If You Build It, Will They Understand? Considerations for Creating Shared Understanding Through Design Artifacts

Effective communication is a critical skill for designers, largely necessitated by the social dynamics involved in the design process. Designers communicate to converge on a design solution with their team members [1,2], explain design concepts to users [3], and persuade potential investors [4]. During each of these communicative acts, designers seek to build a shared understanding of a design solution with other individuals. The establishing of shared understanding within and outside of design teams has been linked with greater team cohesion and performance [5].

Design concepts are often externalized using design representations such as sketches, computer aided design models, and physical prototypes [6–9]. Buciarelli [10] termed design representations as “boundary objects,” as they facilitate communication between stakeholders possessing different knowledge. Designers have been shown to use representations to provide rationale for design decisions and negotiate design features [11] with other stakeholders. A design representation hence acts as a critical tool to achieve shared understanding with other stakeholders—facilitating the establishment of a “common ground” through which further communication can take place.

However, there is little knowledge about how the act of building mental models of design concepts and the creation of a shared understanding between designers, differs between different design representations. Design representations often differ in modality, and this is known to affect aspects of designer behavior, such as idea generation [12]. As different representations involve varied levels of resources in their creation, both monetarily and cognitively [13,14], it is important to understand whether there are trade-offs that exist between the process of creating a representation and the shared understanding it establishes. In this work, we specifically focus on studying the effect of representation modality on the shared understanding achieved between individuals during communication. With this knowledge, designers can create the best representations to communicate more effectively.

The remainder of this paper is structured as follows. First, we review literature related to shared understanding in design, and how design representations facilitate shared understanding within and outside of design teams. Next, we state our research questions, and the methods used to answer them. We then review the results of this work and situate them within the existing body of literature. The paper closes with concluding remarks and opportunities for future work.

Communication in design is often undertaken with the intention of achieving shared understanding with individuals in the design process [15]. In the context of engineering design, shared understanding is defined as the “similarity in the individual perceptions of actors about… how the design content is conceptualized” [16]. The achievement of shared understanding is associated with improved team coordination [17], reductions in reworks [18], and better team performance [5]. The achievement of shared understanding is especially important in a heterogenous activity such as design, where each actor may possess their own background knowledge and assumptions [10].

Shared understanding is facilitated by external representations such as sketches and physical prototypes [19]. These representations, termed as “boundary objects”, allow for the externalization of tacit knowledge and manifestation of design information [20]. The use of these representations is especially common in early-stage design, as each designer may have their own mental models of design concepts. Externalization of these design concepts through representations facilitates communication and the creation of shared understanding. Recent work has also highlighted how design representations support communication outside design teams across various social situations with different stakeholders, underscoring the criticality of these representations [11]. Representations like sketches and physical prototypes are hence “vehicles” of design information [21], as they allow designers to communicate design concepts and rationale [22] and facilitate coordination and negotiation between individuals [11].

Past research has, however, highlighted that the modality and fidelity of a design representation may influence communication between stakeholders in design. For instance, Haggman et al. [23] found that design concepts presented as low-fidelity foam prototypes were perceived as being more creative, comfortable, and aesthetic than those concepts represented as sketches. Deininger et al. [24] found that when obtaining user feedback, physical prototypes led to longer responses from stakeholders, as compared to virtual prototypes, likely due to the tactility the former offered. It is unclear, however, if these results are associated with how individuals build mental models of design concepts through representations. If this process of building mental models differs between representations, it is likely to also affect shared understanding established between individuals, and subsequently team performance. In this work, we seek to understand how the building of mental models of design concepts, and subsequently, the shared understanding established during a communicative act differs between design representations (namely sketches and low-fidelity physical prototypes).

As argued by Maier [25], communication in design is a process of continued understanding and meaning-making—the communicator must both plan on effectively transmitting the message through an appropriate medium, and the receiver must apply effort to understand the information being communicated. Communication in design is hence dependent on the intentional choices made by designers that may impact how understanding and meaning-making occur between individuals engaged in the communicative act.

In this work, we focus on one of these choices that designers make—the visual medium through which design information is communicated. More specifically, we investigate how differences in the modality of design representations (sketches and low-fidelity physical prototypes) may affect the shared understanding between individuals when communicating design concepts. Sketches have been described as a process of visual thinking and reasoning that allows designers to externalize design concepts during early-stage design [26] and clarify concepts to communicate them to others [27]. Here, we conceptualize sketches as any hand-drawn diagram, graphic, or note that a designer creates during idea-generation activities. Physical prototypes are another commonly used design representation [28] that allow for tactile engagements [29] and are useful for spurring discussion [30]. Synthesizing prototyping literature we characterize low-fidelity prototypes as physical models created out of cardboard, foam, tape, and other materials that require little training or prior experience to begin building with. Importantly, low-fidelity prototypes are constructed quickly, allowing designers to rapidly gain feedback on their ideas with little time or resources invested in the physical model. Each representation offers unique advantages during communication. Low-fidelity prototypes, due to their physicality, allow designers and other individuals to demonstrate design concepts and tactilely engage with the design representation. This offers a unique advantage over static sketches, as dynamic motions may contribute to the creation of more accurate mental models [31]. At the same time, however, sketches are an accessible method of visualizing design concepts regardless of sketching ability [32]; they often contain information-rich annotations [27], and this information may be critical to creating accurate mental models of design concepts.

Results from other fields further suggest that the modality of representation may influence the creation of mental models and shared understanding. For instance, in the field of medical education, prior work has shown that students build more accurate mental models of anatomical structures when learning from 3D physical models as compared to 2D drawings [33]. Similar results have been found by researchers studying how students build mental models of atomic structures, in that physical models promoted a better understanding of spatial relationships [34]. Extending these results to engineering design, it is possible that as compared to sketches, the tangibility of low-fidelity physical prototypes may contribute to individuals building a more accurate mental model of a design concept relative to the mental model being communicated, thereby establishing a greater shared understanding between designers.

In the field of engineering design, no work to date has compared how three-dimensional physical prototypes and two-dimensional sketches differ with respect to the shared understanding achieved between designers during communication. With this knowledge, we argue that designers can better select a design representation, namely a sketch or low-fidelity physical prototype, needed to effectively build a shared understanding with stakeholders and achieve successful design outcomes in the context of two stakeholders trying to communicate their ideas to one another to build shared understanding. Synthesizing the literature above, we aim to answer the following research questions in this work:

RQ1: How does the modality of a design representation, specifically two-dimensional sketches and three-dimensional low-fidelity physical prototypes, affect the shared understanding achieved between a communicator and listener when communicating design concepts?

RQ2: What factors of design representations, specifically two-dimensional sketches and three-dimensional low-fidelity physical prototypes, affect the shared understanding achieved between a communicator and listener when communicating design concepts?

This study consisted of 44 participants (22 men and 22 women). The following inclusion criteria were used: participants must be fluent in English, above the age of 18, and enrolled in either mechanical engineering, aerospace engineering, civil engineering, biomedical engineering, or engineering design. These disciplinary constraints were placed to ensure that differences in knowledge due to participants’ programs, i.e., disciplinary diversity, did not confound our results. The programs were selected based on a review of their courses and learning objectives; the research team concluded that the use of technical jargon or disciplinary discourse would not significantly vary across these fields. Participants in these programs have completed a series of design courses in which they are exposed to design thinking methods, solve various design problems, and are trained to develop design representations such as sketches and prototypes.

Purposeful sampling methods were used for recruitment—the research team printed flyers and sent emails to students enrolled in design classes through their instructors. Snowball sampling methods were also used during data collection—participants were requested to share information about the study with any peers who they knew would meet the inclusion criteria. Recruitment began with graduate students, and the sample was eventually expanded to include junior- and senior-level undergraduate students, 14 participants were graduate students, and 30 participants were junior- or senior-level undergraduate students. Participation was limited to graduate students and upper-level undergraduate students with design experience to limit differences in design skills such as sketching and prototyping; 28 participants identified as White, 10 identified as Asian, 2 identified as Hispanic, Latino, or of Spanish origin, 1 identified as Black or African American, 1 identified as White and Asian, 1 identified as Middle Eastern or North African, and 1 identified as White and Hispanic, Latino, or of Spanish origin.

The study involved participants being paired in dyads. Figure 1 shows experiment groups and procedures. Pairs were randomly generated, and each pair was randomly assigned to one of two groups: (1) prototyping and (2) sketching; 22 participants were in the sketching condition, and 22 participants were in the prototyping condition. At the start, participants were briefed about the study and were told their participation was voluntary, per the guidelines of the Institutional Review Board. Once consent was obtained, participants took a short pre-survey to collect demographic information.

After the pre-survey, the facilitator informed the participants about their design task. Depending on the condition they were assigned to, participants were given the following information and were instructed that they would be communicating their solution using the design artifact that they developed:

Prototyping: “You will now be given a design task to complete. You can use as many materials that are given to you and you have forty five minutes to complete the task. You can feel free to sketch out as many ideas as you want, but you will only be allowed to bring your final prototype with you when explaining your design solution. Your final design can be a single idea or a combination of your ideas generated.”

Sketching: “You will now be given a design task to complete. You have twenty minutes to complete the task. You can feel free to sketch out as many ideas as you want, but you will only be allowed to bring your final sketch with you when explaining your design solution. Your final design can be a single idea or a combination of your ideas generated.”

Participants who were asked to prototype were allowed to create sketches initially, due to the detrimental effect on idea generation when designers create prototypes without prior sketching [35]. Additionally, as physical modeling is more time-consuming as compared to sketching [12], participants in the prototyping condition were given 25 more minutes to complete the task (45 min in total) than those in the sketching condition (20 min in total). Participants were not instructed on how long to spend ideating, sketching, or prototyping and thus were able to spend their time as they found necessary.

Participants were then led to different rooms to complete the design tasks. This was done to ensure that participants could not see each other’s design solutions as they were creating them. Participants were then provided their design prompts; each participant was given a different design prompt than the other participants in their dyad. This ensured that participants presented distinct solutions to one another, and familiarity with the prompt did not affect the formation of shared understanding. The prompts were selected from prior work, and their similarity in terms of their structure, complexity, and solvability was validated [36]:

Prompt A: “Design an automatic clothes-ironing machine for use in hotels. The purpose of the device is to press wrinkled clothes as obtained from clothes dryers and fold them suitably for the garment type. You are free to choose the degree of automation. At this stage of the project, there is no restriction on the types and quantity of resources consumed or emitted. However, an estimated 5 min per garment is desirable.” Shown in Fig. 2 is an example sketch and prototype for this prompt.

Prompt B: “Design an automatic recycling machine for household use. The device should sort plastic bottles, glass containers, aluminum cans, and tin cans. The sorted materials should be compressed and stored in separate containers. The amount of resources consumed by the device and the amount of space occupied are not limited. However, an estimated 15 s of recycling time per item is desirable.” Shown in Fig. 3 is an example sketch and prototype for this prompt.

Each participant within a dyad was provided with the same materials, dependent upon experimental conditions. For the sketching condition, this included pencils, papers, and a ruler. For the prototyping condition, in addition to materials for sketching, participants were also given foam core, cardboard, popsicle sticks, rubber bands, wire, thread, utility knife, scissors, tape, cotton balls, tube cleaners, and hot glue. These materials were selected based on pilot studies and participants' prior experiences with low-fidelity prototyping in undergraduate design courses.

After the design task, participants were given a survey through Qualtrics where they provided a written description of their design problem, solution, and how the solution works. After the participants completed the survey, they were asked to return to the same room with their final sketches or prototypes. All subsequent interactions between the participants were audio and video recorded. Participants were then asked to present their solution to each other with the following prompt:

One participant was assigned as the first communicator and the other as the first listener. The communicator was given 5 min to explain their design problem and solution to the listener, after which the listener had 3 min to ask the communicator any clarifying questions. Next, each participant was given a survey through Qualtrics. The listener was asked to describe the design concept that was presented to them: “In as much detail as possible, please recall the solution that was presented to you, and describe what the solution is, the problem it solves, and how it works in your own words.” The participants then switched roles, i.e., the participant who was previously the listener became the communicator, and vice-versa. Once again, the communicator had 5 min to present their solution, and the listener then had 3 min to ask any clarifying questions. They were given the same post-task surveys.

This procedure was first conducted as a pilot study with 14 participants; 30 additional participants were recruited to complete the study. The only change made to the study after pilot testing was the addition of short semi-structured interviews at the conclusion of the experiment. The first author conducted these interviews individually with 30 of the 44 participants (i.e., those not part of the initial pilot studies). This interview was conducted to gain a deeper insight into how participants communicated and built their understanding of the presented design concepts using their representations. Questions in the interview pertained to participants’ experiences of communicating their design concept using their representation (“Do you think the prototype/sketch helped you in explaining your concept? Why?”), understanding the presented concepts (“Do you think the prototype/sketch helped you in understanding the concept? Why?”) and identifying any challenges during communication (“What do you feel inhibited your ability to explain/understand the design concept?”). Participants were debriefed, thanked, and allowed to leave. The study lasted approximately 1 h and 5 min for the sketching condition and 1 h and 30 min for the prototyping condition. All generated sketches and prototypes were photographed and stored digitally.

Prior work in engineering design has used the similarity between written explanations of design concepts as an indicator of shared understanding [2]. We use a similar approach in this work and use the similarity between the communicator’s and listener’s explanation of a design concept as a measure of shared understanding. A variety of methods were explored to quantify the similarity between descriptions of design solutions. Prior work by Fu et al. [2] used a combination of Latent Semantic Analysis (LSA) and cosine similarity to quantify the semantic similarity between design team members’ conceptualizations of a design idea. The limited size of our corpus, however, made it challenging to effectively use a corpus-based measure like LSA to find semantic similarity between documents. We then explored using a combination of pre-trained word embeddings and cosine similarity to measure semantic similarity. Specifically, we used the GloVe word embedding (trained on a corpus of Wikipedia articles) to build a vector space where semantic relationships between words in documents are represented. Cosine similarity was then performed on this vector space to measure the similarity between texts based on the angle between two vectors. Human judgments of similarity were used to ensure that the algorithm was accurately quantifying semantic similarity. Specifically, ratings were performed to agreement by two raters using the semantic anchors scale developed by Charles [37]. However, no correlation was found between human judgments and the similarity scores yielded by the algorithm (Spearman’s ρ = 0.105, p = 0.676), implying that our automated approach did not accurately measure semantic similarity between texts. This might be due to the corpus the word embedding was trained on, as it may not include semantic relationships between commonly used words in engineering design. While human judgment scores could have been used for the analysis, these scores do not capture the complexity of the presented design concepts. For instance, in a dyad, a simple design concept (with lower information content) may yield a higher similarity rating than a much more complex design concept (with higher information content), since the listener has to store a smaller amount of information while creating their mental model.

As a result, our final process to calculate the similarity between descriptions relied on the method developed by Nandy et al. [38], as this approach allows us to capture both the similarity between and complexity of ideas. To compare the similarity between design concepts, Nandy et al. propose a method where each design is represented as a functional structure; the functional structure is then decomposed to either a matrix or network, and finally, the similarity is calculated using the appropriate vector- or network-based similarity approach.

Extending this approach to our work, we used the explanations of an idea from each participant in a dyad as the basis for creating the functional structures of each idea, the corresponding networks, and then the similarity between them. For greater clarity, consider a dyad of two participants A and B, and the calculation of the similarity between the two participants for the idea that Participant A developed. During the study, Participant A completed a survey after the design task, where they were asked to provide an explanation of their design concept. Then, once Participant A presented their idea to Participant B, Participant B was asked to provide an explanation, in their own words, of Participant A’s idea. This provided us with both participants’ explanations, and hence their mental models, of the same design concept.

First, the functional structure taxonomy by Stone and Wood [39] was used to convert each participant’s explanation of their own idea to a functional structure. Rather than using participants’ verbal explanations to one another, the written explanations from participants in the surveys were used to create the functional structures (and networks) of each design concept. This was done because, during verbal communication, participants may have used non-specific, demonstrative language with gestures to explain their design concepts (such as saying “this moves like this” while interacting with their representation). The functional structures were created by the first author and validated with the help of the third author, who has extensive experience in the generation of functional structures [40–42]. The first and third author reviewed 20% of the created functional structures, the consistency in their creation, discussed how accurately the functions and flows represented the written explanations, and ensured that the functional structures did not include and repeats or reiterations. Any changes were then discussed and implemented.

Once the functional structures were generated, they were converted to networks to quantify similarity. A network-based approach was chosen over a vector-based one as the former allows for the capturing of more specific information during the calculation of similarity. A network-based approach captures the flows of functions, in addition to, specific components that provide a specific flow or carry out a function. For example, consider a design solution with the sentence “An iron press would provide heat”—a network-based approach would be able to indicate the function (Transmit), the flow (Thermal energy), and the component where the flow is coming from (Iron press), as each of these would be interconnected nodes in the network. However, in a vector-based approach, only the function (Transmit) and flow (Thermal energy) would be represented in the matrix. In each network, each function, flow, and component is represented as a node, with the connections between each being represented as edges. Edges were weighted in each network to account for repeated functions, flows, and components, resulting in weighted networks that represented each design concept. Because the networks were derived from the functional structures, any repeats, or reiterations, were excluded from the networks’ edge weights. Figure 4 shows the functional structure and weighted network for the following design concept:

To calculate the similarity between two weighted networks, we used a weighted Jaccard Similarity approach. This approach is an extension of the Jaccard Similarity metric, which has been previously used to evaluate design similarity [38], to weighted networks. Consider the networks G and H shown in Fig. 5. The calculation of the weighted Jaccard Similarity begins with the union of the connections between the two networks. In the example, the vector of the union between the networks G and H would be represented as $(G∪H)={(A,B),(B,C),(A,D),(C,D)}$⁠.

In the context of this work, the networks G and H above would correspond to the weighted networks that represent a communicator’s and listener’s mental model of a particular design concept. The weighted Jaccard similarity between these networks hence represents the similarity or achieved shared understanding between the two participants. Therefore, two similarity values were calculated for each dyad, one for each design concept presented.

Idea complexity may affect shared understanding, especially considering the approach used to calculate shared understanding in this work. Consider the examples of Dyads A and B, shown in Fig. 6. While the shared understanding in Dyad A is higher than that of Dyad B (as exhibited by the similarity of the networks), the number of nodes and edges is far greater in B than A. It is possible that the higher similarity in Dyad A is confounded by the lesser amount of information needed to be held in the listeners' working memory to construct a mental model, as compared to Dyad B. In the context of this work, it is then important to measure and control for any differences in the complexity of ideas (i.e., the amount of information being communicated) to isolate the effects of the type of representation on shared understanding.

RQ1: How does the modality of a design representation, specifically sketches and physical prototypes, affect the shared understanding achieved between the communicator and listener during design communication?

All statistical analyses for RQ1 were performed on R version 4.1.1. In addition to p-values, we also report the effect sizes for the statistical test performed. One dyad was removed from the dataset due to incorrectly interpreting the design prompt. No outliers were found in the values of shared understanding.

To isolate the effect of the representation (prototypes versus sketches) on shared understanding, we investigated the effect of idea complexity on the achieved shared understanding within dyads. A Spearman correlation found a significant negative relationship between idea complexity and shared understanding (ρ = − 0.428, p < 0.005), implying that ideas that were more complex were associated with lower levels of shared understanding (Fig. 7). This is expected though—as compared to a complex idea, a simpler design idea requires less information to be processed to create a mental model, thereby increasing the chances of attaining a higher level of shared understanding between designers. As a result, idea complexity was included as a covariate in the analysis, and an ANCOVA was run to assess the effect of the representation on shared understanding, as will be reviewed in the Results section.

RQ2: What factors of design representations, specifically sketches and physical prototypes, affect the shared understanding achieved between the communicator and listener during design communication?

This study used a follow-up explanatory QUANT → qual approach. That is, the qualitative results are used to contextualize and explain the quantitative findings [45]. To understand what factors of design representations affect shared understanding, interviews were conducted with 30 out of the 44 participants. Interviews lasted approximately 15 min. Scribie, a third-party transcription service, was used to transcribe the interview, and the transcripts were then validated for accuracy; the first author read through the transcripts while listening to the interview audio. One participant’s interview was removed from the dataset due to a technical issue during the recording. All transcripts were anonymized using each participant’s unique ID.

The first author coded each transcript using an open and axial coding approach [46] and an abductive coding paradigm. An abductive paradigm involves the use of prior, relevant work during the coding process while acknowledging that the observations may go beyond prior work [47]. This is contrary to a grounded theory approach, which generates a framework at the end of the qualitative analysis. Specifically, in this work, we leveraged the works of Carlile [48] and Broberg et al. [49]. Both works explore the characteristics of boundary objects in engineering design and explore how these characteristics affect communication between individuals. Carlisle [48] argues that effective boundary objects share three characteristics: they establish a shared language between individuals; they can help individuals jointly transform knowledge and can help negotiate differences between individuals. Broberg et al. [49], who investigated the characteristics of boundary objects in participatory design, explored not only the characteristics of effective boundary objects, but also the effect of the environment in which the object is used, and the actors using the boundary object. Not only should a boundary object have built-in affordances to facilitate communication, but the actor also decides how the object is used to achieve certain communicative outcomes. Using these works helped identify which characteristics of sketches and low-fidelity prototypes contributed to effective communication and shared understanding between participants.

RQ1: How does the modality of a design representation, specifically sketches and physical prototypes, affect the shared understanding achieved between the communicator and listener when communicating design concepts?

To identify if the shared understanding differed when sketches and low-fidelity physical prototypes were used as communication tools, a one-way ANCOVA was run with the representation as the independent variable, shared understanding as the dependent variable, and idea complexity as the covariate (Fig. 8). The result was not significant (F(1,40) = 1.151, p = 0.29), and negligible effect size was also observed (partial ω² = 0.02). This implies that shared understanding achieved between participants in a dyad did not differ when sketches and low-fidelity prototypes were used as communication tools.

RQ2: What factors of design representations, specifically sketches and physical prototypes, affect the shared understanding achieved between the communicator and listener during design communication?

Our second research question sought to understand the factors that affect the shared understanding achieved between a communicator and listener. Our research question primarily focused on identifying the factors of design representations, namely sketches and prototypes, that shaped shared understanding between individuals. However, these representations do not exist in isolation and are manipulated by human actors to achieve communicative outcomes. This is acknowledged in the work of Broberg et al. [49], whose findings were kept in mind during the abductive coding process. We hence distinguish the determinants of shared understanding into two categories—representation-level factors and actor-level factors, as listed in Table 1.

Representation-level factors: Through our qualitative analysis, we identified two factors at the representation level that shape the level of shared understanding developed between communicators and listeners. The first factor, “clarity of features in representation,” relates to the representation’s ability to capture the information that individuals need to establish shared understanding. This theme explores how the inherent features of sketches and low-fidelity prototypes lend themselves to being vehicles through which design information is embedded and communicated. Multiple participants alluded to challenges in building mental models of design features not visualized in a representation. For instance, one participant explained how, because a certain process was not represented in a sketch, they could not build an understanding of the working of the solution:

Information not being visualized in a design representation not only leads to challenges in understanding the design solution being presented, but participants may also make assumptions and potentially build incorrect mental models. As noted in the following quote, one participant stated that they assumed how long the conveyor belt in the presented solution was, since the sketch did not present the information they were looking for:

Challenges with building shared understanding due to design features and details not being visualized were especially salient when participants used low-fidelity prototypes to communicate. While low-fidelity prototypes can be created using easily accessible materials such as craft supplies, these may not accurately represent features of a design solution: “There was cotton balls at the bottom of one of the tubes, and I wasn't really sure what that was supposed to convey…I don't really understand what that was about.” In contrast, multiple participants described how sketches allowed them to show a certain level of detail which they could not have achieved with a low-fidelity prototype, which then aided communication and shared understanding:

Additionally, even when features weren’t drawn out accurately, participants stated they either could have compensated or did compensate by adding annotations to communicate what certain visuals represented. This then aided in listeners’ understanding of the design solutions, as seen in the following quote:

The second factor “dynamic and step-by-step demonstrations” relates to a representation's ability to facilitate tactile motions and demonstrations of features. Multiple participants explained how design solutions were easier to understand when the explanations were structured as user interactions. One participant articulated how a “step-by-step process” of the design solution operating helped her better understand the solution presented, while another participant, when asked what made the solution presented to him easier to understand, said: “He [the other participant] did it step by step, kind of like how I did it, so like hitting the switch, watching it go through, so each step, he kinda just walked me through…”. It is likely that low-fidelity prototypes were especially useful to demonstrate such user interactions, as evidenced by participants’ quotes. One participant, when asked to state the factors that helped them understand the design solution stated that the “dynamic aspect of the model was super helpful.” Another said that “having her [the other participant] moving her machine on its conveyor belt, and then also pointing out the different parts specifically” helped in understanding the design solution.

Actor-level representations: Through our qualitative analysis, we also identified two factors at the actor level that capture the individual behaviors of both listeners and communicators and their effect on shared understanding during design communication. Design communication specifically refers to a designer informing a listener of their design concept. The first factor that we identified, “Design problem and rationale definition”, captures how outlining of the design problem and rationale was critical to establishing a shared understanding of the design solution during communication. Listeners articulated that understanding a design solution was challenging without an explanation of the constraints within which the solution was created. For instance, one participant stated how, because the order of operations in relation to the prompt was unclear, they had trouble understanding the presented design concept: “I didn't know what her design problem was, like what is written on the paper. But in my idea, the compression should happen after the trash is accumulated not before it.” Another participant stated that while they eventually understood the problem in the later stages of their communication, understanding the solution was challenging “since it was difficult to see what the problem was like right at the beginning of the interaction that we had.”

Similarly, listeners also stated that an insight into the communicator’s problem-solving process, and their justifications for design decisions, was important to build a mental model of the presented design solution. One participant explained that the solution presented to them could be understood easily because the communicator: “had a lot of reasoning about why you would use the photo sensor, why you used the weights, stuff like that.” Listeners sought to learn both what the design solution was and why it was designed that way as they built their mental models of presented solutions. In the absence of important design rationale, listeners perceived greater challengers in building a shared understanding. This is seen in the following quote from a participant, when asked about what they might change in the communicator’s explanation to make the solution easier to understand: “I think talking about the reason that there are three different holes and how size plays a role in figuring out what the material is made of would have made a difference.”

The second factor “Background knowledge” explores the effect of individual actors’ knowledge on the building of shared understanding during communication. Multiple participants stated that because they were communicating with other engineering students, they assumed the listener possessed some level of pre-existing knowledge. This may have made it easier for some participants to both communicate and understand technical information, as seen in one participant’s explanation of the factors that helped them understand the presented idea: “I think the fact that I was an engineer and I understand the mechanics of it.” However, the absence of background knowledge was also found to impede shared understanding. We hypothesize that this could be due to an assumed level of background knowledge on the part of the communicator; in other words, the communicator may have assumed the listener shared a similar background knowledge of the concept and therefore did not provide critical details to the sketch or prototype. For example, when asked why a certain part of the design presented was hard to understand, one participant stated that a lack of knowledge was the determining factor, despite the visual information the prototype provided:

This study sought to understand how three-dimensional low-fidelity prototypes and two-dimensional sketches differ with respect to the shared understanding achieved between designers during communication. To achieve our objective, we conducted a mixed-methods study with 44 participants to understand if representation modality affected shared understanding, and the factors that shaped shared understanding during communication.

Our results showed that that low-fidelity prototypes and sketches did not significantly differ in terms of the shared understanding of the artifacts facilitated within dyads. This is contrary to prior literature on instructional design which has found that physical representations lead to individuals building more accurate mental models of presented information [33,34]. However, these differences may be attributable to the types of representations used in these studies. The physical models used in this body of work were often 3D printed models of anatomical and atomic structures and possessed the same amount of information as a 2D diagram of the structures. This contrasts with sketches and low-fidelity prototypes, as they more often than not contain different levels of detail and information, which would have then shaped the shared understanding developed between participants.

In addition to our quantitative results, we also identified four factors at the representation- and actor level that influence how shared understanding is built between individuals during design communication. At the representation level, we observed the clarity of representations, or how well a representation captures design information, to be an important determinant of building shared understanding. Specifically, listeners explained that they experienced challenges with understanding design solutions when certain information was missing from the representation or visualized inaccurately. Carlile et al. stated that a key characteristic of an effective boundary object is its ability to “establish a shared syntax or language for individuals to represent their knowledge” [48]. In other words, Carlile argues that a boundary object (a design representation) should be able to capture what all individuals communicating have “at stake.” In our work, we observed that when listeners sought to understand a greater level of detail of the design solution during communication, sketches were perceived as the stronger communication tool than prototypes. On the other hand, however, when the aspect of communication “at stake” was a deeper understanding of the working, or physical motions of the design solution, low-fidelity prototypes were perceived as being the superior communication tool owing to their physicality. We highlight these results do not posit that one representation is a better communication tool than the other. Rather, we argue that it is important for designers to understand the perspectives listeners might come in with during communication, and what information they might seek to learn during the communicative process. Without an understanding of what a listener might seek to gain from communication, designers may create representations that do not accurately represent the information a listener might be looking for. This may then impede communication and the building of shared understanding.

It is important for designers to also build an intimate understanding of the background knowledge that listeners may or may not possess, as noted in the actor-level factor of “Background knowledge.” Similar to prior work by Cash et al. [50], we also find that background knowledge, or prior experience and skills, plays a role in how shared understanding is established between individuals. While the sample in this study was homogenous, in that participants were engineers interacting with other engineers, this may not always be the case. Designers are equally likely to interact with individuals with little to no technical background [51], thereby necessitating the need for effective communicative strategies that can bridge the gaps caused by differences in background knowledge. At the actor-level, we also noted how the outlining of the design problem and rationale was important for listeners to build a shared understanding. In his analysis of design presentations, Swales [52] posited that failing to outline the design problem when explaining a design solution may be a problematic communicative pattern, as the problem acts as the context in which the solution was generated. This aligns with the results from our work—when communicators did not outline the design problem clearly at the beginning of their explanations, listeners were often unsure about how the solution achieved the objectives of the given problem, thereby leading to challenges in building a mental model of the design solution. Prior work has also highlighted the importance of design rationale in communication, as it represents the design space and alternatives explored [53–56]. The importance of design rationale is also demonstrated in our work—as noted in participants’ quotes, listeners highlighted that gaining an understanding of both what the solution was and why it was designed that way, was important for building mental models of design solutions.

Some limitations of this work are as follows. First, the sample of this study was limited to dyads of students enrolled in undergraduate and graduate programs at the Pennsylvania State University. Future work should explore how these results translate to design practice, where designers often communicate and seek to build shared understanding in teams of larger sizes. The quantitative results of our work are also limited by the fairly small sample size of the study (44 participants), and future work should use larger samples to validate the results of this research. As highlighted in Sec. 6, future work should also explore how designers communicate with individuals who come from drastically different backgrounds, as unique results may be identified based on the context of communication. In the current work, two distinct design problems were selected based on prior work demonstrating the similarity of these problems with regard to complexity and problem structure; however, it should be investigated if either of these problems are more or less easily sketched and/or prototyped. If, for example, one of the prompts lends itself more readily to low-fidelity prototyping, this could affect the outcomes of the study and therefore is a limitation of the current work. Further, the current work asked participants to verbally present their ideas to their partners along with their sketches or prototypes. They were also asked to describe their own perceptions of their partner’s ideas via text. In future work, researchers should evaluate the effect of these modalities on the ability of the participants to formulate a more cohesive mental model by comparing verbal pitches and text-based responses to sketches. Allowing participants to both sketch and describe via text their own mental models of ideas may further crystalize mental models possibly affecting the sharedness of mental models between participants. Integrating member-checking into the experimental design, in which communicators would evaluate or rate the similarity of the listeners’ descriptions to their own description of the original idea would significantly bolster the work and allow the research team to triangulate the validity of the similarity scores. This method should be integrated into future studies exploring this phenomenon. Additionally, this work focused on only two representations, namely sketches and low-fidelity prototypes, and their effect on how shared understanding is built in early-stage design. It is possible that rather than randomly assigning participants to sketch or prototype, allowing designers to choose their representation may affect the formation of a shared understanding. Future work can extend our research to examine how designers’ choices of design representation, including representations of higher fidelity, differ in their effect on shared understanding further in the design process and in longer design projects. Finally, the functional structure taxonomy does not currently account for new processes and technologies such as artificial intelligence, and future work is necessary to update the model.

This research sought to learn how representations of different modalities, namely sketches and low-fidelity prototypes, differed in how they helped build a shared understanding between individuals during communication, and what factors of these design representations affect the building of shared understanding. Our results found no significant effect of the representation condition on the shared understanding developed between participants. Through a qualitative analysis, we identified two factors each at the representation level (clarity of features in representation and dynamic and step-by-step demonstrations) and actor level (design problem/rationale definition and background knowledge) that shape the shared understanding built between individuals during communication. Ultimately, this work seeks to give designers the knowledge needed to build a shared understanding with other individuals during communication through design representations and drive further research into the effect of design representations on communicative processes.

There are no conflicts of interest.

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