Proceedings of the ASME 2007 International Design Engineering Technical Conferences & Computers and

Information in Engineering Conference

IDETC/CIE 2007

September 4-7, 2007, Las Vegas, Nevada, USA

DETC2007-34948

INCREASING INNOVATION: A TRILOGY OF EXPERIMENTS TOWARDS A DESIGN-BYANALOGY

METHOD

J. S. Linsey

Manufacturing and Design Research Laboratory

Department of Mechanical Engineering

The University of Texas

Austin, Texas

E. Clauss, K. L. Wood

Manufacturing and Design Research Laboratory

Department of Mechanical Engineering

The University of Texas

Austin, Texas

ABSTRACT

Design by analogy is a noted approach for conceptual  design. This paper seeks to develop a robust design-byanalogy  method. This endeavor is sought through a series of  three experiments focusing on understanding the influence of  representation on the design-by-analogy process. The first two  experiments evaluate the effects of analogous product  description--presented in either domain-general or domainspecific  language--on a designer’s ability to later use the  product to solve a novel design problem. Six different design  problems with corresponding analogous products are  evaluated. The third experiment in the series uses a factorial  design to explore the effects of the representation (domain  specific or general sentinel descriptions) for both the design  problem and the analogous product on the designer’s ability to  develop solutions to novel design problems. Results show  that a more general representation of the analogous products  facilitates later use for a novel design problem. The highest  rates of success occur when design problems are presented in  domain specific representations and the analogous product is  in a domain general representation. Other insights for the  development of design by analogy methods and tools are also  discussed. 

J. P. Laux

The Similarity and Cognition Lab

Department of Psychology

The University of Texas

Austin, Texas

A. B. Markman1

The Similarity and Cognition Lab

Department of Psychology

The University of Texas

Austin, Texas

1. INTRODUCTION

Design-by-analogy is a powerful tool for developing ideas.  Designers frequently base their designs on products they have  seen before. Professional designers often use analogies  [1,2,3]. Numerous examples of innovative products based on  analogies may be found in technology magazines and related  literature. A recent example is a retractable mast with sail  designed after studying bird and bat wings [4]. This sail is also  useful for cargo ships to harness wind power, reducing fuel  costs [5]. 

Design-by-analogy is clearly a powerful tool in the design  process but numerous questions surround its use. What will  make the designers more successful? What do designers not  do well? What are typical wrong turns or places designers  have difficulties? What makes a good analogy? What tools do  designers need to support this process?   To more fully understand the use of analogy in design and  to serve as a basis for the development of a robust design-byanalogy  method, we have performed a series of controlled  experiments. This paper begins with a description of the  motivation for this work and then describes research and  theory on innovation and the underlying cognitive processes  that support it. Next we present a series of experiments exploring the way the representations of prior knowledge and  new design tasks affects the use of analogy. Finally we  discuss the implications of these results for a robust theory of  design-by-analogy and for the development of tools to support  innovation processes.   Figure 1: The sails of this cargo ship are designed  based on an analogy to a bat’s wing [4,5].  2. MOTIVATION AND PREVIOUS WORK

Prior work in the design research field has focused on the  development of formal design-by-analogy methods and  understanding relevant cognitive processes.   2.1. Formal Design-by-Analogy Methods  A few formal methods have been developed to support designby- analogy such as Synectics [6], French’s work on  inspiration from nature [7], Biomimetic concept generation [8]  and analogous design through the usage of the Function and  Flow Basis [9]. Synectics is a group idea generation method  that uses four types of analogies to solve problems: personal  (be the problem), direct (functional or natural), symbolic and  fantasy [6]. Synectics gives little guidance to designers about  how to find successful analogies. Other methods also base  analogies on the natural world. French [7,10], highlights the  powerful examples nature provides for design. Biomimetic  concept generation provides a systematic tool to index  biological phenomena [8,11]. From the functional  requirements of the problem, keywords are derived. The  keywords are then referenced to an introductory college  textbook and relevant entries can be found.   Analogous concepts can also be identified by creating  abstracted functional models of concepts and comparing the  similarities between their functionality. Analogous and nonobvious  products can be explored using the functional and  flow basis [9]. This approach requires a database of products  represented in the function and flow basis.   2.2. Representation  A representation is a physical or mental item that stands for  another thing. Hence, there are four necessary parts to a  mental representation: the physical or mental item serving as  the representation, the domain being represented, rules  (usually implicit) that map parts of the first to parts of the   second, and a process (also usually unstated) that is capable of  performing the mapping and using the information [12].  Understanding the design process requires understanding both  the internal mental representations of designers as well as the  external representations (e.g., sketches, function and flow  basis diagrams) that are used during the design process.   2.3. Prior Analogy in Design Experiments  Human-based design methods require a deep understanding of  the processes people use and the areas where guidance or  assistance could improve the process. This knowledge is  gained largely through experimental work. Even though  design-by-analogy is a well-recognized method for design,  few human experiments exist. Notable results from these  experiments include the work of Casakin and Goldschmidt,  Ball et al., Kolodner, and Kryssanov et al. Casakin and  Goldschmidt. Casakin and Goldschmidt found that visual  analogies can improve design problem solving by both novice  and expert architects [3]. Visual analogy had a greater impact  for novices as compared to experts. Ball, Ormerod, and  Morley investigated the spontaneous use of analogy with  engineers [13]. They found experts use significantly more  analogies than novices do. The type of analogies used by  experts was significantly different from the type used by  novices. Novices tended to use more case-driven analogies  (analogies where a specific concrete example was used to  develop a new solution) rather than schema-driven analogies  (more general design solution derived from a number of  examples). This difference can be explained because novices  have more difficulty retrieving relevant information and have  more difficulty mapping concepts from disparate domains due  to a lack of experience [14].   A structured design-by-analogy methodology would be  useful for minimizing the effects of the experiential gap  between novices and experts and to further enhance experts’  abilities. The cognitive analogical process is based on the  representation and processing of information, and therefore  can be implemented systematically given appropriate  conceptual representations and information processing tools  [15,16].   Prior research in analogical reasoning found the encoded  representation of a source analogy (the analogous product)  can ease retrieval if it is entered into memory in such a way  that the key relationships apply in both the source and target  problem domains [17,18]. This work shows that the internal  representations in memory play a key role in retrieval. The  analogies and problems used in these experiments were not  specific to any domain of expertise and used fantasy problems  relying on strictly linguistic descriptions. Little work has been  carried out based on a strong psychological understanding of  analogical reasoning combined with the design knowledge of  analogies for high-quality designs. This paper takes a  distinctive interdisciplinary route to combine these threads of  research to develop a more complete understanding of the use  of analogy in engineering design and to provide the basis for   2      

 formal method development. Designers rely on both internal  mental representations and numerous external representations  ranging from sketches to specialized diagrams such as black  box models. The use of various representations in the design  process warrants further understanding. The following  experiments further investigate visual and semantic  representation effects on design-by-analogy and lead to a  deeper understanding of how to enhance the design-byanalogy  process.   2.4. Cognitive Processes: Design-by-Analogy  Understanding the cognitive process involved in the formation  of analogies is important for understanding the concept  generation process. Analogy can be viewed as a mapping of  knowledge from one situation to another enabled by a  supporting system of relations or representations between  situations [19,20,21]. This process of comparison fosters new  inferences and promotes construing problems in new  insightful ways. The potential for creative problem solving is  clearest when the two domains being compared are very  different on the surface [22].   Research has been carried out in the field of psychology  to understand the cognitive processes people use to create and  understand analogies [21,22,23,24,25]. Figure 2 shows the  basic process steps involved in reasoning by analogy, the most  cognitively challenging step, and the design methods that are  available to support each step.   Understanding the cognitive process involved in the  formation of analogies is important for understanding the  concept generation process. These processes are best  understood by referring to a predicate-argument representation  like that used in logic, artificial intelligence, or linguistics.  Within this framework, a predicate is a statement that is  asserted of a subject or subjects, and arguments are the  subjects of which predicates are asserted. These predicates  are partitioned into attributes (defined as taking single  arguments) and relations (which take two or more arguments).  For example, in the statement ‘the boot is brown’, brown is an  attribute, and can be written using the one argument predicate  Brown(boot). In the form ‘Brown(boot)’, ‘Brown’ is a  predicate and ‘boot’ is its argument. In the statement ‘the boot  is larger than the shoe’, larger-than is a relation, which would  be written with the two argument predicate Larger_than(boot,  shoe). Relations can connect other relations as well as  objects. The most common example of this is the relation  cause, as in ‘Bob is taller than Sam, causing Bob to jump  higher than Sam’, written Cause (Taller_than (Bob, Sam),  Can_jump_higher_than (Bob, Sam)). Such a relation is  known as a higher order relation (for a fuller account, see  [21]).   Methods Supporting Design-by-Analogy  1) TRIZ Relationship Matrix  2) Function and Flow Basis  1) TRIZ Relationship Matrix  2) Function and Flow Basis  3) Biomimetic Concept Generation  4) Synectics  Encode the source  Retrieve the appropriate  analogy (source)  Mapping between the  target problem and  the source is found  Inference based on  the mapping are  found (solution)  Relatively Straight  Forward Steps  Steps in Human Reasoning by Analogy  Cognitively  Difficult Step  Figure 2: Steps in human reasoning by analogy and  the current methods available to support those  processes.   Analogy has traditionally been viewed as a comparison  between two products in which their relational, or causal  structure, but not the superficial attributes match [20,22]. For  example, an airplane wing and a hydrofoil can be viewed as  analogous because of how they work; the colors they are  painted is irrelevant. This process of comparison fosters new  inferences and promotes construing problems in new,  insightful ways. The potential for creative problem solving is  most noticeable when the situation domains are very different.   In the psychological literature, there has been a great deal  of interest in the roles of analogy and expertise in problem  solving. When working with undergraduate students who  have no specialized domain knowledge, a classical finding is  that analogies are helpful in solving insight problems, but are  difficult to retrieve from memory [26]. Conversely, naturalistic  research with experts typically finds that analogies are often  used [e.g., 27, 2, 3]. This dichotomy may reflect that experts  can see the deeper, logical structure of situations while those  without domain expertise are mainly aware of only the  superficial features [cf. 28, 29, 30].   To clarify and more fundamentally understand these  issues, laboratory research, which affords good experimental  control, needs to be conducted with burgeoning domain  experts. Such individuals are capable of recognizing the  causal structure of products, but could also be distracted by  superficial features. These characteristics make them the  appropriate test bed for determining the role of source  representation in analogical reminding. Moreover, it has been  suggested that implicit processes could mediate analogical  problem solving [31]. That is, problem solving can occur  without being aware of the analogous solution in memory.  Therefore, it is important to assess when participants find the  solution and recognizing the analogy, separately.   2.5. Semantic Memory Retrieval  Designers frequently base their concepts on ideas they have  seen and experienced previously. These designs are retrieved  directly from their long-term memory, specifically semantic  memory. Semantic memory refers to the storage of  meaningful, factual information. Semantic memory is  contrasted with the storage of personal experiences (episodic  memory) or skills (procedural memory). In the psychological  literature, the structure of human semantic memory is often  conceptualized as a network of concepts that are associated  with each other. For example, in Figure 3, the concept of a bed  is represented by a node in a somewhat chaotic web of  associations. When one thinks about beds, the node  representing that concept becomes active, and this activation  can spread out along its associative links to other connected  ideas. Another concept is remembered when it becomes  sufficiently activatived. However, nodes pass along only a  fraction of their activation to neighboring nodes, and so the  activation weakens as it gets more distant from the source of  activation. The probability that a concept will be remembered  increases as the path distance (i.e. number of links traversed)  shortens, or if multiple active paths converge on it. Nodes that  are more general concepts, such as “substance”, tend to be  connected to a much greater number of other nodes, becoming  hubs in the network. Thus, linking new concepts through them  shortens path distances, increasing the probability of retrieval  [12,32,33,34].   substance  inflate  fill  air  water  mattress  clutter  spring bed  soft  stuffed  Figure 3: Example Semantic Network   3. EXPERIMENTAL APPROACH AND  RESEARCH  QUESTIONS  Designers need a predictable method for developing  innovative solutions to difficult design problems. Thus, it is  crucial that we understand the relationship between the  representation of the design problem and the representation of  analogous product descriptions in memory Our goal is to  explore the factors that make previously seen analogous  products easier to retrieve and use in solving the problem. The  problems used in these experiments have many viable  solutions. The goal of the experiment is not to determine if the  participant can find solutions to the design problem but to  explore the factors that affect the use of analogous solutions.  The solutions of interest for this experiment are the ones based  on products presented in the first part of the experiment.  These analogous products represent a useful source for  finding solution to the design problems. The experiments use  a combination of visual and semantic information to represent  the source design analogy.   In this context, we seek to answer the following research  questions:   •  Question 1: Is prior product knowledge more likely to be  retrieved and used in innovative design when it is  described using domain-general or domain-specific  language? Prior psychological literature [17,18] implies  that the domain-general descriptions should be more  likely to be retrieved but this needs to be validated and  explored in a more realistic situation.  •  Question 2: How does the representation of the problem  statement affect the ability of a designer to retrieve and  use a relevant analogous product to expose a solution to a  new design problem?  •  Question 3: Usually, when a designer is solving a novel  problem, the representation of appropriate analogous  products is not known to the designer. What is the best  way to represent a design problem in this situation and  what implications does this have for a design-by-analogy  method?  4. EXPERIMENTAL APPROACH  To further explore the effects of representation on analogy use  for real-world problems and to further understand how  supporting methodologies should be created, a series of three  experiments was implemented. This series of experiments  controlled how participants learned about a series of products  and therefore also controlled how the products were  represented in their memories. This allowed the predictions  from psychological models of analogical reasoning and  semantic memory to be evaluated. These models, along with  additional knowledge gained from experimentation, will be  used as the basis for methods development. These experiments  were conducted over three semesters with senior mechanical  engineers with instruction in design methodology including  idea generation and from two professors’ classes. Each  experiment contained a unique set of participants.   The first and second experiment explored the effects of the  analogous product representation on a designer’s ability to  later use the product to solve a novel problem. A total of six  design problems with corresponding analogous products were  explored to more fully understand the influence of semantic  representation and other factors in analogical design. The  analogous solutions were semantically described using either  domain specific or more general terms that applied across both  the problem and the analogous product domains.   The third experiment evaluated the representation effects  for both the analogous product and design problem. A 2 X 2  factorial experiment design was employed which resulted in  four different experimental groups, Table 1. For both the  analogous product and the problem description, two levels of  participants were compared, a “Domain Specific Description”  Group and a “General Description” Group, Table 2. All  experiments used a combination of visual and semantic  information to represent the source design analogy.

 5. OVERVIEW  OF THE ANALOGOUS PRODUCT  REPRESENTATION EXPERIMENTS 1 & 2  The experiments consisted of two tasks: Memorize the  Analogous Products and Solve the Design Problems with a  week in between for most participants. Normally when faced  with a design problem, a useful analogous product has not  been seen immediately beforehand, but the analogous product  is stored in a person’s long term memory. A week was chosen  as a relevant time because any analogies retrieved will be  taken from long-term memory and this time frame has been  used previously [35]. Multiple solutions were encouraged for  all parts of the experiment. Participants were told the  experiment evaluated various skills in the design process.   The two analogous product representation experiments  were virtually identical with the exception of the design  problems and analogous products being evaluated. Results  from Analogy Experiment 1 left many unanswered questions  and showed some short-comings in the analogies that were  chosen [36]. To further understand design-by-analogy, a  second set of analogies were evaluated in Experiment 2. For  Experiment 1, the football to the raft analogy required the  mapping of visual rather than semantic information and the  sketches of the flour sifter device were difficult to interpret.  For Experiment 2, two innovative products found in the  literature were used. The first, a kayak with a hydrofoil could  have been based on an analogy to an airplane wing [37]. The  second, a set of dirt bike racer goggles was based an analogy  to film in a camera [38].   5.1. Procedure for Experiments 1 & 2  For the first task, Memorize the Analogous Products,  participants were given five short functional descriptions of  products along with a picture (example in Table 2 and see  appendix for complete descriptions) and asked to spend thirty  minutes memorizing the descriptions. The products were  functionally described in a few short sentences either with a  more general description that applied in both the source  analogy and target design problem domains, or with a domainspecific  description. An example of the descriptions used for  the film in a camera is shown in Table 2. The product  descriptions and the design problems included meaningful  pictures. The semantic descriptions of the devices were varied  but the pictures were identical for both conditions. The focus  of these experiments was on the linguistic representations of  the devices, but visual information was also present.   Both groups were then given up to fifteen minutes to  answer a quiz, requiring them to write out the memorized  descriptions. Finally the groups spent up to ten minutes to  evaluate their results. Three of the products acted as source  analogies for the design problems in the second task, Solve the   Design Problems, and three were distracter products that  shared surface similarities with the design problems (Figure 4  and Figure 5).   Table 1: Overview of the Factorial Design for  Experiment 3    Factor 1: Analogous Product Representation  Factor 2: General Domain Specific  Design  Problem  General Group 1:  General, General  Group 2:  General, Domain  Representation Domain  Specific  Group 3:Domain,  General  Group 4:  Domain, Domain   Time limitations were based on a pilot experiment with  graduate students with no time limits. Time limits were set to  be longer than the amount of time required by most  participants in the pilot experiment. For certain tasks and  phases, it was clear participants were not spending enough  time on the task, so the time limits were actually extended  well beyond the time required in the pilot experiment.   In the second task, Solve the Design Problems,  participants were given three design problems to solve in a  series of the following five phases:   Phase 1: Open-ended design problems, few constraints   Phase 2: Highly constrained design problems   Phase 3: Identify analogies and try using analogies   Phase 4: Informed task 1 products are analogous   Phase 5: Target analogous product from task 1 is given   and participants need to find the solution   Phases one and two were completed for the three  problems followed by phases three through five. Throughout  all phases participants were given the general idea generation  guidelines to (1) generate as many solutions as possible with a  high quality and large variety, and (2) to write down  everything even if it did not meet the constraints of the  problem including technically infeasible and radical ideas.  Participants were also instructed to use words and / or  sketches to describe their ideas. They were asked not to  discuss the experiments with their classmates until all the  experiments were completed.  In phase 1, the problems were initially presented with few  constraints. Participants received eleven minutes to generate  ideas for the open-ended design problems and then eleven  additional minutes to create more solutions to the same  problem with additional constraints. The additional constraints  limited the design space increasing the chance the participants  would retrieve the desired source analogy. Next they had a  five minute break.   In phase 3, participants spent fifteen minutes listing any  analogies they had used and also used analogies to develop   5       

 Table 2: An example of the domain specific and general device descriptions given to participants for task 1.   Sentence / General (G) or Domain (D) Specific  1 G Two reels move a surface in the path of incoming substance.  D Two reels feed the film in front of the stream of light.  2 G The surface collects the substance and then a new unchanged surface is moved into place.  D The film captures the image and then a new unexposed section of film is moved into place.   Air Mattress Water-filled Travel Weights  Source Product  Analogies  Target Problem  Solutions  Problem 1  Problem 2  Problem 3  Distracter Products  Whisk  Bullet Raft  Football  Flour Duster  Pan Cake  Flipper  Travel  Cart  Figure 4: Source products analogies,  corresponding target problem solution and  the distracter products for experiment 1.   additional solutions. In phase 4, the participants were told that  products from the first task were analogous, to mark their  solutions that used the analogy and to generate additional  solutions using the products from the first task (Memorize the  Products). Finally, participants were given the target analogy  for each problem, asked to place a check where they had used it  and asked to generate more ideas if they had not used the  described analogy. This final phase serves as a control to verify  that the analogies being used are sensible, are useful for these  particular design problems and to facilitate data evaluation. At  each phase, participants used a different color of pen, thus  identifying the phase. A short survey at the conclusion of the  experiment evaluated English language experience, work  experience, if the participant had heard about the experiment  ahead of time, functional modeling experience, if they felt they  had enough time and prior exposure to the design problem  solutions. For Experiment 2, the survey also included a  question asking the participants to write down a list of features  they used from the targeted analogous product from task 1 to  find their solution to the design problem. For the sketches that   were difficult to interpret, the additional survey questions  assisted in evaluating if the appropriate features from the  analogous product had been used to solve the design problem.  Results from the first task were matched to the second task.  The entire experiment required about three hours.   Airplane Kayak  Source Product  Analogies  Target Problem  Solutions  Problem 1  Problem 2  Problem 3  Distracter Products  Toy  Dirt Bike Racer Goggles  Film in a Camera  Flour Duster  Pepper  Grinder  Football  Figure 5: Source products analogies,  corresponding target problem solution and  the distracter products for experiment 2.   5.2. Metrics  Each analogy produces a set of solutions, not a single solution.  Participants also created a large number of solutions which  were not based on the analogies provided. We were primarily  interested in the phase of the study at which participants  produce a solution to the constrained design problem based on  the targeted analogy and also the phase at which they identified  the analogy that they used. As we will see, people often show  evidence of being influenced by an analogous product without  explicitly recognizing where the idea came from. Two  evaluators coded the data independently, recording when the  analogous solution was found. Initial agreement was  approximately 80% across the three experiments and disagreements were readily resolved through discussion. The  most common reason for the initial differences was the  participant referenced solutions that appeared on different  pages of the results.   5.3. Results and Discussion: Analogous Product  Representation Experiments 1 & 2  Example solutions are shown in Figure 6. Figures 8-13 show  the cumulative percentage of participants who found a valid  solution to the constrained design problems based on the  appropriate analogous product. The analyses excluded  participants who remembered seeing the expected solution  prior to the experiment. The expected solutions are actual  products so it is possible for the participants to have seen the  products prior to the experiment. In addition, a verbal  description without a picture of the water weight example is  given in the textbook used in the participants’ design methods  class but the section is in the optional readings for the class. It  is unlikely that any of the students read this section of the book  prior to the experiment.   The representation of the analogies in the participants’  long-term memory affected the probability the analogous  product would be used to solve an appropriate design problem  for certain problems (Figure 7 and Figure 12). Appropriate  representations can improve the success rate in design-byanalogy.  For two of the design problems, participants who  received general descriptions of the analogous products had  statistically higher probabilities of success. The results for  phase 4, are statistically significant for design problem 1  Experiment 1 and design problem 3 Experiment 2 (Figure 7  and Figure 12). Using a binomial probability distribution [39],  the probability that the domain specific description group is  from the same distribution as the general description group is  almost zero.   We should note that we reported the data from Experiment  1 in a prior paper [36], though it was analyzed differently there.  In the previous paper, we did not distinguish between when  participants used the analogous solution and when they  explicitly mentioned the analogy they had used. Looking more  carefully at our data, however, participants frequently find the  analogous solution without realizing the source of the idea.  This issue is discussed in detail in Section 5.5.   The semantic representation has an impact on analogy  retrieval but other factors also influence the process. The key  features to be mapped in Experiment 1, problems 2 and 3 were  visual information, the shape of the football and the varying  spacing of the wires in the whisk. The semantic representation  may only influence the analogical reasoning process when the  information that must be accessed and mapped is stored  verbally instead of visually. This proposal is consistent with our  observation that nearly all prior studies of analogical reasoning  involve semantic materials (typically written stories) [17,18].  Visual and verbal information are two distinct types of  knowledge stored in long-term memory [40].   Tightlycoil one wire into the shape ofafootball. Bend it to fillithfl let it otThese sections have larger spaces  between the wires and sift the flour as  the two halves move back and forth.  Flour  Pin  Tightly coil one wire into the shape of a  football. Bend it to fill with flour or let it out.  Wire is tight together.  Twist it to insert /  release flour  The pin is attached to the wire spiral. When the pin  is pulled, the flour falls through the cracks.  Figure 6: Examples of flour sifter solutions found by  the participant based on the analogy to the toy,  Experiment 2.  Percentage of Participants with a Solution based on the Correct  Analogous Product: Travel Water Weights  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1: Open- Ended Design  Problem  Phase 2:  Constrained  Design Problem  Phase 3: List/  Use Analogies  Phase 4: Use  Task 1 Products  Phase 5: Given  Analogy, Find  Solution  %  of  participants  with  a solution  using  the correct  analogous product  General (n=16)  Domain (n=15)  Air Water-filled Travel Weights  Figure 7: Design Problem 1, Experiment 1. A general  description facilitated retrieval and use of the  analogous product to solve a novel design problem.   For Experiment 1 design problems 2 and 3, it was also  much more difficult to evaluate the appropriateness of the  solution and to isolate the features that had been mapped. This  would lead to more inaccurately mapped solutions being  counted and erroneous results. This issue was corrected in  Experiment 2. For more details of the Analogous Product  Representation Experiments 1 see Linsey et al. [36].   7      

 Percentage of Participants with a Solution based on the Correct  Analogous Product: Bullet Raft  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1: Open- Ended Design  Problem  Phase 2:  Constrained  Design Problem  Phase 3: List/  Use Analogies  Phase 4: Use  Task 1 Products  Phase 5: Given  Analogy, Find  Solution  %  of  participants  with  a solution  using  the correct  analogous product  General (n=20)  Domain (n=16)  Bullet Football  Figure 8: Design Problem 2, Experiment 1. A more  general semantic representation does not ease  retrieval and use when visual information must be  mapped from the analogous solution.  Percentage of Participants with a Solution based on the Correct  Analogous Product: Flour Sifter  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1: Open- Ended Design  Problem  Phase 2:  Constrained  Design Problem  Phase 3: List/  Use Analogies  Phase 4: Use  Task 1 Products  Phase 5: Given  Analogy, Find  Solution  % of participants  with a solution using  the  correct analogous product  General (n=21)  Domain (n=16)  Whisk Flour Duster  Percentage of Participants with a Solution based on the Correct  Analogous Product: Kayak Design Problem  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1:  Open-Ended  Design Problem  Phase 2:  Constrained  Design Problem  Phase 3:  List/ Use  Analogies  Phase 4:  Use Task 1  Products  Phase 5:  Given Analogy,  Find Solution  % of participants  with a solution usingthe  correct analogous productGeneral (n=15)  Domain (n=11)  Figure 10: Design Problem 1, Experiment 2. This  problem required participants to use their knowledge  of fluid dynamics to appropriately choose the right  characteristics from the analogy.  Percentage of Participants with a Solution based on the Correct  Analogous Product: Dirt Bike Goggles Design Problem  80%  90%  100%  % of participants  with a solution usingthe  correct analogous productGeneral (n=12)  Domain (n=12)  Dirt Bike Racer  Goggles  Percentage of Participants with a Solution based on the Correct  Analogous Product: Bullet Raft  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1: Open- Ended Design  Problem  Phase 2:  Constrained  Design Problem  Phase 3: List/  Use Analogies  Phase 4: Use  Task 1 Products  Phase 5: Given  Analogy, Find  Solution  %  of  participants  with  a solution  using  the correct  analogous product  General (n=20)  Domain (n=16)  Bullet Football  Figure 8: Design Problem 2, Experiment 1. A more  general semantic representation does not ease  retrieval and use when visual information must be  mapped from the analogous solution.  Percentage of Participants with a Solution based on the Correct  Analogous Product: Flour Sifter  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1: Open- Ended Design  Problem  Phase 2:  Constrained  Design Problem  Phase 3: List/  Use Analogies  Phase 4: Use  Task 1 Products  Phase 5: Given  Analogy, Find  Solution  % of participants  with a solution using  the  correct analogous product  General (n=21)  Domain (n=16)  Whisk Flour Duster  Percentage of Participants with a Solution based on the Correct  Analogous Product: Kayak Design Problem  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1:  Open-Ended  Design Problem  Phase 2:  Constrained  Design Problem  Phase 3:  List/ Use  Analogies  Phase 4:  Use Task 1  Products  Phase 5:  Given Analogy,  Find Solution  % of participants  with a solution usingthe  correct analogous productGeneral (n=15)  Domain (n=11)  Figure 10: Design Problem 1, Experiment 2. This  problem required participants to use their knowledge  of fluid dynamics to appropriately choose the right  characteristics from the analogy.  Percentage of Participants with a Solution based on the Correct  Analogous Product: Dirt Bike Goggles Design Problem  80%  90%  100%  % of participants  with a solution usingthe  correct analogous productGeneral (n=12)  Domain (n=12)  Dirt Bike Racer  Goggles  that the word “film” is readily retrieved for the design problem.  This likely caused there to be no influence for the analogous  product description (Figure 11). It was as easy for participants  in the domain condition to remember the appropriate analogy  as it was for participants in the general condition.   Experiments 1 and 2 highlight the effects of sentential  representations on the design-by-analogy process. Other  factors that influence the design-by-analogy process are also  alluded to by these experiments.   Figure 9: Design Problem 3, Experiment 1. The   70%   semantic representation did not influence analogy   60%   use for this problem.   50%   Again in Experiment 2 more additional factors which   40%   influence the design-by-analogy process were observed. In   30%   Figure 10, the participants who received the domain specific   20%   descriptions had a higher success rate in solving the design   10%   problems. This was not the hypothesized results. The kayak  problem, design problem 2, required domain knowledge of  fluid mechanics to select and map the appropriate  characteristics of the airplane onto the kayak. All participate  were expected to have the required domain knowledge but it is  clear that some did not. The domain knowledge influence  dominated this design problem. It also appears it was easier for  participants to retrieve the required domain knowledge when  the analogous product was described in domain specific terms  (for example airplane) than when it was described generally  (Figure 10).   For the dirt bike racer design problem, a retrospective  evaluation of the domain description and the problem revealed   0%   Phase 1: Phase 2: Phase 3: Phase 4: Phase 5:  Open-Ended Constrained List/ Use Use Task 1 Given Analogy,  Design Problem Design Problem Analogies Products Find Solution   Figure 11: Design Problem 2, Experiment 2. The word  “film” in the domain specific description also mapped  well for the design problem. 

 Percentage of Participants with a Solution based on the Correct  Analogous Product: Flour Sifter Design Problem   100%   0%  10%  20%  30%  40%  50%  60%  70%  80%  90% % of participants  with a solution usingthe  correct analogous productGeneral (n=15)  Domain (n=15)  Phase 1: Phase 2: Phase 3: Phase 4: Phase 5:  Open-Ended Constrained List/ Use Use Task 1 Given Analogy,  Design Problem Design Problem Analogies Products Find Solution   Figure 12: Design Problem 3, Experiment 2. A  general description facilitated retrieval and use of the  analogous product to solve a novel design problem.   5.4. Method for Effects of the Design Problem  Representation Experiment 3  The third experiment evaluated representation effects for  both the analogous product and design problem. A 2 X 2  factorial experiment design was employed which resulted in  four experimental groups, Table 1. For both the analogous  product and the problem description, two levels of participants  were compared, a “Domain Specific Description” Group and a  “General Description” Group, Tables 2-3. In addition only the  travel water weight and flour duster design problems along  were used (Figure 13). The focus of this series of experiments  was on understanding how general and domain specific  descriptions influenced a designer’s ability to use an analogy to  solve a problem. Only two of the analogies from Experiments  1 and 2 (water weights and flour sifter) isolated the sentential  representation from other factors that influence design-byanalogy.  Therefore only these two analogies were used for  Experiment 3 which focused on further exploring the influence  of sentential representation. .   The third experiment used the same procedure as the first  two with the following exceptions. In the second task, Solve  the Design Problems, participants were given two design  problems to solve in a series of the following seven phases:   Phase 1: Open-ended design problems, few constraints   Phase 2: Highly constrained design problems   Phase 3: Identify analogies and try using analogies   Phase 4: Continue using analogies   Phase 5: Try to use a function structure to help you find a   solution   Phase 6: Informed task 1 products are analogous   Phase 7: Target analogous product from task 1 is given   Phase 3, list any analogies used and try using analogies,  was 10 minutes long instead of 15 as in experiments 1 & 2  since the experiment 3 contained two design problems instead  of three. An open question from one of our prior experiments   [36] was if the participants were given more time to use  analogies, would they be more likely to find the source analogy  from task 1? Therefore, following the initial phase using  analogies, participants were given ten additional minutes to  continue to use analogies to create solutions.  Table 3: Domain Specific and General Problem   Statements  Domain  Specific  General  Problem Statement for Design Problem 2  Design a kitchen utensil to sprinkle flour over  a counter.  Design a device to disperse a light coating of a  powdered substance that forms clumps over a  surface.   Air Mattress Water-filled Travel Weights  Analogous Products Innovative Solution  Based on  Analogous Product  Problem 1  Distracter Products  Travel  Cart  Toy Flour Duster  Problem 2  Airplane  Pan Cake  Flipper  Figure 13: Analogous products and solutions  based on the analogies, Experiment 3.  Next, participants were shown a set of six function structures  (three per design problem) and asked to develop more solutions  to the constrained design problem (Figure 14). This phase  provided a foundation for evaluating the effectiveness of  function structures for generating novel design solutions.  Function structures are representations used in engineering  design (see 9 and 41 for more detail). When function structures  are created for novel design problems, process choices must be  made. The process choices for the function structures were  made so that they are consistent with the solution based on the  analogous product and were expected to improve participants’  ability to generate a solution. Process choices include using  human energy to actuate the device as opposed to a battery and  electric motor or a gasoline engine. The purpose of  implementing the function structures was to evaluate if this  representation has potential. This experiment does not address  how these functional representations with appropriate process  choices are developed.   9      

 Contain  SubstanceDisperse  PowderedSubstance onSurfaceImport  SubstanceclumpedsubstancepowderhumanenergypowderRelease  SubstanceChangeShape toDisperseChange Shapeto ImportChange Shapeto ContainhumanenergyclumpspowderedsubstancehumanenergytranslationalmotionConvert humanenergy totranslational motionContain  SubstanceDisperse  PowderedSubstance onSurfaceImport  SubstanceclumpedsubstancepowderhumanenergypowderRelease  SubstanceChangeShape toDisperseChange Shapeto ImportChange Shapeto ContainhumanenergyclumpspowderedsubstancehumanenergytranslationalmotionConvert humanenergy totranslational motionContain  Substance  Disperse  Powdered  Substance on  Surface  Import  Substanceclumped  substance  powder  human  energy  powderRelease  Substance  Change  Shape to  Disperse  Change Shape  to Import  Change Shape  to Contain  human  energy  clumps  powdered  substance  human  energy  translational  motion  Convert human  energy to  translational motion  Contain  SubstanceDisperse  PowderedSubstance onSurfaceImport  SubstanceclumpedsubstancepowderhumanenergypowderRelease  SubstanceChangeShape toDisperseChange Shapeto ImportChange Shapeto ContainhumanenergyclumpspowderedsubstancehumanenergytranslationalmotionConvert humanenergy totranslational motionContain  SubstanceDisperse  PowderedSubstance onSurfaceImport  SubstanceclumpedsubstancepowderhumanenergypowderRelease  SubstanceChangeShape toDisperseChange Shapeto ImportChange Shapeto ContainhumanenergyclumpspowderedsubstancehumanenergytranslationalmotionConvert humanenergy totranslational motionContain  Substance  Disperse  Powdered  Substance on  Surface  Import  Substanceclumped  substance  powder  human  energy  powderRelease  Substance  Change  Shape to  Disperse  Change Shape  to Import  Change Shape  to Contain  human  energy  clumps  powdered  substance  human  energy  translational  motion  Convert human  energy to  translational motion  Figure 14: One of the functional model for design problem 2: flour sifter   During one of session experiment 3, a fire alarm occurred  during phase 2. The data was reviewed and little impact was  observed. These four participants were spread across the  conditions and are included in the results.   5.5. Results and Discussion for Effects of the  Design Problem Representation Experiment 3  The analogous product representation and the problem  representation had a clear influence on the designers’ ability to  use the analogy to generate a solution to the design problems.  The trends are similar across the two design problems and  similar to the first two experiments. A summary of the results  for experiment 3 which are relevant to the research questions  are presented in this paper. More detailed results are presented  in [42]. Figure 15a-b show the percentage of participants at  each phase who were able to generate a solution to the design  problems based on the analogous product. Participants who had  previously seen the solution to the design problems based on  the analogous product were removed from the data set (twenty  participants for design problem 1 and three participants for  design problem 2). Participants who memorized the analogous  product in a general form had the highest rate of success. This  result is shown by the green line in the figures, where success  rate increased by up to 40%.   A two-predictor logistic model [43] was fit to the data for  problem 1 at phase 4 to evaluate the statistical significance of  the effects. A multivariate approach could not be used because  too many of the participants had scores for only one of the  design problems since many had previous experience with the  solution for design problem 1. The logistic model for problem 1  at stage 4 shows no significant interaction between the two  predictors and therefore the interaction was removed from the  model (p>0.4). The remaining predictors show the design  problem representation to be a statistically significant predictor  (.=-1.6, p<0.06) and the analogous product representation to  be non-significant (.=1.0, p>0.2). Clearly from the results  plots, the general/domain condition is different from the other  three conditions. Using a binomial probability distribution with  pairwise comparisons between the conditions, the  general/domain condition is statistically significantly different  from the other three conditions (p<0.008, p<0.002, p<0.001)  [43]. The representation of the design problem has a large  effect on the analogies designers retrieve to assist in developing  a solution. The representation of the design problem and the   representation in memory significantly impact the designers’  abilities. Most of the time, the form of representation in  memory is not known so multiple design problem  representations should be used to retrieve more analogies.   A two-predictor logistic model [43] was also fit to the data  for problem 2 at phase 4 to evaluate the statistical significance  of the effects. None of the predictors were statistically  significant. This is likely caused by the low power, due to  limited sample size, of the experiments. Clearly from the plots,  the general/domain condition is different from the other three  conditions. Using a binomial probability distribution with pair  wise comparisons between the conditions, the general/domain  condition is statistically significantly different from the  domain/general condition (p<0.0001) [43].   Figure 15a-d shows when participants found a solution  based on the analogy and also explicitly referenced which  product from task 1 was analogous. Participants could have  labeled the analogy as early as phase three when they were told  to try using design-by-analogy to try to solve the design  problem, but none of the participants explicitly identified the  analogous product until phase five when they were given a  functional model.   6. ADDRESSING THE RESEARCH QUESTIONS  The data provide important insight into the effects the  representation of the problem and representation of analogous  products have on design-by-analogy. The following discussion  provides further elaborates on some of these insights.   6.1. Question 1: How does the linguistic  representation affect a designer’s ability to later use  the analogous product to solve a novel design  problem?  General linguistic representations, which apply both in the  analogous product and design problem domain, increase the  success rate more than domain specific representations. If a  designer stores analogous products in memory in more general  representations, they are more likely to be able to later use  these analogies to solve novel design problems (Figure 15a-d).   10      

 Percentage of Participants with a Solution based on the Correct  Analogous Product: Design Problem 1  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1:  Open-Ended  Design  Problem  Phase 2:  Constrained  Design  Problem  Phase 3:  List/ Use  Analogies  Phase 4:  More  Analogies  Phase 5:  Try a  Functional  Model  Phase 6:  Use Task 1  Products  Phase 7:  Given  Analogy, Find  Solution  %  of  participants  with  a solution  using  the  correct  analogous  productMemory Rep. / Problem Rep.  General / Domain (n=7)  Domain / Domain (n=8)  General / General (n=8)  Domain / General (n=7)  Percentage of Participants with Innovative Solutions Based on  Correct Analogous Product: Design Problem 2  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1:  Open-Ended  Design  Problem  Phase 2:  Constrained  Design  Problem  Phase 3:  List/ Use  Analogies  Phase 4:  More  Analogies  Phase 5:  Try a  Functional  Model  Phase 6:  Use Task 1  Products  Phase 7:  Given  Analogy, Find  Solution  %  of  participants  with  a solution  using  the  correct  analogous  productMemory Rep. / Problem Rep.  General / Domain (n=13)  Domain / Domain (n=13)  General / General (n=11)  Domain / General (n=10)  Figure 15a: Percentage of participants with a  solution based on the target analogous  product at each phase, Design Problem 1.  Figure 15b: Percentage of participants with a  solution based on the target analogous  product at each phase, Design Problem 2.  Percentage of Participants with a Solution based on the Correct  Analogous Product and Identified the Analogous Product:  Design Problem 1  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1:  Open-Ended  Design  Problem  Phase 2:  Constrained  Design  Problem  Phase 3:  List/ Use  Analogies  Phase 4:  More  Analogies  Phase 5:  Try a  Functional  Model  Phase 6:  Use Task 1  Products  Phase 7:  Given  Analogy,  Find Solution  %  of  participants  with  a solution  using  the correctanalogous  productMemory Rep. / Problem Rep.  General / Domain (n=7)  Domain / Domain (n=8)  General / General (n=8)  Domain / General (n=7)  Percentage of Participants with Innovative Solutions Based on  Correct Analogous Product and Identified the Analogous Product:  Design Problem 2  0%  10%  20%  30%  40%  50%  60%  70%  80%  90%  100%  Phase 1:  Open-Ended  Design  Problem  Phase 2:  Constrained  Design  Problem  Phase 3:  List/ Use  Analogies  Phase 4:  More  Analogies  Phase 5:  Try a  Functional  Model  Phase 6:  Use Task 1  Products  Phase 7:  Given  Analogy, Find  Solution  %  of  participants  with  a  solution  using  thecorrect  analogous  productMemory Rep. / Problem Rep.  General / Domain (n=13)  Domain / Domain (n=13)  General / General (n=11)  Domain / General (n=10)  Figure 15c: Percentage of participants who  had a solution based on the target analogous  product and also identified the analogy at  each phase, Design Problem 1.  Figure 15d: Percentage of participants who  had a solution based on the target analogous  product and also identified the analogy at  each phase, Design Problem 2   This result has important implications for teaching  designers to think about and remember design solutions they  encounter. If they seek representations that apply across more  domains and in more general forms, they will be much more  likely to be able to use the design in the future. For example,  framing an air mattress as “a device that uses a substance from  the environment it is used in”, rather than “a device that is  filled with air” makes it much more likely to be used in future  design problems that seek innovative solutions.   6.2. Question 2: How does the representation of  the problem statement affect the ability of a designer  to retrieve and use a relevant analogous product to  find a solution to a new design problem?  The representation of design problems clearly influences a  designer’s ability to generate analogous solutions (Figure 15ad).  The representation that will give the designer the highest  probability of exposing or generating an analogous solution  depends on how the analogous solution is stored in memory. If  the analogous product is stored in a general form, then a  domain specific representation is the most efficient means to  retrieve it. Generally, it is not known in advance what  representation is most likely to retrieve the desired information.  This means that the best approach for seeking analogous   solutions is to use multiple representations that vary across the  range of domain specific or domain general.   This experiment also provided a basic study of the potential  for function structures (functional models) to enhance the  design-by-analogy process. Participants were given function  structures with process choices which are consistent with the  analogous solutions. These function structures also included  linguistic functional descriptions that were different from the  given problem statements. This experiment does not address  how the participants would go about developing these  particular function structures. This experiment addresses the  question that if given an appropriate function structure, does it  increase the likelihood of generating an analogous solution?  From the results, there is a clear increase in phase six when  participant use the function structures to assist in generating  solutions. This result is exciting and a validation of anecdotal  claims about an important role of functional modeling in  design. Function structures are another potential representation  that will enhance the design process and should be included in  the search for analogous solutions. Diagrammatic  representations merit further investigation.   6.3. Question 3: What is the best way to represent  a design problem when the representation in memory  11      

 is not known and what implication does this have for  a design-by-analogy method?   For any design task, a number of representations should be  created with a varying semantics. Typically it is not known how  relevant analogies are represented in memory and which  retrieval cues are required. Therefore a number of  representations and therefore retrieval cues should be created to  maximize the probability a useful analogy will be found.   7. DISCUSSION  OF ADDITIONAL RESULTS AND  OTHER IMPLICATIONS  The series of experiments provide results and implications  beyond the research questions.   Analogy identification and implications for naturalistic  analogy research   Designers frequently use analogies to solve design  problem without realizing the source of the idea. The  participants used analogies to solve the design problems, but  did not mention that they were using analogies and/or did not  realize that their solutions were analogous to previously  experienced products until a later phase (Figure 15a-d). If the  designers had realized the source of the idea, they would have  listed the analogy at a much earlier phase (Figure 15c-d).  Instructing subjects to use analogies and list the analogies they  had used caused little effect.   Our findings replicate the work of Schunn and Dunbar  [31], but for an independent data set and in the engineering  domain. Schunn and Dunbar found that participants often used  analogies to solve difficult insight problems, but the subjects  did not realize they were doing so. One implication of this  result is that analogies play an important role in problem  solving, but do so, at least in part, outside of awareness.  Another implication is that, in naturalistic observation studies,  simply recording how often people say they are using analogies  is likely to underestimate their true frequency. For example,  imagine an investigator who seeks to determine how important  analogies are in generating new designs. This researcher  decides to observe expert designers at their workplace  generating novel designs and counts the number of times the  experts say “this is just like [some other product]”. Intuitively,  this procedure seems reasonable, but our data suggest that it  will underestimate the role of analogies. These results also  indicate that designers frequently use analogy without  recognizing it. This implies that design by analogy has an even  greater impact on the design process than what is currently  indicated by the anecdotal evidence.   7.1. Design constraints guide move designers to  search particular areas of the design space.  We hypothesize the application of design constraints can lead  designers to search particular regions of the design space.  Systematically adding and removing constraints may assist the  designer in thoroughly searching portions of the design space.  This approach has potential as part of a design method. The  experiments presented in this paper intentionally constrained   the design space to areas where it was known that good  solutions existed. The constraints required participants to  search particular regions of the design space.   8. CONCLUSIONS AND FUTURE WORK  Design-by-analogy is a powerful tool in a designer’s toolbox,  but few designers have the methods to harness its full capacity.  Simply recognizing its potential and attempting to search  mentally for analogies is not enough. Designers need methods  and tools to support this process. They need approaches for  when they feel they have run out of ideas. They need methods  to represent the problem in a multitude of representations. The  right representations have the potential to increase a designers’  probability of success by up to 40%. These methods need to be  built on a solid understanding of human capacity combined  with scientific design knowledge. This experiment  demonstrates, at least foundationally, the impact the right  representation has on the design by analogy process.   This paper explores a limited set of influences on the  design-by-analogy process and highlights only a few of the  potential levers that a design methodology may take advantage  of. The analogy between the airplane and the novel kayak  design illustrate the influence of domain knowledge in this  process. This warrants further experimental exploration. A  design methodology will needs tools to highlight areas where  domain knowledge is lacking and approaches to facilitate the  recognition of the underlying principles.   Design-by-analogy is a common occurrence in the design  process. Designers frequently use analogous products without  recognizing the origin of the idea. Participants who have been  exposed to the technique of design-by-analogy will  spontaneously use it when asked to generate design solutions.  Design-by-analogy is not limited to an elite few designers who  learn to harness its power but it is a commonplace approach.   A deeper understanding of the mechanism behind  analogical reasoning and their implications within design will  guide the development of drastically improved design-byanalogy  methods and tools for design innovation.  Representation clearly matters and seeking improved  representations has great potential for significantly enhancing  the innovation process.   8.1. Future Work  Future work will focus on developing new design approaches  and methods to increase the quantity, quality, novelty and  variety of innovative solutions based on the knowledge gained  from the experiments presented and other relevant literature.  Greater exploration of the use of functional models and other  representations for assisting in the design process will also be  investigated. New methodologies will be validated through  controlled experiments and with professional designers.   ACKNOWLEDGMENTS   The authors would like to acknowledge the support provided  from the Cullen Endowed Professorship in Engineering, The   12       

 University of Texas at Austin and the National Science  Foundation under Grant No. CMMI-0555851. This research  was also supported by a Fellowship in the IC2 Institute given to  Dr. Arthur Markman. 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 sponsors.  The authors would also like to thank Rachel Kuhr for her  assistance in the development of the graphics.   REFERENCES   [1] Christensen, B. T., and Schunn, C. D. “The relationship of  analogical distance to analogical function and pre-inventive  structure: The case of engineering design.” Memory &  Cognition, (in press).  [2] Leclercq, P.  and Heylighen, A. “5,8 Analogies per Hour,”  Artificial Intelligence in Design '02. J. S. Gero (ed.), 285303,  2002  [3] Casakin, H., and Goldschmidt, G., 1999, “Expertise and the  Use of Visual Analogy: Implications for Design  Education,” Design Studies, 20(2), pp. 153-175.  [4] “Wings Take to the Water,” 2000, Tuesday Nov. 7th, BBC  News, http://news.bbc.co.uk/1/hi/sci/tech/1011107.stm,  accessed 4/2006.  [5] Reed, J., 2006, “The Future of Shipping”, Popular Science,  May, pp. 50-51.  [6] Gordon, W. J. J., 1961, Synectics: The Development of  Creative Capacity, Harper and Brothers, New York.  [7] French, M., 1996, Conceptual Design, Springer-Verlag,  London, UK.  [8] Hacco, E., and Shu, L. H., 2002, “Biomimetic Concept  Generation Applied to Design for Remanufacture,”  Proceedings of the DETC 2002, ASME 2002 Design  Engineering Technical Conferences and Computer and  Information in Engineering Conference, Montreal, Quebec,  Canada.   [9] McAdams, D. and Wood, K., 2002, “A Quantitative  Similarity Metric for Design by Analogy,” ASME Journal  of Mechanical Design, 124(2), pp. 173-182.  [10] French, M., 1988, Invention and Evolution: Design in  Nature and Engineering, Cambridge University Press,  Cambridge,UK.  [11] Shu, L. H., Hansen, H. N. Gegeckait, A., Moon, J., Chan,  C., 2006, “Case Study in Biomimetic Design: Handling and  Assembly of Microparts,” Proceedings of ASME Design  Theory and Methodology Conference, Philadelphia, PA.  [12] Markman, A.B., 2002, “Knowledge Representation”, in D.  Medin, & H. Pashler (eds.) Stevens’ Handbook of  Experimental Psychology, 2, 3rd ed. Wiley, New York, pp.  167.  [13] Ball, L. J., Ormerod, T. C., Morley, N. J., 2004,  “Spontaneous Analogizing in Engineering Design: A  Comparative Analysis of Experts and Novices”, Design  Studies, 25(5), pp. 495-508.  [14] Kolodner, J. L., 1997, “Educational Implications of  Analogy: A View from Case-Based Reasoning”,  American Psychologist, 52(1), pp. 57 – 66.  [15] Kryssanov, V. V., Tamaki, H. and Kitamura, S., 2001,  “Understanding Design Fundamentals: How Synthesis  and Analysis Drive Creativity, Resulting in Emergence,”  Artificial Intelligence in Engineering, 15, pp. 329 – 342.  [16] Goldschmidt, G. and Weil, M., 1998, “Contents and  Structure in Design Reasoning”, Design Issues, 14(3), pp.  85-100.  [17] Clement, C. A.,1994, “Effect of Structural Embedding on  Analogical Transfer: Manifest versus Latent Analogs,”  American Journal of Psychology, 107(1), pp. 1-39.  [18] Clement, C. A., Mawby, R. and Giles, D. E., 1994, “The  Effects of Manifest Relational Similarity on Analog  Retrieval,” Journal of Memory & Language, 33(3), pp.  396-420.  [19] Chiu, M., 2003, “Design Moves in Situated Design with  Case-based Reasoning”, Design Studies, 24, pp. 1 – 25.  [20] Gentner, D., 1983, “Structure Mapping – A Theoretical  Framework,” Cognitive Science, 7, pp. 155 – 177.  [21] Falkenhainer, B.F., Forbus, K.D., and Gentner, D., 1989,  “The Structure Mapping engine: Algorithm and Examples,”  Artificial Intelligence, 41(1), pp. 1-63.  [22] Gentner, D. and Markman, A.B., 1997, “Structure  Mapping in Analogy and Similarity,” American  Psychologist, 52, pp. 45-56.  [23] Gentner, D., Holyoak, K. J., and Kokinov, B. (Eds.), 2001,  The Analogical Mind. The MIT Press, Cambridge,  Massachusetts.  [24] Blanchette, I., and Dunbar, K., 2001, “Analogy Use in  Naturalistic Settings: The Influence of Audience, Emotion,  and Goals,” Memory and Cognition, 29(5), pp. 730-735.  [25] Hummel, J. E., and Holyoak, K. J., 1997, “Distributed  Representations of Structure: A Theory of Analogical  Access and Mapping”, Psychological Review, 104(3), pp.  427-466.  [26] Gick, M. L., and Holyoak, K. J., 1980 “Analogical  Problem Solving”, Cognitive Psychology,” 12, pp. 306355.  [27] Dunbar, K., 1997, “How Scientists Think: On-line  creativity and conceptual change in science,” In T.B.  Ward, S.M. Smith, and J. Vaid (eds.), Creative Thought:  An investigation of conceptual structures and processes.  American Psychological Association, Washington, DC.  [28] Chi, M.T.H., Feltovich, P.J., and Glaser, R., 1981,  “Categorization and Representation of Physics Problems  by Experts and Novices,” Cognitive Science, 5, pp. 121132.  [29] Gentner, D. and Landers, R., 1985, “Analogical  Remindings: A good match is hard to find,” Proceedings  of the International Conference on Systems, Man, and  Cybernetics. Tucson, AZ.  13       

 [30] Novick, L.R., 1988, “Analogical Transfer, Problem  Similarity, and Expertise,” Journal of Experimental  Psychology: Learning, Memory, and Cognition, 14(3), pp.  510-520.  [31] Schunn, C.D., and Dunbar, K., 1996, “Priming, Analogy,  and Awareness in Complex Reasoning,” Memory &  Cognition, 24(3), 271-284.  [32] Anderson, J.R., 1983, “A Spreading Activation Theory of  Memory.” Journal of Verbal Learning and Verbal  Behavior, 22, pp. 261-295.  [33] Collins, A.M., and Loftus, E.F., 1975, “A Spreading- Activation Theory of Semantic Priming.” Psychological  Review, 82, pp. 407-428.  [34] Roediger, H.L., Marsh, E.J., and Lee, S.C., 2002, “Kinds  of Memory,” in D. Medin, and H. Pashler (eds.) Stevens’  Handbook of Experimental Psychology, 2, Wiley, New  York.  [35] Thompson, L., Gentner, D., and Loewenstein, J., 2000,  “Avoiding Missed Opportunities in Managerial Life:  Analogical Training More Powerful than Individual Case  Training”, Organizational Behavior and Human Decision  Processes, 82(1), pp. 60-75.  [36] Linsey, J. S., Murphy, J. T., Wood, K. L., Markman, A. B.,  and Kurtoglu, T., 2006, “Representing Analogies:  Increasing the Probability of Success,” Proceedings of  ASME Design Theory and Methodology Conference,  Philadelphia, PA.  [37]  Regenold, S., 2006 “The Kayak that Flies Over  Water,”Popular Science, April, pp. 26.  [38] Kelley, T., and Littman, J., 2001, The Art of Innovation:  Lessons in Creativity from IDEO, America's Leading  Design Firm, Doubleday Publishing.  [39] Devore, J. L., 1999, Probability and Statistics for  Engineering and the Sciences, Duxbury, United States.  [40] Schooler, J. W., Fiore, S. M., and Brandimonte, M. A.,  1997, “At a Loss From Words: Verbal Overshadowing of  Perceptual Memories,” In. D. L. Medin, The Psychology  of Learning and Motivation, Academic Press, New York,  pp. 291-340.  [41] Otto, K. and Wood, K., 2001, Product Design Techniques  in Reverse Engineering and New Product Development,  Prentice Hall, Upper Saddle River, New Jersey.  [42] Linsey, J., Laux, J., Wood, K., and Markman, A., 2007,  “Effects of Analogous Product Representation on Future  Design-by-Analogy,” Proceedings of the 2007  International Conference on Engineering Design, Paris,  France.  [43] Kutner, M. H., Nachtsheim, C. J., Neter, J., Li, W., 2005,  Applied Linear Statistical Models, McGraw-Hill, Boston.  APPENDIX A: SOURCE ANALOGY AND DESIGN  PROBLEM DESCRIPTIONS FOR EXPERIMENT 2    (see [37] for Experiment 1 descriptions)   Design Problem 1   Design a fast kayak.   Design Problem 1- Additional Constraints   Design a fast kayak.   •  A person is the only available power source.  •  It must have a top speed of greater than 14 mph. Currently,  typical human-powered boats have a top speed of less than 6  mph even for top athletes.  •  The top speed is limited by drag, the faster a boat goes the  greater the drag.  •  Your design must reduce the drag.  Design Problem 2   Design a set of goggles that remove dirt and mud from a dirt  bike racer's goggles.   Design Problem 2- Additional Constraints   Design a set of goggles that remove dirt and mud from a dirt  bike racer's goggles.   •  Forcing the dirt and mud across the goggle's surface creates  scratches. The goggle system must not scratch the surface of  the goggles.  •  The dirt and mud can not be forced across the surface of the  goggles.  •  The dirt bike racer's hands cannot leave the handle bars of  the bike.  •  A section of the goggles at least 1" by 2” must be completely  clean.  Design Problem 3   Design a kitchen utensil to sprinkle flour over a counter.   Design Problem 3- Additional Constraints   Design a kitchen utensil to sprinkle flour over a counter.   •  The only material that is available to build the kitchen utensil  from is various thicknesses of stainless steel wire.  •  The entire kitchen utensil must be made from only one  thickness of wire.  •  The kitchen utensil must be manufactured by bending and  cutting the wire only.  •  The kitchen utensil must be capable of containing the flour  and carrying the flour 1 meter without losing the flour.  14       

 Pepper Mill   A small motor inside the pepper mill spins and  grinds the peppercorns into ground pepper. A  person presses a button to actuate a small dc  motor.   Device to create a fine powder    A small motor inside the device turns and breaks down the  substance into a fine powder. A force presses a switch to turn  on a small actuator.   Film in a camera   Two reels feed the film in front of the stream  of light. The film captures the image and  then a new unexposed section of film is moved into place.    Film in a camera   Two reels move a surface in the path of the incoming  substance. The surface collects the substance and then a new  unchanged surface is moved into place.   Football   A person throws the American football. As it  flies through the air it spirals. This spiraling  reduces air friction allowing the ball to travel farther.    Moving Device   Another object gives energy to this device.  As it moves through a substance it turns  about. This motion changes the forces  allowing the device to move more.   Airplane   An airplane flies through the air allowing  rapid flights. The airfoil shape of the Section view  airplane's wings causes a lift force as the of a Wing  plane flies through the air. As the plane increase altitude there  is less drag due to the air being less dense.  Device for Rapid Travel  This device moves through a fluid allowing rapid travel. The  shape of the device’s extensions causes a net force as the  device moves through the fluid. As the device changes position  there is less resistance due to the fluid being less dense.  Shape-o-toy  This toy serves a number of purposes. The  two halves contain the blocks allowing them  to be carried. A child pulls the two halves  apart and the blocks fall out.  Device to hold and release substances  This device serves a number of functions. The two sections  hold the substances allowing them to be moved. A force  separates the sections and the substance is released.    APPENDIX A: DESIGN PROBLEM DESCRIPTIONS  FOR EXPERIMENTS 3    Source analogies are the same as Experiments 1 & 2   General: Design Problem 1   Design an exercise device that can be carried for long distances  in a 3 ft3 container   General: Design Problem 1- Additional Constraints   Design an exercise device that can be carried for long distances  in a 3 ft3 container. Here are the additional requirements:   •  Provides at least 15 lbs of resistance  •  Adds less than 4 lbs to the 3 ft3 container.  •  Maximum volume is 120 in3 (~750 cm3).  •  It must be capable of being used for movements normally  done with hand weights  •  It cannot use strips or cords of elastomer (rubber) for  resistance.  Domain: Design Problem 1   Design a piece of exercise equipment that can be carried in a  suitcase.   Domain: Design Problem 1- Additional Constraints   Design a piece of exercise equipment that can be carried in a  suitcase. Here are the additional requirements:   •  Provides at least 15 lbs of resistance  •  Adds less than 4 lbs to the suitcase  •  Maximum volume is 120 in3 (~750 cm3) or about half the  size of a briefcase.  •  It must be capable of being used for exercises normally  done with hand weights  It cannot use strips or cords of elastomer (rubber) for resistance  General: Design Problem 2  Design a device to disperse a light coating of a powdered  substance that forms clumps over a surface.  General: Design Problem 2- Additional Constraints  Design a device to disperse a light coating of a powdered  substance that forms clumps over a surface.   •  The only material that is available to build the device from  is various thicknesses of stainless steel wire.  •  The entire device must be made from only one thickness of  wire.  •  The device must be manufactured by bending and cutting  the wire only.  •  The device must be capable of containing the powdered  substance and carrying the powdered substance 1 meter  without losing the powdered substance.  Domain: Design Problem 2   Design a kitchen utensil to sprinkle flour over a counter.   Domain: Design Problem 2- Additional Constraints   Design a kitchen utensil to sprinkle flour over a counter.   •  The only material that is available to build the kitchen  utensil from is various thicknesses of stainless steel wire.  •  The entire kitchen utensil must be made from only one  thickness of wire.   •  The kitchen utensil must be manufactured  by bending and cutting the wire only.  •  The kitchen utensil must be capable of  containing the flour and carrying the flour  1 meter without losing the flour.  15