1 Introduction

Conceptual design, in which ideas are generated, is one of the most important tasks in engineering product development cycle (Wang et al. 2002). Various theories have been proposed as to how idea generation takes place; few of these are as follows: convergent–divergent theory (Guilford 1967; Zhang and Sternberg 2005), analogy and inspiration (Eckert and Stacey 2000; Vattam et al. 2010), Wallas model (Wallas 1926; Torrance 1988), Barron’s psychic creation model (Barron 1988; Plsek 1996), Osborn’s seven-step model (Osborn 1963), and the Creative Problem-Solving (CPS) model (Isaksen and Treffinger 1985; Parnes 1992). Although these theories and models are different from one another, an underlying assumption in most is that designers search for new, appropriate, and improved solutions. Thus, searching for ideas is a major activity of designers during conceptual design. Many researchers hold this view explicitly. For instance, De Silva Garza and Maher (1996) and Gero and Kazakov (1996) argue that design is a phenomenon involving search of new ideas, and Benyon and Imaz (1999) expressed also similar views, discussing ideation as phenomena and proposed usage of experimental cognition in understanding ideation. Roozenburg and Eekels (1995) argue that search of ideas is similar to idea finding, since both are divergent processes, where many ideas need to be considered before selecting the best ones. According to Stal and Turkiyyah (1996), creative design involves generation of new search spaces.

Search is used in finding an improved design within a given design space (Gero and Kazakov 1996). It is also considered as a process that finds new ideas (points) that are better than previous ones. Thus, search is a process of finding new or improved designs in a design space. Footnote 1

Gero (1990) established that there are mainly two broad classifications of design—routine designing and non-routine designing. In non-routine designing, all the variables, which specify designs, are already available. The space of possible designs is known, and the space can be predicted, constructed, and evaluated. Search is required only to locate the best design in that space. On the other hand, in non-routine designing, search need to be conducted for both ‘best’ space of possible designs as well as the ‘best’ design possible from the space. Thus, search is useful for routine design, and exploration is useful for non-routine design. Gero and Kazakov (1996) used shape grammar to represent knowledge. They formulated both routine and non-routine, or creative design problem using state space and extended state space shape grammar. They then produced two algorithms, which have potential of producing better designs. Gero and Kazakov (2000) argue that in a function–behavior–structure mode, exploratory processes have potential for increasing the space of each of these three individual design spaces (i.e., function, behavior, and structure spaces).

Gelsey et al. (1998) observed that automated search of a space of candidate designs is an attractive way to improve the traditional engineering design process. Langdon and Chakrabarti (1999) discuss some of the positive effects and some limitations of exploration. They argue that effective support for conceptual design should help designers to obtain a thorough overview of the solution space, as well as a detailed understanding of its individual solutions. However, Bryant et al. (2005) state that there are only few computational tools that exist to assist designers in the conceptual phase of design. For example, Bryant et al. (2005) state that there are only few computational tools that exist to assist designers in the conceptual phase of design. They developed concept generator, an automated design tool. They found that the tool is a promising one for supporting designing. Kurtoglu et al. (2005) discuss a methodology to extract design knowledge from an online library of components in the form of grammar rules. They mention that initial implementation of forty-five rules that have been compiled from 15 components extracted of three products. In another interesting work, Nagai and Taura (2006) proposed a design synthesis process and stated that it is key to creative design. They proposed through design study several primitive activities in ideation, such as, concept abstraction and concept blending and discussed how these are related to creativity. Potter et al. (2003) argue that even though design textbooks suggest a number of techniques for supporting design synthesis, synthesis of solutions still predominantly depends on the creativity, skill, and experience of the designers. While it is often argued that factors such as quality of the ideas generated during conceptual design influence creativity of the outcomes of a design process, it is unclear as to what is meant by quality of ideas, and there is little empirical study to establish what influence quality of ideas and experience have on creativity (Potter et al. 2003).

This research thus aims at understanding the process of search and its effect on conceptual design using two sets of design experiments. For this, we categorize the types of searches that occur during problem understanding, solution generation, and solution evaluation and selection, and map them with the creative outcomes of these experiments. Additionally, we also study the effect of other important factors such as experience of participant designers, duration of the design experiments, and creativity and problem-solving style of the designers on the design outcomes, especially creativity of the final solution.

2 Understanding the process of search through design experiments

To understand the process of search, two sets of design experiments (set of design sessions) are conducted. They are as follows:

  1. (i)

    Design experiments with compulsory use of design methods

  2. (ii)

    Design experiments without the use of any design methods

2.1 Design experiments with compulsory use of design methods

The first set of experiments, henceforth called ‘initial design experiments,’ involved eight design sessions, carried out by two groups of designers, three in each group, using four different design methods: Brainstorming, Ideal design, Functional analysis, and Innovation Situation Questionnaire (Chakrabarti 2003). All these designers had formal training in designing at the postgraduate level; some had experience of designing products in industry.

2.1.1 Experimental setup

The experiments were conducted in a laboratory setting, where the steps of each method were provided in a printed sheet, and blank sheets were provided to the designers to work on and express their outcomes. The designers were asked to discuss audibly while designing, so that their verbal expressions could be captured on videos. Two problems (Table 2) were interchangeably used, and both the groups solved both the problems using all the four methods. Even though there was no time constraint, the average duration of each conceptual design session was about an hour. All sessions are videotaped and subsequently analyzed using video protocol analysis.

Two design problems have been used in these design experiments, as shown in Table 1. Initial Problem 1 (IP1) is related to the design of a system for locking in which one do not require key or remember numbers or words to use it. Initial Problem 2 (IP2) is related to finding out suitable ways of removal of dry leaves from a given place. Designers are asked not to reuse any solution from previous experiments, if any.

Table 1 Initial design experiments (average duration is 60 min)
Full size table

2.1.2 Analysis and results

Each utterance (each statement that the designers made during designing) of each designer in the video protocols of these eight (4 × 2) sessions (on average 300 utterances per experiment) is separated and enlisted. The contents of the transcribed protocols from each session are categorized into one of these three phases: (i) problem understanding phase, where the given problem is analyzed, and requirements, related problems, and constraints are identified; (ii) solution generation phase, where potential solutions are generated; and (iii) evaluation and selection phase, where evaluation criteria are determined, the ideas generated earlier are evaluated, and final design outcome is selected.

In these protocols, for the solutions generated, we identified all the solutions generated during each design session and compared their similarities with the solutions previously generated in the same session. We observed, as shown in Fig. 1 (for IP1), that designers find new solutions:

Fig. 1
figure 1

Transcript of a Brainstorming session by three designers while solving IP1

Full size image

  • either by searching for similar solutions,

  • or by generalizing a previously generated solution,

  • or by detailing a solution.

The right-hand column of Fig. 1 shows the results from this analysis.

Next, two individual researchers, each with over three years of research experience who had earlier transcribed many video protocols, are asked to cluster all the solutions generated in two of the design sessions, based on the similarity of the ideas. These researchers found that four distinct types of solutions (see ‘Appendix 1’ for details) were generated by the designers:

  • Type 1: A designer generates a solution by extending a previous solution by adding details to it. For instance, while solving IP1 (Table 1), a designer expresses that locking could be done remotely using a remote locking device. Next, the designer adds that the remote locking device would have a particular range of frequency. This is an instance when a designer adds details to a solution. We name this pattern as ‘detail solution search.’

  • Type 2: This occurs when a designer finds a specific solution within a generic solution space. For instance, for IP1, the solution of using one’s own ‘DNA’ that is ‘specific to a person’ as a means of identifying that person as a ‘key’ (see Fig. 1) is an example of this pattern. Here, the idea ‘DNA’ can be taken as a solution that belongs to a space in the generic idea of ‘specific to a person.’ We name this type of solutions as ‘local solution search,’ as the solutions seem to be localized within a generic design space.

  • Type 3: This occurs when a designer finds a solution that is different from all the solutions found before in the session, and is generic in nature. However, such solutions are known to exist as potential solutions. For instance, a designer generates ‘passwords that are common to a person’ (see Fig. 1) as a solution to IP1. Here, the designers were generating ideas that belong to the generic idea of ‘specific to a person’ (eye balls and thumb impressions ideas belong to this space), and later, they started generating ideas that belong to ‘passwords that are common to them’ (personal numbers idea belongs to this space). We name this type of solution as ‘global solution search.’

  • Type 4: This occurs when the designer finds either a novel idea or a new application of an existing idea. This pattern is similar to global solution search, except that the solutions did not exist before and are not known as potential solutions for similar problems solved earlier. Though found rarely in transcripts, it does not fall into any of the above three categories. For instance, in solving the locking problem (IP1), a designer generates ‘we (would) have something which traces the thought coming in our mind, we can (think) unlocking the system so it will unlock. So, if we can trace the thought and if we say unlock, it will unlock.’ Thought tracking devices are still in a research stage and have never been used for a locking device. We call this as ‘new solution search.’ A ‘new solution search’ is activated when a designer generates an idea (the same is also applicable to finding problem or evaluation criteria), not previously known to the designer. These kinds of solutions are potentially novel solutions. However, these have been generated during the idea generation phase of these design experiments and are yet to be evaluated for their suitability as a selected solution. Often, these solutions are new applications of an existing technology in another field. Finding a search space of this kind that is not known at the starting point of the design process should increase the possibility of finding optimum designs (Stal and Turkiyyah 1996) and influence the occurrence of other kinds of search. The low frequency of occurrence of this kind of search could be attributed to the difficulty of identifying these spaces. Stal and Turkiyyah (1996) stated that generation of new search spaces influences the creative outcome of a design process.

We express these findings in the form of a flow diagram, see Fig. 2. As presented in Fig. 2, ‘global search’ represents search in a space (global) that is less specific than that of the local and detailed spaces.

Fig. 2
figure 2

Representation of a ‘design space’ and various types of search. The bold arrows show typical processes for finding potential solutions by a designer

Full size image

Figure 2 represents that a designer, while generating solutions, finds different potential solutions in different design spaces, mainly through new or global searches. Once a potential solution is found, the designer starts exploring the same design space both in breadth and in depth, and often finds several other potential solutions though local and detailed searches. In the later part of this paper, as shown in Fig. 4, we see that finding new and global searches helps in finding many local and detailed searches.

As explained above, ideas for solutions generated in conceptual design are a result of searching different design spaces, and each of these ideas can be classified into one of these four types: ‘new solution search,’ ‘global solution search,’ ‘local solution search,’ and ‘detail solution search.’ Further analysis of problems and evaluating criteria identified shows that similar kinds of search also occur in problem understanding and solution evaluation phases (see ‘Appendix 1’ for details). Consequently, a general pattern of search in conceptual design emerged: designers employ twelve (4 × 3) different types of search (i.e., four types of search in each of the three design phases) in order to search for problems, generate solutions, and identify criteria for evaluation and selection of solutions.

3 Design experiments without the use of any design methods

Since each initial design experiment involved obligatory use of a method, it is possible that these methods had influenced the kinds of search that occurred in these experiments. To ascertain the frequency of different kinds of search taking place in a generic design process, and to propose general conclusions based on them, another set of design experiments was conducted without prescription to use of any design method.

3.1 Experimental setup

In these experiments, eight design experiments were conducted, with four novice designers and four experienced designers, all of them working individually. Two types of problems were provided to the designers to solve—one requires engineering design knowledge and the other requires only general design ability. One problem was given (P1) to two novice and two experienced designers and the other (P2) to the remaining two novice and two experienced designers. The two problems solved are given below:

  • Problem 1 (P1- requires general design ability): Design a handheld utensil cleaning system for urban middle-class women.

  • Problem 2 (P2- requires engineering design knowledge): Design a drilling machine that can drill a hole of variable diameter in a metal block in any direction and is capable of changing its direction even while the drilling is going on inside the block.

All designers had undergone a four-year engineering course (mostly mechanical engineering) at the undergraduate level and a formal two-year course in product design at the postgraduate level. The experienced designers had between two and eight years of design experience in design firms, where they designed tangible products.

Designers participating in these experiments were different from those in the first set of design experiments, in order to reduce the ‘subject effect’—the influence of the subjects used on the findings. The problems in these design experiments were interchanged, as shown in Table 2, among the designers to eliminate the ‘problem effect,’ so that the designers could not reuse solutions from their previous design experiments. The experiments were conducted in a laboratory setting without any intervention. The experimental room was a large mostly vacant room, without any clutter, to avoid designers being getting inspired with the products in the room. Blank sheets were provided to the designers to work on, or express their solutions. Each designer was asked to express audibly while designing, and think aloud protocol (Jaaskelainen 2010; Sarkar and Chakrabarti 2013) was used. A video camera was used to capture verbal expressions. The duration of these experiments varied between 47 and 184 min (see Table 2).

Table 2 Searches in the main design experiments
Full size table

3.2 Results of the protocol analysis of these design experiments

The transcribed protocol from each design experiment was categorized. Apart from the twelve categories discussed in the previous section (viz. new problem search, global problem search, local problem search, detail problem search, new solution search, global solution search, local solution search, detail solution search, new evaluation search, global evaluation search, local evaluation search, and detail evaluation search), the following categories were used to categorize the remaining portion of the transcripts: ‘agree,’ ‘disagree,’ ‘clarification,’ ‘method clarification,’ and ‘selection’ (see ‘Appendix 1’ for details). The results of categorization are shown in Table 2, and further analyses are presented in the next subsections.

The inter-coder consistency was 88 %, which was assessed by comparing a large portion of coding done by the authors and two different coders, each with three years of transcribing and coding experience. After discussion between them, the entire code, as originally coded by the authors, was accepted by the two coders.

3.3 Analysis

With the results tabulated in Table 2, various analyses are conducted to seek deep relationships, if any, among search, idea generation, time, experience, creativity, and creativity style (using KAI (Kirton 2012)).

3.3.1 Effect of search in new design spaces

In the introductory part of the section, we mentioned that once a designer jumps into a new design space through a new or global search, other searches such as local and detail search follow. Now, we investigate as to whether finding new or global design spaces helps in finding more ideas in that space, or helps in problem solving leading to development of more appropriate ideas. To do this, we ascertain the relative influence of each kind of searches on one another. Standard Pearson’s correlation (Gravetter and Wallnau 2008) is used on the data, and the results of the correlation are shown in Table 3.

Table 3 Correlations among searches (refer to Table 1 for source data)
Full size table

Solution generation: There are fifteen new solution searches (ns) (see Table 2) that were previously not known to the designers and were found while solving problems. From Table 3, it can be interpreted that new solution search positively influences global solution search (gs). This in turn influences the occurrence of local solution search (ls) and subsequent detailed solution search (dp), see Fig. 3 for an example. Presence of new solution search influences the number of searches in all other types of search. Also, the number of detailed solution searches is influenced by the total number of searches of all other types. Hence, designers who find many design spaces have an increased possibility of finding a large number of solutions.

Fig. 3
figure 3

New solution search (ns) versus global solution search(gs). ‘Y’ axis shows number of search

Full size image

Problem understanding: As the total number of new problem (np) searches are very few (only two, see Table 2), its relationships with other terms are ignored. From Table 3, it can be concluded that global problem search (gp) influences local problem search (lp), and this in turn influences detailed problem search (dp). Also, global problem search influences both local and detailed problem searches. Again, detailed problem search is influenced by the presence of all other searches at the higher levels of the hierarchy.

Solution evaluation: There was no new evaluation search (ne) in any of these experiments. Table 3 indicates that global evaluation search (ge) positively influences local evaluation search (lp), which in turn influences detailed evaluation search (de). Global evaluation search influences both local and detailed evaluation search. It can be concluded that detailed evaluation is influenced by the number of other types of evaluation search carried out in the process.

Thus, we argue that the various types of search spaces (viz. new, global, local, and detail) in all the three phases of design problem solving (viz. problem understanding, solution generation, and evaluation/selection) form a hierarchy and mutually influence each other as shown in Fig. 4.

Fig. 4
figure 4

Search hierarchy

Full size image

From Table 3, it is observed that on average:

  • For each new problem found, 17 global problems, 10 local problems, and 33 detailed problems were generated, and for each global problem, 0.5 local problems and 2 detailed problems were found.

  • A single new solution space found was associated with 8 global solutions, 3 local solutions and 42 detailed solutions. Moreover, each global solution search led to the generation of 0.3 local and 5 detailed solution searches, respectively.

  • There was no new evaluation search found. For each global evaluation, on average, 0.2 local and 0.8 detailed evaluation searches, respectively, are found to occur.

From the above observations, we see that presence of generic solutions (i.e., new and global searches) encourages many other similar solutions to be generated. This strengthens our argument that to support conceptual design, designers should explore new design spaces or be exposed to new generic design spaces. Ideas belonging to a different design space should help in the generation of other ideas that are different from that already generated by the designers during solving a problem.

4 Finding related problems and clarifying them to help in solution generation

Smithers et al. (1992) stated that analyzing design problem’s characteristics creates and bounds the space within which possible design solutions can be located. Similarly, researchers such as Nidamarthi (1999) have shown that better problem understanding helps better solution generation in terms of requirement satisfaction. Now, let us see whether the results of the experiments reflect this.

To find relationships, if any, among the different kinds of search that took place in the design experiments, the total number of searches within each phase (e.g., problem understanding, solution generation, and solution evaluation) was correlated with that of those in the other phases. The total number of problem search, for instance, is taken as the sum of the number of searches in all four types of problem search. Similar processes are followed for the other outcomes also.

In Table 4, it can be noticed that problem search influences evaluation search (Row 4). Contrary to our expectations, no correlation was found between the numbers of problems identified (total number of searches in all problem search types) by designers and the total number of solutions generated (total number of searches in all solution search types). Nidamarthi (1999) did not differentiate between clarification (general clarification and method clarification) and problem searches; it was found, however, that the amount of solution search is influenced by the presence of clarifications and problem searches (Row 5). Thus, search in solution space is influenced by the search of the problem space. Table 4 also shows that there is a fair correlation between clarification and solution generation, hinting that clarification of a given problem enhances the generation of a number of potential solutions.

Table 4 Relationship among total number of searches of each type (refer Table 2 for source)
Full size table

5 Time spent in solving a problem positively affects the design outcome

Does spending more time in conceptual design help generate a larger number of ideas? The effect of the length of the experiment (i.e., the duration of designing) on design outcomes was assessed by finding its correlation with the number of different kinds of searches in the design process (see Table 5).

Table 5 Effect of duration of the experiment on searches (see Table 2 for data)
Full size table

Table 5 shows that as the duration of the design process increases, the total number of solutions generated, total number of searches generated, total number of problems, and clarifications generated, as well as the total number of global, local, and detailed solutions generated increase. Thus, as designers spend more time on solving a problem by finding out different kinds of search, the number of potential solutions increases. This shows that time is an important deciding factor for creative design: as designers spend more time in thinking and generating ideas, the chances are higher that the outcome will be more creative.

6 Creative ability and experience of designers positively affect the design outcome

It has been already established that creativity of individuals affects the outcome of their activities (Shalley 1991; Woodman et al. 1993; Amabile 1996). Creative outcome also affects company performance (Amabile 1988). In this section, we investigate how this is reflected in the experiments conducted. Creativity of each of designer involved in the main experiments (i.e., the second set of experiments) was assessed using the outcomes of the design sessions in which they were involved.

Sternberg and Lubart (1999) defined creativity as that which ‘produce work that is both novel (i.e., original, unexpected) and appropriate (i.e., useful, adaptive concerning task constraints).’ Similarly, Weisberg (1993) defined it in terms of ‘novel and valuable products, capacity to produce such works, and the activity of generating such products.’ In a recent work, Sarkar and Chakrabarti (2011), proposed: ‘Creativity in design occurs through a process by which an agent uses its ability to generate ideas, solutions, or products that are novel and valuable (useful).’ According to this definition, the core components of creativity are ‘novelty’ and value (usefulness).’

Sarkar and Chakrabarti (2011) proposed that difference of products in terms of their characteristics can be employed to determine the relative degree of novelty of products. For assessing usefulness or value of ideas, they advocated that products that are good for the society, and are used by or benefit many people for a long period, are more useful than those that are not. Usefulness is expressed in terms of ‘importance of usage of a product,’ ‘its popularity of usage,’ and ‘the rate or duration of usage,’ and used as a criterion for assessing creativity (Sarkar and Chakrabarti 2011).

Chulvi et al. (2012) analyze the influence of several design methods (Brainstorming, Functional Analysis, and SCAMPER method) on the degree of creativity of the design outcome using metric of Moss(1966), the metric of Sarkar and Chakrabarti (2011), as stated before, and the evaluation of innovative potential. They found that Brainstorming provides more creative outcomes than when no method is applied. Another study by Kudrowitz and Wallace (2013) explores a metric for evaluating large quantities of early-stage product sketches and tests the metric through an online service called Mechanical Turk. They found that ‘clarity of the sketch positively influenced ratings of idea creativity. Additionally, the quantity of ideas generated by an individual participant had a strong correlation with that participant’s overall creativity.’ They believe that a metric of three attributes to be used as a first pass in narrowing a large pool of product ideas to the most innovative: novel, useful (or valuable), and feasible (as determined by experts). While in another related study, Oman et al. (2013) presented several methods used to assess the creativity of similar student designs using metrics and judges to determine which product is considered the most creative. They developed a critical survey that provided, along with a comparison of prominent creativity assessment methods for personalities, products and the design process.

In each of these sets of design experiments, the participant designers are selected after one final solution after evaluating all the solutions generated during that session. Two groups, each consisting of two experienced designers, assessed the creative outcomes from each design experiment, using the creativity measurement method proposed by Sarkar and Chakrabarti (2011) as discussed in the last paragraph. The novelty and usefulness of the final designs were assessed, and using these data, creativity was assessed (see Table 6). Table 6 shows that designer ‘Dh’ has the highest creativity, while designer ‘Su’ has the lowest.

Table 6 Measure of creativity
Full size table

Table 6 shows that the average creativity of these experienced designers were substantially high compared to that of the novice designers. The average creativity value of the experienced designers is 5.87; the same for the novice designers is 3.12. This hints that experience often has a positive effect on the creative outcome of an individual (other researchers such as, Maher et al. (1995), Amabile (1997), and Kletke et al. (2001) also conveyed similar statements), however, to find out deep relationships, if any, further analysis is conducted as discussed below.

We investigate how creativity of designers affects their design outcomes. For this, we assess the correlation between ‘the number of different types of searches found by the designers’ and ‘the novelty, usefulness, and creativity of the final selected solution,’ in each design session. Among many possible correlations among search types and novelty and usefulness, only major correlations are shown in Table 7, which shows that search occurring at the higher levels of the search hierarchy (e.g., new and global search) influences novelty of the outcome. Searches at the lower levels of hierarchy influence usefulness of the outcome. The creativity of the outcome is influenced by the total number of searches occurring in each phase of conceptual design, both individually and combined together.

Table 7 Correlation between novelty, usefulness, and creativity ranks with searches
Full size table

7 Effect of creative problem-solving style on design outcomes

Analyses done in the previous sections are based on the outcome of the design experiment. Even the assessment of creativity of individuals as proposed by Sarkar and Chakrabarti and used in this work is also based on assessing individuals based on their creative outcome. In this subsection, we investigate whether designers’ intrinsic abilities have any effect on the design outcome. For this, we use Kirton’s Adaptive Innovative Inventory (KAI) on all participating designers (Kirton 1977) to assess one of the major intrinsic ability, the creative problem-solving style of the designers (see Table 8).

Table 8 Searches in the main design experiments
Full size table

It was found that almost all designers involved are innovative in their problem-solving style (Table 8); thus, we are unable to map different creative problem-solving styles (viz. innovators and adaptors, as proposed by Kirton 1977) with design outcomes. One possible explanation for this observation is that product designers are generally innovative. Similarly, Kirton mapped people with different occupations in the USA, UK, and Europe, and found that engineers have an average KAI score between 95 and 97, R&D managers have KAI between 101 and 103, and people involved in fashion have KAI between 104 and 110 (Kirton 1977, 2012).

8 Influence diagram

Influence diagrams are visual representations showing how different factors influence one another, in a decision-making system. The results of the experiments, as described above, indicate that creativity is influenced by several factors such as total number of searches (including the number of solutions, Sect. 3.3), amount of time spent in solution generation (Sect. 3.8), and experience of the designers (Table 6). Based on these findings, we can create an influence diagram for creativity, as Fig. 5, showing how different factors affect the design outcomes and creativity, as observed in this work.

Fig. 5
figure 5

Influence diagram

Full size image

9 Discussion and conclusions

Shalley and Gilson (2004) state that many factors affecting creativity have been identified in literature; however, it is still unclear which of these are major influences. While some factors such as fluency and flexibility (Torrance 1979) have been proposed after empirical studies, others (such as being able to predict outcomes or bureaucratic procedures) are based on logical arguments only (Sarkar and Chakrabarti 2008). Given the amount of time and energy required to influence each factor, it is not feasible for a company or individual to work on all the factors. The work reported in this paper highlights some of the important factors (that is, amount and types of search, duration, and experience) that affect the creative outcome of designers.

Within the limitations of the think aloud protocol analysis method (Cross et al. 1996), the number of experiments conducted (16 experiments, 8 initial, and 8 main), the number of designers participated (14 designers, 6 in initial, and 8 in main), the amount of time per experiment (an average 60 and 77 min of length for each initial and main design experiment, respectively), and the limited number of problems used (2 for initial and 2 for main), we were able to investigate influences of only some of the factors on designing and its creativity aspects.

The findings in this research work are summarized as below:

  • It has been found that search of design spaces takes place in all phases of conceptual design, where designers search these spaces to identify related problems, generate solutions, and identify associated evaluation criteria.

  • During designing, designers enter a new design space through the generation of an idea.

  • It was also found that there are four types of search, ‘new,’ ‘global,’ ‘local,’ and ‘detail’ that take place in each phase of conceptual design. These searches influence each other in all phases of design. These can be interpreted as follows: designers who find many design spaces have an increased possibility of developing a large number of solutions.

  • The searches influence both the quality and quantity of the designs created.

  • Occurrence of higher-level searches (new and various ideas) in the hierarchy seems to enhance the occurrence of lower level searches (more number of detailed ideas).

  • Finding generic problems, solutions, or evaluation criteria (e.g., ‘new’ and ‘global’ search) should help in finding specific problems, solutions, or evaluation criteria (e.g., ‘local’ and ‘detail’ searches).

  • It is also observed that searches occurring at the higher levels of the search hierarchy have a major influence on the novelty of the outcomes. This is further supported by the findings of Srinivasan and Chakrabarti (2010) work, which indicates that novelty of concept space is dependent on its variety.

  • Searches at the lower levels of hierarchy (more number of detailed ideas) primarily influence the usefulness of the outcomes.

  • The creativity of an outcome is influenced by the total number of searches (many different detailed ideas) occurring in each phase of conceptual design, both individually and taken together. This understanding could be used to make search more efficient at the initial phases of design, if the focus of the design is known—whether on novelty, usefulness, or both (creativity), thus controlling the outcome.

In some way, the results support that presence of certain personal abilities aid creative outcomes, as identified by Torrance (1979) who later developed Torrance Tests of Creative Thinking (TTCT). These abilities are as follows:

  1. (1)

    Fluency: the ability to produce a large number of ideas (total searches positively influences creativity, see Table 7).

  2. (2)

    Flexibility: the ability to produce a large variety of ideas (presence of ns and gs—see Table 7).

  3. (3)

    Elaboration: the ability to develop, embellish, or fill out an idea (presence of ds—see Tables 3 and 7).

  4. (4)

    Originality: the ability to produce ideas that are unusual, statistically infrequent, not banal, or obvious (presence of us—see Sect. 3.2–3.6).

For instance, following Torrance’s work, if fluency is expressed in terms of the total number of ideas or the total number of searches, then fluency influences the creative outcome of an individual. Additionally, if we map novelty and value (or usefulness) also, we find that novelty is influenced by originality (new solution search), fluency (all solution searches), and flexibility (new, global, and local solution search), whereas value (usefulness) is influenced by the designers’ fluency (local solution search) and elaboration (detailed solution search).

Yamamoto et al. (2009) attempted to capture the nature of concept generation process by finding an effective thinking pattern for creativity. They considered design spaces that are made of a chain process of concepts that are both explicitly evoked in concept generation process and inexplicitly imagined as a thinking space. First, they refer to the structure of the space, and second, they refer to the latent concepts in that space. They found significant correlation between the structure and creativity. Additionally, Harakawa et al. (2005) found strong relationships between extension of thinking during designing and the level of creativity in the ideas generated; they expressed that extension of thinking space is also a structural feature, and it strongly affects creativity. To some extent, new and global spaces in our work refer to ‘structure’ in Yamamoto et al. (2009) work, while local and detail searches refer to extension of space in both the above works. As stated by these above two researchers, creativity is enhanced both by the finding of space and its extension; however, in our work, we find a more detailed reason for it. Novelty is influenced predominantly by the new and global space findings, and usefulness is by the local and detailed space findings (from Table 7, new, global, and local problem searches and novelty: 0.85 (p < 0.01, very high correlation)). Since creativity is a product of both novelty and usefulness, creativity is affected in both the cases (from Table 7): 0.76 (p < 0.02, high correlation).

This study indicates that the creative quality of design outcomes is positively influenced by the total number of searches, amount of time spent in designing, and experience of designers. Thus, to enhance the creative quality of design outcomes, the following could be considered:

  • Encourage designers to generate a large number of problems, solutions, and evaluation criteria related to the given set of requirements. Gelsey et al. (1998) suggested that automated search of a space of candidate designs seems an attractive way to improve the traditional engineering design process.

  • Encourage designers to spend adequate amount of time in solving the design problem.

  • Consider having more numbers of experienced designers. Importance of experience in design has also been echoed by several researchers, such as, Kletke et al. (2001), Amabile (1997), and Maher et al. (1995).