen2014).

Also, in research X design, it has been tried to find out how design works in a scientific and research-oriented way through archival documents, content analysis, interviews, questionnaires, audio and video recording, etc., by mentioning two points: 1. These activities often not focus on the outcome and the final product and are more concerned with design cognition (different from research for design) and 2. The recording of documents was not done at the moment of the process, but after its completion, in such a way that in some cases it seems to be a narrative of design in the past (different from research through design). While both through-design and X-design emphasize the inner workings of design cognition, their temporal stance differs significantly; one engages in-process observation, while the other constructs narratives retrospectively. This distinction was identified as aligning with the time-based lens, which provided a framework for nuanced classification.

4.2. Based on research attitude

Literature review shows that, next to the time-based approach, another relevant concept in design studies is the necessity of heeding the effective attitude in every study. In most of the reviewed literature, knowledge production was reported as being positioned in close proximity to the practitioner’s attitude. This knowledge relates to Sanders’ expert mindset and Findeli’s practitioner and educator attitude constructs. In this research study, the people engaged in the practice of architecture, ranging from students to experts, were positioned as practitioners. A second subset of the analyzed studies was marked as practitioner + user, indicating a participatory dynamic similar to Sanders’ work. In addition, a researcher’s attitude, in line with Findeli’s approach, marked another group of the selected studies. Aside from the three aforementioned groups, some studies developed knowledge almost exclusively based on user attitude, with minor indirect input by practitioners. Such studies commonly utilized post-occupancy evaluation methods, usually including user interviews. Concurrently, the increasing use of artificial intelligence (AI) has led to the emergence of a practitioner + AI attitude in some of the studies examined.

Therefore, after a thorough review of each study and brainstorming, the studies were categorized according to the attitude that guided the research direction. These research attitudes were identified in five modes: (1) practitioner; (2) practitioner with user; (3) practitioner with AI; (4) researcher; and (5) user. This classification was observed to show diversity in design research not only in terms of roles, but also in the nature of decision-making and motivation. Therefore, coding was followed to identify indicators, sub-factors and factors in each study in five main groups (as mentioned at the end of the research methodology section). Proper identification of factors and sub-factors may help to get a network with related concepts (Figure 7).

Figure 7.

Relationship between attitude, factor, sub-factor and indicators in the research.

The authors further read each study and compiled concepts that seemed relevant, and then grouped these concepts into the attitudes as depicted in Figure 8. For example, the concept of “drawing” was found mainly in the attitude of practitioner but also occurred in practitioner working with user. This overlap was recorded as showing that, while research attitudes provide a useful framework, many concepts in design research appeared across multiple orientations, reflecting the hybrid nature of real-world design practice.

Figure 8.

Concepts coded in the research.