A Comparative Analysis of Student Perceptions of Recommendations for Engagement in Design Processes

Engineering designers rely on various forms of support, such as design tools, training, and process models, to address complex and multifaceted design problems. Designers, from novices to experts, leverage design process models—particularly prescriptive process models—to support systematic approaches to their design work. Systematic approaches are especially helpful in situations where failure can have grave consequences, where there is a high probability of being wrong, or where problems are complex [1]. Novice designers, with limited experience and a higher likelihood of making mistakes [2], are especially well-situated to benefit from such models—as intentional practice is crucial in developing expertise [3]. Furthermore, research supports the efficacy of models as pedagogical tools across different domains, from supporting students' understanding of atoms and molecules [4] to fostering scientific creativity [5].

While numerous design process models have been developed to support design training and practice, we do not have a deep understanding of their impact on model users. For example, related research describes understandings of process models based on “throwaway comments of interviewees and complaints of project managers” (p. 9) and informal follow-up discussions with design managers [6]. Thus, we cannot leverage knowledge of design process models' impact to support model users in choosing and using such models. The lack of research on the impacts of design process models on model users stands in contrast to the numerous studies in science literature that have investigated students' understandings after engagement with visual representations to inform pedagogy (e.g., [7–9]). Furthermore, while there is prior research on the visual representations of design process models, much of this work focuses on researchers' analysis of the visual representations (e.g., [10]). The use and development of design process models that align with recommended practices for user-centered design [11,12] should include the perspectives of model users, such as design practitioners, students, and educators.

Additionally, while it is crucial for mechanical engineers to incorporate social and contextual dimensions into their decision-making to avoid causing harm and also to better human welfare [13], many visual representations of design process models (e.g., [2,14]) lack explicit references to some of these dimensions. Understanding the guidance provided to designers, especially novice designers, by design process models and how the inclusion of social and contextual dimensions affects this guidance is essential. Therefore, we interviewed novice designers about three design process models known to be used in undergraduate engineering design education, including one that explicitly incorporated a range of social and contextual dimensions. The outcomes of this work can support design pedagogy and inform model users' selections and use of the design process models investigated.

Given the complexity of design [15] and the abstract nature of all models [6], there is no single comprehensive design process model. Further, process models are affected by model creators' priorities and representation affordances. Thus, evaluation of process models should focus on fulfillment of their intended purposes rather than trying to determine absolute “correctness” [6].

There are numerous ways to express a model, including graphically, mathematically, or in text [15]. Many design process models are represented with shapes, arrows, and text labels highlighting stages of and pathways through design work [14]. When using models that have a visual representation, some model users may draw on textual descriptions of a process in addition to the model's visual representation. However, the principle of salience for visual displays says that the most important information should be salient within a display [16–19]. Thus, following the principle of salience, a good visual representation of a design process model represents the most important parts the model creator wanted to emphasize. Additionally, the benefit of visualizations is reflected in their ability to be used as a map or guide for approaching design work. While a textbook may describe a design process model, in practice, an engineer will probably not have the time to reread all available text to review a model, but they can have a printout of the visual representation on their desk for quick reference.

One way to classify process models is whether they have prescriptive and/or descriptive aspects. Prescriptive models are “those that prescribe how the design process ought to proceed” [20], and descriptive models are those that focus on describing how design occurs in practice [20]. Importantly, a single design process model can have both prescriptive and descriptive elements [20,21]. Prescriptive design process models are a particularly important type of process model for supporting design practice as they necessarily aim to guide designer behavior.

Prescriptive design process models are developed to support communication within and across disciplines, prevent omissions in the process, enable education, and facilitate planning [22]. Process models are leveraged in industry projects to support planning a design process [23]. While more research evidence is needed to verify the assumption that following a prescribed process leads to better design outcomes [20,24], recent design studies suggest a connection between using a prescriptive design process model and quality design outcomes. Moraes and colleagues [25] introduced the “W-model” as a prescriptive design model for pre-college novices and found informed designer behaviors—defined by the Informed Design Teaching and Learning Matrix [26]—in student teams using this model. Another study on a freshman-level design-and-build course found a moderate-to-strong correlation between student teams' adherence to a prescriptive design cycle and client satisfaction for the most difficult project, leading the authors to suggest that the impact of prescriptive design cycles may be stronger for more difficult projects compared to simpler ones, as ad hoc solutions may satisfice in simpler projects [27].

In fields that utilize models, such as the sciences, scholars have outlined various roles models play in supporting processes and outcomes within the discipline. Models in science serve a multitude of purposes, including visualization, hypothesis formulation, critical analysis of ideas, examination of theories, and derivation of relationships [28–31]. In engineering, models are used in similar ways, including to emphasize key aspects of design processes (e.g., [32]), whether describing or prescribing design practice. While researchers have compared design models from academia and industry—based on aspects of the models such as stages included—by conducting literature studies and analyzing the models themselves (e.g., [10,22,33,34]), limited research exists on designers' perceptions of design process models. Eckert and Stacey [6] noted that the engineers and project managers they interviewed across several studies lacked awareness of diverse interpretations of models and assumed a single correct interpretation of the model [23,35,36]. The researchers found various interpretations of a process model across participants and related different interpretations to model users' experiences, priorities, assumptions, and views of a model as a mandate or guideline [6].

Research on science models has shown models' role in how students interpret phenomena (e.g., [37,38]). Further, students are more likely to assume and accept the accuracy of models rather than recognize where a model is reflective of reality and where it is not [39–41], suggesting that students do not seek to build a deeper understanding of the reality the model represents. Studies in science have also focused on what understandings students have after they engage with visual representations to guide development of learning tools and experiences that recognize students' interpretations of common representations (e.g., [7–9]). In the context of engineering design, we have a limited understanding of how students interpret the visual representations of existing design process models and, thus, cannot leverage this knowledge to guide how we support designers in choosing and using design process models. Given this limited research, we explored mechanical engineering students' perceptions of a single design process model in a prior study. We found consistent perceptions of the model—for example, the iterative nature of the model—as well as perceptions that varied, for example, relationships between different visual components [42]. In a follow-up study, we identified eight dimensions that students used to describe the usefulness of three process models and two features—iteration and level of detail—that students named as key distinguishing features across models [43].

Our study investigated engineering student perceptions of design approaches recommended by three design process models. Our study aligns with the design research methodology framework as a “Descriptive Study I” [44], which is focused on description rather than prescription. In other words, our paper describes and comments on mechanical engineering students' perceptions of three design process model visualizations. However, we do not prescribe specific changes to any of the models. We aligned with Eckert and Stacey's [6] view that prescriptive process models should be evaluated based on their suitability for a particular purpose. Thus, the final decision on what, if any, updates should be made to the three process models investigated in our study is a case-by-case determination dependent on a particular model user's purpose for a model. We focused solely on the visual representations of design processes as they are more likely to be referenced by designers throughout design work. This approach allowed for a more controlled comparison of participant perceptions across multiple models.

Our study was guided by the following research questions:

1. What recommendations for design process approaches and considerations do upper-level mechanical engineering students perceive to be present or absent in various design process models?

2. How do student perceptions of recommendations vary by design process model?

In the theory of planned behavior, intentions are determined by one's attitude, perception of social pressure, and perception of ease of performing a particular behavior [45]. Behavior is predicted by intention and one's perception of ease of performing the behavior [45]. This theory implies that designers' claims about what a model suggests they do contribute to predicting their actual behaviors. Guided by the theory of planned behavior, we propose that design process models shape designers' intentions by influencing their attitudes about design, their conceptions of engineering design norms, and their perceptions of the ease of specific design behaviors [43].

Participants included 18 undergraduate mechanical engineering students recruited from a public Midwestern university. Our sample size is appropriate for a qualitative study focused on a nuanced understanding of participant perceptions and is consistent with the range of participants included in other qualitative design studies (e.g., [46,47]). We recruited students who had completed at least two courses of a three-course design and manufacturing sequence to (1) ensure their exposure to some design coursework and (2) reduce the number of factors influencing student perceptions. Recruitment involved sending emails to a university listserv for undergraduate mechanical engineering students and engineering student groups as well as a posting on a capstone course website. The emails and posting shared that we were interested in understanding students' views of design processes through interviews about a series of design process models and a link to a short screening survey that collected demographic and background information for those interested in participating. Participant characteristics are summarized in Table 1.

One researcher conducted a semi-structured interview with each participant about their perceptions of three different design process models. Prior research has shown sharing different visual displays and asking people to describe them to be useful in comparing the effectiveness of the visualizations [16]. Semi-structured interviews supported consistency across interviews in terms of the main questions covered while allowing the interviewer to ask follow-up questions to gain clarity and a fuller understanding of participants' perceptions [48]. All interviews were conducted via Zoom, audio-recorded, and automatically transcribed by Zoom.

The interview aimed to explore participant perceptions of different design process model visualizations. Participants were asked about their perceptions of three design process models: the Center for Socially Engaged Design's socially engaged design process model (referred to as SED) [49], Dieter and Schmidt's first three phases of an engineering design process (referred to as EDP) [50], and Ullman's spiral development of mechanical systems (referred to as SPIRAL) [51]. These models were selected because they all have prescriptive aspects, with SED including social and contextual dimensions in its visualization, and EDP and SPIRAL being commonly found in engineering design course textbooks for undergraduate students [52].

The interviews began by describing the study purpose and the interview format and acknowledging that no single design process model can capture everything a designer should do in every situation. Then, each model was presented individually, allowing participants to focus on one model at a time. Participants were given time to review and write notes about the model before answering questions about it. The order in which the three process models were presented was counterbalanced across the 18 participants, as shown in Table 2, to mitigate any order effects. Interviews concluded with participants viewing all three models together and answering comparison questions. Interviews lasted between 42 min and 72 min, and participants received compensation (gift cards of $20). This study includes the data from the portions of the semi-structured interviews when participants considered each model individually, which was approximately the first three-quarters of the interview. The interview protocol questions for this portion of the interviews are included in Table 3.

Participants were only shown visual representations of the process models, along with the model's name and citation in “Author, Year” format. No verbal or written descriptions of the models were provided to participants. The visual representations presented to participants are shown in Figs. 1–3.

Some participants had, at least potentially, seen some of the process models used in this study before. Thus, at the end of each interview, participants were asked about their familiarity with each of the three models. Participant familiarity is summarized in Table 4, with “Unsure” indicating that participants found the model familiar or similar to something they had seen before but were uncertain if it was the exact process model they had encountered previously.

Our data analysis process drew from King's [53] overview of template analysis and Saldaña's [54] methods of theming data. In addition, our data analysis approach aligned with Walther et al.'s [55] recommendations for supporting validity. For example, we decided to use iterative analysis procedures to support theoretical validation, and we required explicit discussion of topics to support process reliability. Another example is we started with in vivo coding to support communicative validation.

First, we familiarized ourselves with the data by reviewing, correcting, and reformatting the transcripts generated by Zoom. We then used in vivo coding—using participants’ own language as codes [54]—of two participant transcripts (P4 and P17) to generate an initial list of codes. These codes focused on the perceived presence and absence of recommendations for design. An initial codebook, akin to an initial template in template analysis [53], was developed by grouping together similarly coded data related to recommendations for engagement in design work, following a process similar to pattern coding [54]. We then used nvivo [56], a qualitative data analysis software, to apply codes to interview transcripts from a larger subset of participants (P2, P4, P15, and P17). New codes were added to the codebook as additional recommendations were identified. The codebook was refined throughout the process of coding five additional transcripts. The finalized codebook remained constant for coding the remaining nine transcripts. Once all 18 transcripts had been coded, we conducted matrix coding queries in nvivo to summarize participant counts by code and design process model.

In alignment with recommendations by Maxwell [57], we determined the appropriate unit of analysis for our research questions was the participant level, meaning that while some participants named the same characteristic from a model multiple times, we did not inflate the findings by counting the number of mentions, but rather counted the number of participants who reported the characteristic. This focus on the individual as the unit of analysis has also been leveraged in other design scholarship (e.g., [47,58]).

Guided by our research questions, we analyzed participant counts to identify the most salient patterns of discussed recommendations. Our findings focus on codes with relatively high participant counts and those that demonstrated variation across the three models or within a theme. Presenting our findings by using the themes as the main structure follows one of the approaches suggested for template analysis [53]. We further modified code names to explicitly connect participants' discussions with the text included in the visualizations. The refinement of code names and code descriptions ensured alignment with coded participant excerpts.

Procedural validation [55] was supported by regular meetings between the first author—the main coder—and the second author. At the time of analysis, the first author was a graduate mechanical engineering researcher with prior professorial experience as a product development engineer and training in qualitative methods, while the second author is an experienced qualitative researcher and design researcher. Both authors have a close relationship with one of the process models under investigation and have taught undergraduate engineering students other process models not under investigation in this study. Given our positionalities, we paid particular attention to representing participants' views of the models rather than our own. Consistency in our analysis was maintained by requiring explicit participant discussions of those topics represented by our codes. Double coding, or applying multiple codes to the same excerpt, was allowed unless otherwise noted in the findings. Our findings reflect participants' perceptions related to the three models and are limited to those perceptions that participants expressed during the interviews. In the following findings section, terms such as “perceived,” “noticed,” “described,” and “discussed” indicate perceptions conveyed by participants in their interviews. Pragmatic validation [55] is achieved by the transparency of data, including descriptions that allow readers to consider the transferability of the findings to other contexts.

Participants' perceptions of the three design process models were categorized into five areas of design work: (1) starting and moving through a design process, (2) gathering information, (3) prototyping, (4) evaluating or testing, and (5) aspects of focus in a design process. The following subsections discuss participant perceptions of model recommendations across these five areas and, at times, contrast their perceptions of model recommendations with their perceptions of what is absent from the model—including both what is missing (i.e., should be included) and what is simply not present (i.e., not a judgment on if it should be included or not). A summary of participant counts of perceived recommendations is provided in Table 5.

Participants noted recommendations they perceived from the models about how one gets started and moves through a design process, including recommendations on first steps in design and approaches to iteration within a design process.

Participants discussed different approaches to starting a design process based on the three models. In some cases, participants noted the labels within the models as emphasizing a place to start design work. Seven participants perceived they should start design by exploring based on SED, 10 participants perceived starting with a problem based on EDP, and nine participants perceived they should start with requirements based on SPIRAL. In contrast to the specific language in the model, one participant perceived SPIRAL to recommend starting with a problem and one participant perceived SED to recommend starting with a problem. Interview excerpts that serve as examples of how participants perceived ways to start design work are presented in Table 6.

Some participants perceived that EDP and SPIRAL were missing early steps of the design process. For example, gathering information should precede defining the problem, or defining the problem should come before creating an initial design. In addition, some participants highlighted that SPIRAL did not indicate the need for a provided problem definition in order to start with developing requirements. Interview excerpts that provide further elaboration on these ideas are provided in Table 7.

Participants not only discussed the recommended starting points for design but also perceived recommendations for progressing through design, particularly regarding the presence or absence of iteration. They discussed two types of iteration. The first type, referred to as “feedback iteration,” involved going back to a previous point in the design process, i.e., it acted as a feedback loop. The second type, termed “flexible iteration” in line with our previous research [42], involved non-linear movement, including back-and-forth transitions between stages of a design process. While SED includes the language “Iteration—Non-linear move to another stage,” participants perceived both feedback and flexible iteration in this model. Among our participants, SED was the only one perceived to recommend flexible iteration (N = 14). Sixteen participants perceived SPIRAL to recommend feedback iteration, while six participants perceived SED to recommend feedback iteration. No participants mentioned EDP as recommending either kind of iteration. Instances where participants mentioned the term “iteration” without providing specific details were included in the counts for feedback iteration. In addition, although flexible iteration encompasses feedback iteration, instances of flexible iteration were not double coded with feedback iteration in order to highlight differences in participant counts.

Participants also provided critiques of certain process models, expressing that iteration was missing as the process model did not include a recommendation for a designer to revisit previous stages in the design process. Most of the participants (N = 16) perceived EDP to be missing iteration, and a few participants (N = 3) felt the same about SPIRAL. Interview excerpts that serve as examples of how participants perceived iteration are presented in Table 8. Additionally, some participants described EDP as requiring thoroughness or doing the best with the one chance at each design stage; example interview excerpts are included in Table 9.

Participants discussed the recommendations of the process models regarding gathering information, including sources of information. Some participants perceived SED (N = 10) and the EDP process model (N = 5) as recommending information gathering information or conducting research but not the sources or methods from which information could be gathered. Although SED and EDP include “Gather information,” a few participants (N = 2 for EDP and N = 1 for SPIRAL) noted instances where they felt the process model was missing information gathering. Table 10 provides interview excerpts illustrating participants' perceptions of general information gathering recommendations.

Participants also identified specific features of information gathering, particularly regarding information sources, and noticed a lack of recommendations for gathering information from stakeholders in the process models. Some participants perceived EDP (N = 9) and SED (N = 3) as encouraging information gathering from multiple sources. For EDP, participants often referred to the text within the “Gather information” box in the model, while for SED, participants pointed to the text after the “Gather information” undercurrent in the model. EDP contains “consultants,” and three participants described it as recommending gathering information from consultants or engineering professionals. No participants mentioned these domain experts specifically when discussing the recommendations that they perceived from SED and SPIRAL. While SED includes the language of “stakeholders,” all three models prompted at least one participant to describe gathering information from stakeholders (N = 7 for SED, N = 1 for SPIRAL, and N = 1 for EDP), including people that will be affected by a design, end-users, customers, outside parties, and communities. Some participants also perceived a lack of recommendations for gathering information from people, either throughout the process or in specific parts, in SPIRAL (N = 6), EDP (N = 4), and SED (N = 1). Table 11 presents interview excerpts exemplifying participants' discussions on the presence or absence of information gathering source recommendations in the process models.

Participants discussed the inclusion or absence of recommendations for prototyping in the process models. While SED and SPIRAL include the word “prototype” at least once, participants had varying perceptions of the recommendations for prototyping across the process models. Some participants perceived a recommendation for prototyping in SPIRAL (N = 13), SED (N = 6), and EDP (N = 3). Only one participant perceived a lack of prototyping, which they noted in response to EDP. Interview excerpts that serve as examples of how participants discussed prototyping are shown in Table 12.

Some participants perceived recommendations regarding the evaluation or testing of designs, prototypes, products, materialized concepts, or simulations (SPIRAL N = 16, EDP N = 3, and SED N = 2). However, several participants felt the models did not provide recommendations for evaluating or testing (EDP N = 8, SED N = 6, and SPIRAL N = 1). Here, we captured participants' attention to evaluating developed “designs” that were materialized, simulated, or prototyped (which could include prototypes of processes). Interview excerpts that serve as examples of how participants perceived evaluating and testing are shown in Table 13.

Participants discussed multiple considerations as present or absent in the three process models. The salient aspects identified were economic, environmental, social, and technical. Some participants perceived SPIRAL (N = 7) and EDP (N = 2) recommendations for consideration of economic aspects, while no such recommendations were attributed to SED. Relatedly, some participants perceived an absence of considerations of economic aspects from EDP (N = 6) and SED (N = 3). Only SED (N = 1) was perceived by participants to include a recommendation for consideration of environmental aspects, while some participants perceived EDP (N = 5), SPIRAL (N = 2), and SED (N = 1) to lack such recommendations. Some participants perceived SED (N = 15) and EDP (N = 1) to recommend consideration of social aspects, but no participant perceived SPIRAL to recommend it. Some participants described the absence of recommendations for consideration of social aspects (EDP N = 9, SPIRAL N = 6, and SED N = 1). EDP (N = 11), SPIRAL (N = 7), and SED (N = 3) were perceived by some participants to recommend consideration of technical aspects, while participants noted the absence of such recommendations in SED (N = 8), EDP (N = 4), and SPIRAL (N = 1). Table 14 provides interview excerpts exemplifying participants' perceptions of these four aspects.

The findings revealed what stood out to participants about how to engage in a design process based on three different process models. We did not ask participants if specific elements were present or not in the models, but rather open-ended questions about what the models recommended and what important aspects of design engagement were not conveyed by the models. This approach then allowed us to identify salient participant perceptions related to engagement in an engineering design process. Participant perceptions varied across models as well as within models. Also relevant from our findings is what areas of design work many participants did not comment on. We discuss these findings in the following sections.

Our findings revealed several patterns that suggest different design process models make perceptions of certain recommendations more salient than others. For example, the types of iteration most frequently noticed by participants varied by the process model they were considering. Participants described two main types of iteration in our study: feedback iteration, where they could return to a previous stage, and flexible iteration, allowing unrestricted movement between stages. Sixteen participants named an absence of iteration in EDP, 16 participants noted feedback iteration as recommended in SPIRAL, and 14 participants identified flexible iteration as recommended in SED. This lack of iteration or type of iteration pursued in a design process has implications for design outcomes. Iteration is recognized as a fundamental aspect of design [59] and is commonly observed in real-world projects [60]. However, research suggests that beginning designers navigate design processes haphazardly or follow linear approaches [26]. Combining this scholarship with our findings suggests that scaffolded representations of iteration that progressively enable more complex movement—such as the absence of iteration in EDP, feedback iteration in SPIRAL, and flexible iteration in the SED process model—could benefit beginning designers.

Most participants perceived SPIRAL to recommend prototyping, but fewer perceived prototyping recommendations in SED, and none reported recommendations in EDP. How models direct designers to prototype is important; recent design literature recognizes prototyping as a continuous tool or activity throughout the design process rather than a single stage [61–63], yet novice designers often view prototypes as trial builds of final products or for testing functionality, with less common understandings focused on communication, feedback gathering, and decision support [63]. As the uses of prototypes are broad, including enabling communication, informing decision-making, and aiding learning [61,63], models that limit designers' recognition of their uses may cause this tool to be underleveraged to support design success.

The variations we observed in participant perceptions across process models align with our prior work, where participants used dimensions of usefulness, along with iteration and detail level, to distinguish the three different process models [43]. Variation across the models reflects their nature as abstractions, which inherently limits the information they convey [6]. Thus, trade-offs are inevitable when determining what to include in a particular process model. Another example was that participant perceptions of how to start a design process varied, including exploring, starting with the problem, or starting with requirements. Exploring involves divergent thinking, searching for and generating multiple alternatives [64]. Defining the problem focuses on convergent action, describing the end goal [64,65], without emphasizing the importance of exploring potential design problems. Starting with requirements also involves convergent action, listing criteria for a successful solution [50,51]. Crismond and Adams' [26] scholarship of integration study highlighted that beginning designers treat problems as well-structured and prematurely jump to solution generation, while more informed designers delay decision-making to explore and iteratively frame the problem. Of course, how one begins a design process is dependent on the design context such that, at times, beginning with convergent actions will be appropriate, while at other times beginning with divergent actions will be better suited. The segmentation of perspectives across these three process models on how to begin a design process suggests that novice designers' awareness of these different approaches is something that could be facilitated by leveraging multiple process models.

Following the theory of planned behavior [45], participants' perceptions of recommendations (e.g., regarding iteration, prototyping, and starting a process) can influence their intentions and, ultimately, their engagement in design work. Thus, our findings suggest that participants would pursue distinct design paths based on the model or models they rely on to guide their work.

In addition to perceiving different models differently, participant perceptions varied for a single model. For example, three participants perceived EDP to recommend evaluating and testing designs. In comparison, eight participants perceived an absence of recommendations to evaluate and test in EDP. This variety of perceptions for a single process model aligns with prior research showing that engineers can interpret the same model differently [6,42], even when those same engineers assume there is only one “sensible interpretation” [6]. Furthermore, we found this variation in participant perceptions of the same process model when all participants were from the same university and had taken some of the same design coursework, emphasizing that a process model will not be universally interpreted in the same way by all model users.

Our findings highlight several areas of design work where, regardless of the process model, many participants did not comment on a particular area of design work. For example, many participants did not perceive recommendations for information gathering from domain experts and stakeholders across the three models. Gathering diverse perspectives, including from across stakeholders and domain experts, can improve designers' ability to adequately address a range of needs and maximize impact [66–70]. However, prior research has shown that beginning students gather less information and less varied information than more advanced students [71] and that novice teams prioritized domain expert perspectives [66]. The lack of design process models to encourage attention to diverse stakeholders and domain experts throughout design could allow these behaviors to perpetuate and support the false conception that engineering design does not require stakeholder engagement.

The presence or absence of environmental aspects was rarely discussed in any of the three models. Other important considerations were discussed relatively more frequently by participants across the three process models; however, each process model highlighted a certain aspect more than others. Participants most frequently noted consideration of economic aspects in SPIRAL, consideration of social aspects in SED, and consideration of technical aspects in EDP. The lack of explicit attention to certain aspects in a single model or collection of models may have varying impacts depending on the disciplinary norms. Engineering has historically been presented as a purely technical discipline [72], with emphasis on technical analysis [73], technical decisions [74], technicist identities [75], and cost considerations [50,51,76], while social aspects have received less attention. Calls for engineering education to develop students' social and technical expertise persist [77,78], as engineers face challenges attending to social aspects [79] and students lose interest in attending to social aspects [80]. Relatedly, engineering education that integrates technical aspects with sustainability (economic, environmental, and social aspects) has some support [81], although sustainability within engineering often prioritizes economic and environmental aspects over social aspects [82,83]. Explicit representation of environmental and social aspects in design process models may be a tool for shifting engineering design behaviors to attend to these aspects.

One limitation is that participants only had access to visual representations of the process models. This choice ensured consistency in the information and focus on key information, following the principle of salience for visual displays [16–19], but did not allow supplementary text or verbal explanations to support students' deeper understanding of the models. Further, while each of the process models had words within their visual representation, this study did not focus on the specific choices of what words were included but rather on participant perceptions of each model as a whole.

People's perceptions of design process models may be affected by many factors apart from the information contained in the models' visualizations, for example, their thoughts about and previous experiences with design. Our sample had zero non-binary people, consisted entirely of folks in their early twenties, and most (13/18) participants identified as White, Asian, or both. A more diverse sample of participants may have shifted which perceptions of recommendations were most salient. Additionally, participants had limited time to review the models, and their perceptions may have been different with more time or design practice with the models. Lastly, due to word count limitations, we focused on salient patterns and could not include all participant perceptions across the models.

Model users—especially educators—can use our findings to guide the development of engineering design education and training. Our findings highlight several areas of design work that many participants did not comment on, regardless of the design process model they were reviewing. For example, the noticeable omissions we observed included discussions of gathering information from a diverse set of stakeholders and discussions of environmental aspects. These omissions suggest a need for further development of engineering design pedagogy to emphasize the importance of attention to social and environmental aspects.

Design educators can also use our findings to guide how they teach when using any of the three process models explored in this study as they are. For example, relatively few students discussed prototyping when reviewing the SED process model even though the word “prototype” appears in the visual representation of that model. Instructors who choose to use the SED process model could have students develop prototypes at each stage of their process as a way of highlighting the undercurrent of the SED process model.

In addition, our findings can inform updates to the models by their creators or users based on the specific use purpose. While we do not suggest specific updates to any of the models, our findings do support informed iterative development where there is misalignment for model creators, instructors, students, and design practitioners by providing empirical evidence of ways in which the three process models are currently perceived by 18 mechanical engineering students.

Our research reinforces the importance of using multiple design process models for instruction and training, especially for novice designers. In this study, different participants perceived the same process models in different ways. Our findings aligned with prior research that found engineering and project managers had diverse interpretations of the same process model [6]. Thus, different models are needed to communicate the same concept to different people.

Multiple process models are also important for an individual model user. We found that not everything present in the visual representation was translated into meaning for our participants, but certain visual representations were more likely to translate certain meanings. Following the theory of planned behavior [45], which connects individuals' perceptions to their engagement in design work, model users can draw on knowledge of which visual representations are more likely to translate to certain meanings to purposely select model(s) that will support particular kinds of engagement. For example, model users who aim to promote a holistic design practice may strategically select a collection of models that will likely highlight a range of considerations.

In our study, we examined mechanical engineering students' perceptions of three design process models: the Center for Socially Engaged Design's socially engaged design process model [49], Dieter and Schmidt's engineering design process model [50], and Ullman's spiral development model [51]. We focused on participant perceptions of recommendations for design approaches and considerations. Participant perceptions varied across the three process models and varied for a single process model. Thus, while certain process models make some recommendations for design work salient to model users more frequently, we cannot expect a process model to be universally perceived one way. Variation in perceptions across design process models included initiating and progressing through design work, gathering information, prototyping, evaluating or testing, and aspects of focus. Furthermore, our findings highlight several areas of design work that many participants did not discuss. These findings can inform the development of design pedagogy, guide the use and refinement of models, and support the value of using multiple design process models in engineering education and practice.

We would like to thank Diana Karlsson, a member of the Daly Design and Engineering Education Research Group,² for their assistance with interview protocol development. We would also like to thank our study participants and the anonymous JMD reviewers for their constructive feedback.

• This material is based upon work supported by the National Science Foundation under Grant No. 2013410. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

Kelley E. Dugan and Shanna R. Daly are both affiliated with the Center for Socially Engaged Design. Dugan worked as a facilitator at the Center for Socially Engaged Design for several years while completing their Ph.D., while Daly is a co-founder of the center, current director of research and evaluation, and part of the team that developed the socially engaged design process model investigated in this study.

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.