This article was downloaded by: [University of Michigan] On: 03 October 2011, At: 14:13 Publisher: Taylor & Francis Informa Ltd Registered in England

and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Engineering Design
Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/cjen20

Incorporating user shape preference in engineering design optimisation
Jarod C. Kelly , Pierre Maheut , Jean-François Petiot & Panos Y. Papalambros
a a a b b

Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA
b

Institut de Recherche en Communications et Cybernétique de Nantes, Ecole Centrale de Nantes, 1, rue de la Noe, Nantes, 44321, France Available online: 24 Jun 2011

To cite this article: Jarod C. Kelly, Pierre Maheut, Jean-François Petiot & Panos Y. Papalambros (2011): Incorporating user shape preference in engineering design optimisation, Journal of Engineering Design, 22:9, 627-650 To link to this article: http://dx.doi.org/10.1080/09544821003662601

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-andconditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Journal of Engineering Design Vol. 22, No. 9, September 2011, 627–650

Incorporating user shape preference in engineering design optimisation
Jarod C. Kellya * Pierre Maheutb , Jean-François Petiotb and Panos Y. Papalambrosa
a Department

Downloaded by [University of Michigan] at 14:13 03 October 2011

of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA; b Institut de Recherche en Communications et Cybernétique de Nantes, Ecole Centrale de Nantes, 1, rue de la Noe, Nantes 44321, France
(Received 16 July 2009; final version received 26 January 2010 )

Form versus function is a classic design debate. In this article, a practical approach to combine shape preference (form) and engineering performance (function) under a design optimisation paradigm is proposed and implemented. This synthesis allows form and function to be considered in quantitative terms during the design process to identify shapes that can benefit overall product design. Two methods of preference modelling, PREFMAP analysis and conjoint analysis, are used to model user preference as a mathematical function of design variables. Physics-based models express engineering performance as functions of the same design variables. The models are combined in an optimisation formulation to capture the design trade-offs involved. A simple illustrative study of bottle design is presented. A divergence is found between the most preferred shape and the technically optimal shape; a Pareto frontier provides insight into the trade-off between these two goals. Keywords: preference modelling; shape preference; decision-making; product design

1.

Introduction

Mathematical design optimisation traditionally deals with design considerations related to product functionality because the availability of physics-based models allows for quantitative expression of product performance as a function of design variables (Papalambros and Wilde 2000). Extending such a quantitative approach to include subjective design considerations makes design optimisation more valuable but also substantially more challenging. Modelling methods from the behavioural sciences offer a foundation for developing quantitative behavioural models, such as user preferences for certain design aspects. Shape is a product aspect that often affects its performance. Shape also affects user preference for the product, often associated with aesthetic preference. Shape preferences may be motivated by non-aesthetic considerations, e.g. ergonomics, and so it is difficult to extract an objective metric for purely aesthetic shape preference. Preference modelling for optimal product design related to shape would be more manageable if the shape preference metric does not attempt to include causality for the preference. This is
*Corresponding author. Email: jckelly@umich.edu

ISSN 0954-4828 print/ISSN 1466-1837 online © 2011 Taylor & Francis DOI: 10.1080/09544821003662601 http://www.informaworld.com

g. In this article. manufacturing limitations. As noted earlier. while unscientific. in industrial. and proposed ‘quantitative methods that matched the perceptual and mental processes of (2D interface) users’. Even without explicit understanding of preference causality. graphic or architectural design.C. We further postulate that one may be able to identify combinations of shape attributes that yield an optimally preferred design for a particular artefact within a particular context. Norman 2002. Kansei Engineering. the context of the object’s function. Park et al. a and b. Industrial and architectural design has a rich tradition of studying aesthetics (e. a b 2 (1) Downloaded by [University of Michigan] at 14:13 03 October 2011 . defined as the ratio of the length of two line segments. indeed beauty. Application of these heuristics does not inherently dictate a product’s appeal. engineers and product planners. Orsborn and Cagan 2009.g. 2005). including quantitative shape metrics such as balance. That work specifically provided a numerical quantification of the effects of symmetry.628 J. 2005). 2003. applicability of design rules becomes tenuous. 2009). for example. MacDonald 2001. mathematical metric for user shape preference is plausible. Liu (2003b) has noted that. Combining marketing models with engineering ones in a decision-making framework is a more recent topic in the emerging area of design for market systems. shape preference simply means that one product shape is more liked than others in the given context of how the product will be used. However. Then. Petiot and Chablat 2003. or heuristics. The marketing literature on utilising preference to guide product design is expansive. within the confines of shape constraints (e. consistent with modelling approaches in marketing (Tybout and Hauser 1981). Orsborn et al. devised formulas to account for each of these compositional elements for 2D interfaces and validated them through human experimentation. such that the ratio of a to b is equal to the ratio of a + b to a: √ a+b a 1+ 5 = ≈ . a quantitative study of trade-offs relating preferred product shape to functionally optimal product shape. ergonomics).g. symmetry. The golden section is the best-known quantification of shape preference. We cannot extend the findings or shape preference from the present work to suggest a universal shape preference. designers successfully apply rules. subject to the physical constraints dictated by required engineering performance. we limit the notion of shape preference and say nothing about the inherent ‘beauty’ of a product. Pye 1978. can provide valuable insights to designers. This tradition suggests that a quantitative. 2003b) has suggested the idea of engineering aesthetics to seek methods that help designers make better decisions regarding the subjective aesthetic qualities of a product. these design heuristics ‘offer important insights into aesthetic questions and provide useful perspectives from which we can examine aesthetic concepts’. Such a combination of approaches (and models) is necessary to avoid dislocated and even infeasible designs (Michalek et al. Lidwell et al. Bauerly and Liu (2006) focused on the graphical layout of displays and web pages. Petiot and Grognet 2006. 1997). rhythm and proportionality. balance and compositional blocking. Park 2004. Preferences are also limited to the geometric qualities of the product that users are permitted to vary. Bauerly identified these as the three compositional elements of aesthetic judgment. In the following. because this does not account for many contextual issues. Liu (2003a. utilises semantic information to match user wants and expectations in product design (Nagamachi 1995. initially developed in the 1970s and often referred to synonymously as Emotional Engineering or Emotional Design. Kelly et al. Assessing subjective tastes of users and utilising that information in the product design process is critical for success in a competitive marketplace and has received significant research attention (e. heuristics provide designers with a structure around the creative process of design (Tjalve 1979). preference for compositional characteristics can be context dependent.

a respondent would be shown several different designs defined by variations in the design variables. x2 ) is the model of preference. . This provides response data that can then be used in a mathematical regression. x1 and x2 are design variables and B = {b0 . A designed object examined in PREFMAP is composed of design attributes that are continuous. 2. Specific designs from within this bounded design space can then be used to query users and develop a preference model. The respondent would then assign a numeric value. to the paraboloid defined in Equation (2). Coxon et al. associated with their preference. to each design that they observe. Equation (2) gives the definition of this form in the case of two design variables. x1 and x2 . The first psychophysical tests on this ratio (Fechner 1897) have been methodologically challenged (Green 1975). the stimuli space is based on data obtained independently to the preference assessment (Chang and Carroll 1972. b5 }. then we could determine brightness preference by mapping the light samples’ preference onto a stimulus space of lumens as the external map. Section 2 gives some background on PREFMAP and conjoint analysis. The governing equation for the PREFMAP model is 2 2 Pmodel (x1 . B = {b0 . all of the coefficients must be estimated in this phase. This paper examines how preference models of shape can be utilised within design optimisation for product development. called phases. are defined (Figure 1). and a specific case study that incorporates shape preference models within design optimisation. 1982). Phase I is the most general model. So. b1 . if several examples of light were shown to participants and the participants rated their preference for each sample. four types of models. . and a discussion on the obtained optimisation results. 2. in a PREFMAP survey. . The constants. are determined by minimising the Euclidean . b1 . . The model fits user response data. Section 5 offers conclusions and suggestions for future work. In an external mapping of preference. PREFMAP PREFMAP analysis is a tool that relates preference data to a stimuli space in order to generate an external mapping of preference (Chang and Carroll 1972). but bounded. It corresponds to an elliptical paraboloid that can be rotated within the plane. .Journal of Engineering Design 629 Downloaded by [University of Michigan] at 14:13 03 October 2011 This ratio has been extensively studied and found to be relevant to many mathematical and biological phenomena (Green 1995). This includes the examination of methods for understanding shape preference. x2 ) = b0 x1 + b1 x2 + b2 x1 + b3 x2 + b4 x1 x2 + b5 . . Section 3 describes modelling examples to further examine the preference models. individually or in the aggregate. Section 4 presents the linking of behavioural (preference) and engineering models in a design optimisation framework for a bottle.1. b5 } are constants. In the following. The different phases of the model are defined according to the nullity of certain coefficients. which are presented in Figure 1. According to the complexity of the model. . but the golden ratio appeal remains strong and is still used in current products. The principle of PREFMAP is to model the preference by a quadratic form. the two preference modelling methods used in the subsequent study. For instance. . Preference modelling with PREFMAP and conjoint analysis A brief review of PREFMAP analysis and conjoint analysis provides some necessary background directly related to the studies in Section 3. (2) where Pmodel (x1 .

(4) The basic idea behind PREFMAP is that each individual has an ideal point of maximum preference and is capable of ranking different stimuli in such a way that the ideal point is revealed (Coxon et al. b4 x1 x2 . j ))2 . For Phase IV. from Equation (2). 1982). .630 J. The preference space (Pmodel in Equation (2)) would be defined by two design variables. Downloaded by [University of Michigan] at 14:13 03 October 2011 distance between observed user response data. ∂ 2 Pmodel 2 ∂x1 ∂ 2 Pmodel 2 ∂x2 − ∂ 2 Pmodel ∂x1 ∂x2 2 < 0. A circular shape indicates an equal influence of each variable on the resulting preference. The model is described by six parameters. and Pmodel (x1 . consider a person that likes large vehicles with high amounts of power. Kelly et al. This can be done using the following optimisation formulation: min F (B) = i j (Pmodel (i. The Phase II portion of the PREFMAP model is obtained by removing the interaction term. As an example related to vehicle preference. The different phases of PREFMAP are suited to different situations. Phase I or II. . The external stimuli space is simply the two design variables weighted on a linear basis. Phase II is like Phase I. Figure 1 visually shows the difference between the four phases. x2 ). the possibility of a saddle point exists. B = {b0 . (3) which is simply the criterion for least-squares linear regression. j ) − Pobs (i. The model of preference would describe a paraboloid whose function value was maximal at a design point of large size and high power. Phase I x2 Phase II x2 x1 Phase 1V b2 = b3 = b4 = 0 x2 x2 x1 x1 b4 = 0 x1 Phase III b4 = 0 b2 = b3 Figure 1. Pobs (x1 . . one can imagine that the ideal point predicted in Phase III is very far away from the actual stimuli tested. Phase III is simpler yet. . The ideal point assumption appears to be axiomatic. in either case. b5 }. However. respectively). A linear model only indicates the direction of increasing preference. except that the paraboloid is not rotated (thus the variables are decoupled).C. Rotated ellipses indicate a coupled relationship between design variables while a non-rotated ellipse indicates that one variable has a greater impact on preference than the other. as the paraboloid is circular. . perhaps corresponding to a sport utility vehicle. Saddle points occur when the determinant of the Hessian matrix is negative. x2 ). Phase IV is a vector model where the vector can be thought of as pointing in a direction of ever increasing preference for the given design attributes associated with the stimuli space. b1 . thus iso-preference curves of the circular map are nearly parallel and suggest a gradient of ascent towards the ideal point. vehicle size and vehicle power (x1 and x2 in Equation (2).

Or. While δj mn represents a dummy variable that is equal to unity if the level of attribute m in product j is n. 2001) and marketing communities. thus treating a group as a single individual and negating the need for the q term. ‘as large as an orange’ or ‘as large as a grapefruit’. Conjoint analysis Conjoint analysis is a method for modelling preference that has been used extensively within the social science (Ryan and Farrar 2000. m. (7) vj = m=1 n=1 βmn δj mn . The method is based on the principles of utility. and σ is a scale parameter. viq . . 2000). we consider n to be the same for all m. A no-choice alternative is included in the product offerings. a popular form of evaluation is through selection of one product among a set. Downloaded by [University of Michigan] at 14:13 03 October 2011 (5) The logit model assumes that the unobserved error term is randomly distributed and has a double exponential probability distribution (Gumbel distribution) (Guadagni and Little 1983. has been shown in studies to agree well with the results of a normally distributed probability function. but in general the number of levels. The mathematical formulation of viq can be broken down.Journal of Engineering Design 631 2. the attribute-level can be more technical. primarily with reference to the use of the logit model (Louviere et al. and the notion that consumers attempt to maximise product utility when they make choices. When using discrete choice analysis for the assessment of utility functions. The attribute may be ‘size’ and the levels may be ‘as large as a marble’. do not have to be equal for each attribute. as in ‘0–60 mph acceleration time’. Michalek presented a review of the foundations of discrete choice analysis. respondents are shown several descriptions or images of potential products. which has more theoretical validity (Michalek 2004). These attributes are further decomposed into levels. Green et al. where i and q are the product offering and individual. (8) Here βmn represents the part-worth of level n for attribute m. each having n-levels. such that each i consists of an attribute term.2. iq . This type of conjoint analysis is known as discrete choice analysis. This assumption. is composed of a deterministic term. m. Michalek 2004). (6) Here μ is a location parameter. with levels of ‘10’. Louviere et al. uiq = viq + iq . f ( ) = exp − exp − −μ σ . it is assumed that all individuals act in a consistent and similar manner. it is assumed that utility. concept evaluation. the new model becomes Pi = M e vi J vj j =1 e N . Each product is of a similar nature. Otherwise δj mn is zero. Presuming that the deterministic component of utility can be predicted through regression of observed choice data to mathematical models of utility yields the mathematically tractable multinomial logit (MNL) model. The goal of conjoint analysis is to determine the ideal combination of feature attributes based on the preference responses of a participant or group. It is frequently used in industry to determine information related to product design. In order to collect data for a conjoint analysis. Further. ‘12’ and ‘15 s’. and an error term. product positioning and market segmentation (Green and Srinivasan 1990). Therefore. 2000. Respondents are then asked to evaluate the products in some fashion. n. respectively. while not being based on theoretical grounds. but the product is decomposed into characteristic attributes (variables). uiq . In our case.

The two analysis models were then tested in their ability to reproduce this function using only information requested by the querying tool of each technique and subject to that model’s own constraints. or ‘goal’ preference. while the model with interaction consists of (m × (n − 1) + p × (p − 1) + 1) terms. and that the levels in p are exactly the same as the similarly associated levels in n.632 J. So. only the difference between levels matter. Here. This discrete model of preference can then be made continuous by fitting a cubic spline through the design points (Michalek et al. In utility. This can be modelled using a form that takes into account second-order effects between the attributes. when reporting β values we will include the redundant terms. Kelly et al. Each attribute is considered independent of the other attributes in affecting utility. While this makes axiomatic sense to the authors. a model is again presented that consists of m-attributes and n-levels: vj = m n βmn δj mn + o p βmnop δj mnop . Note that the attributes comprising set o are exactly the same attributes as those comprising set m. respectively. and to further explain PREFMAP analysis and conjoint analysis to the reader. A unimodal mathematical function has a single global maximum. 2005). and 29 are needed to account for interaction effects. or βmnop . We explore both unimodal and bimodal mathematical goal functions because they represent different levels of . The number of parameters to estimate in this model depends on the number of attributes and levels examined. where it is likely that the utility of an offering is not composed of independent attributes. it is possible that discontinuities exist. This is not a comparison of PREFMAP analysis against conjoint analysis. we used a test function representing ‘real’. in this paper. while conjoint analysis is applied to discrete variables. both PREFMAP and discrete choice analysis were used to develop a mathematical representation of both a predefined unimodal and bimodal mathematical function. is equal to zero. In these experiments. thereby yielding a total of 11 (or 36) terms depending on the model type. PREFMAP is based upon continuous variable values. and this is often done such that the sum of each βmn . The model without interaction terms consists of (m × (n − 1) + 1) terms.C. It can be difficult to conduct conjoint analysis studies as the number of attributes and their associated levels increase because the number of responses required of a participant becomes very high. 3. (9) Downloaded by [University of Michigan] at 14:13 03 October 2011 Now. This assumes that preference is continuous in relation to continuous variables across the entire design domain. for a two-attribute five-level model. while a bimodal mathematical function has two local maxima. However. However. Evaluating the merit of one model explicitly against the other is not realistic in this study because the necessary input data for each model is very different. it is an illustration of these two different modelling tools and the preference models that result from their application. Conjoint analysis is often conducted on groups in order to obtain generalised information regarding preference for various attribute levels for use in designing products. such as shape preference. The inclusion of interaction terms in an analysis is warranted in situations. a total of nine parameters are needed not accounting for interaction. thus it is necessary to normalise the β values. δj mnop = 1 when alternative j possesses attributes m and o at levels n and p. Preference tool examination To illustrate some differences between the PREFMAP analysis and MNL conjoint analysis modelling techniques. it is quite possible that variations in one attribute will impact the other attributes’ contribution to utility. Doing this allows us to model the design space with a continuous mathematical model that can yield gradient information. The model must also contain a term for the no-choice option. The above formulation only considers that main effects impact utility.

This information was then used in Equation (3) to create a predicted model of preference.86) within the intervals [−2. Unimodal examination We defined an example unimodal ‘goal’ preference function as f (x1 . A design point is simply one instance of the design defined by the combinations of the design variables.1. So. The surface could represent the preference that an individual has for the location of a clock on a wall with relation to the centre of the wall. the design space was discretised into a 5-by-5 grid of equally spaced design points (uniform design). Figure 2. design point (1. with the bimodal being more complex. Contour plot of model for unimodal comparison of preference tools.1. not on ratings. x2 ) = (0. Contour separation = 100. PREFMAP For the PREFMAP query. We used this model and the discretised decision space as a way to inform both the PREFMAP and discrete choice query tools.1. x2 ) = (1. 2] for x2 . we rated the 25 design points by scaling the function values at the discrete points to be integers between 1 and 9.5)(x2 + 3)(x2 + 2)]2 . The conjoint analysis MNL input data are based on discrete choice selections. .Journal of Engineering Design 633 modelling complexity. where 1 corresponds to least liked and 9 to most liked. x1 and x2 .86. 3. 1). 3. We are interested in observing how well the two preference modelling techniques represent this ‘goal’ preference model.1) would be the design associated with (x1 . 0. could be vertical and horizontal location. This discretisation was chosen because it is a logical way of discretising the space. The goal is to show that these two techniques yield different results and must be applied judiciously to design problems. x2 ) = [(x1 − 2.5)(x1 + 3)(x1 + 2)]2 + [(x2 − 2. as shown in Table 1. To provide preference information. (10) Downloaded by [University of Michigan] at 14:13 03 October 2011 This polynomial surface (see also Figure 2) has ridges and a single maximum (x1 . respectively. 2] for x1 and [−2. The design variables. This table is associated only with PREFMAP evaluations because it is a rating of the various designs. This scale is typical of actual PREFMAP surveys. This information was then regressed to the characteristic PREFMAP equation.

This shows us that the generality of this model to certain unimodal design situations may yield results that are desirable by the market of interest.C. but it should be noted that identifying which variables to examine is a difficult issue in a real design problem. However. Table 1.82 PREFMAP yielded an elliptical paraboloid centred at (x1 .59 b3 −0. Table 2. PREFMAP interpretation of proposed unimodal preference model.54 PREFMAP solution b-values.46). x2 x1 −2 −1 0 1 2 −2 1 2 4 5 2 −1 2 2 4 6 3 0 4 4 7 8 5 1 5 6 8 9 6 2 2 3 5 6 3 Downloaded by [University of Michigan] at 14:13 03 October 2011 Figure 3. It can be used to determine some valuable information about the market data. . it would be impossible for a paraboloid model to identify the ridges associated with Equation (10). Clearly. we do notice that these results provide a solution that identifies the appropriate quadrant of the design map for further investigation and it identifies the general trends well. b1 0. Its optimal point indicates that it would provide the producer with a design point that closely matches the actual optimal design point.54 b2 −0.46. Contour separation = 0. Kelly et al. 0.01 b5 6.2. Using the results of this PREFMAP study could allow a producer to properly identify an area of interest in the design space.68 (68% of preference variance is explained by the model). Values used to answer unimodal PREFMAP survey.634 J. The b-values associated with Equation (2) are shown in Table 2.59 b4 −0. Figure 3. x2 ) = (0. b0 0. The R 2 value for this model is 0.

40 × 16. and the resulting part-worths for each attribute level were analysed in an MNL model.1. This random double exponential distribution term. This use of conjoint analysis is not typical.2. These values are used in Equation (8). such that the option presented in the set with the greatest functional value was chosen from the set. with part-worth values given in Table 3. a Figure 4. and one was a no-choice option. The natural cubic splines fit to the data were used in order to reconcile the discrete nature of the conjoint analysis survey with the continuous nature of the design variables.1. Forty ‘participants’. was added to the value derived in Equation (10). The cubic splines fit to the data were specified to be natural cubic splines and the second derivate of the spline end conditions were set to zero. Each unique survey consisted of 16 questions. were used to answer the survey with their preferences defined by the function in Equation (10). MacDonald et al. We then fit natural cubic splines to this data to obtain a continuous and differentiable model of preference. In the PREFMAP model. 40 × 16 × 5. Contour separation = 0. To highlight one assumption of the MNL model. These data were then analysed with Sawtooth’s SMRT (Sawtooth’s Marketing Research Tool) module. Second. MNL model interpretation of proposed unimodal preference model without error distribution term. The premise is to determine spline-interpolated part-worths for the continuous set of design options. with integer ratings between 1 and 9. characterised by μ = 0 and σ = 200. 2005. each question was answered using Equation (10) along with an error term having a double exponential distribution. 3200 were needed in modelling. It was then possible to use these cubic splines to determine the probability of selection of any design within the design space. 640 choices from a set of.Journal of Engineering Design 635 3. each question had five options: four were designs selected from the discrete set of equi-spaced designs consisting of two attributes and five levels. causing some ‘improper’ choice selections. but is also not unprecedented (Michalek et al. 2007). First. So. This describes how each design option is preferred relative to every other option. modelled by computer agents. Conjoint analysis Downloaded by [University of Michigan] at 14:13 03 October 2011 The discrete choice analysis query was formed using Sawtooth Software’s Choice-Based Conjoint module (Orme 1999). for the MNL model. The MNL model derived from the first set of answers mentioned earlier is shown in Figure 4. we used Equation (10) to answer these questions in two different ways. the agents answered each question using Equation (10). only 25 data points were collected. . 2007). however. thus creating an interpretation of the full-factorial marketplace (MacDonald et al.

however. MNL model part-worths for data without error distribution term. The optimal point is located near the discrete choice (x1 . with part-worth values presented in Table 4. However.75 β12 −28. the contours are less polarised towards that point and more gradient 2 information is available to understand the preference space. Using ‘perfect’ preference data without an error term. x2 ) = (1. the MNL model would specify Figure 5. Figure 5 presents the MNL model results using the second set of answers.05 J. using survey information from a large number of respondents answering perfectly according to a specified preference model. The model described 2 by Table 3 has an Rcox = 0.14 MNL model part-worths for data with error distribution term. This model has an Rcox = 0.95.05. Kelly et al. 1. it should not be considered as directly equivalent to the R 2 that are reported in most regression situations. The ideal point coincides with the ideal design option available in the discrete set.31 β21 −1. is much more successful than that shown without the error distribution accounted for. the MNL model quickly identifies the most preferred option and defines the design space so that the most preferred option takes outstanding preference over all other options. β0 −36.50 β22 −28. a pseudo-R 2 value can be calculated.82.636 Table 3. MNL model interpretation of proposed unimodal preference model with error distribution term.53 Downloaded by [University of Michigan] at 14:13 03 October 2011 part-worth must also be included for the no-choice option. x2 ) = (1.50 β21 −45.50 β23 25.06) and appears relatively insensitive to the ridges of Equation (10). The pseudo-R 2 used in this 2 article is the Cox and Snell R 2 .C. Table 4. The model is polarised at an optimal value of (x1 . β0 . This model.82 β25 −0. Indeed.59 β12 −0.06. the error term is an important assumption of discrete choice analysis.31 β13 25.82 β13 0.32 β22 −1. Contour separation = 0.85 β14 1. β0 −17. At the extreme. Logistic regression models do not have an equivalent R 2 analysis to describe their goodnessof-fit. β11 −1.36 . β11 −45. which will be termed Rcox (Allison 1999). which included the random error term.52 β14 60. while not fully able to recreate the original model.04 β23 0.63 β24 59.90 β24 1.90 β25 −11. But.86 β15 −0.04 β15 −11. 1) available in the survey.

shown in Figure 6. A marketing survey will never have perfect data. and further.2).3)2 +(x2 +1. it will always contain human error and individual differences in preference. −1. 2] for x1 and [−2. x2 ) = (1. if the most preferred design was technically infeasible. 3. We would therefore hope that a model of preference would be capable of understanding such a bimodal response of users to such an incident. which might occur in a marketing survey. The ‘goal’ model. These two points have the same function value. one at (x1 . thus neither dominates the other. x2 ) = (−1. This is accounted for in the mathematical model with the inclusion of simulated error.2)2 ) + 1 e((x1 +1. Contour separation = 0. (11) This mathematical model has two points that represent locations of highest preference.2. . We used this model and the discretised decision space as a way to inform both the PREFMAP and discrete choice query tools. be able to accurately identify those two optimal locations. Mathematical model used to inform bimodal conjoint analysis and PREFMAP queries. The design space was again discretised into a 5-by-5 grid of equally spaced points. then identifying acceptable alternative designs would be very difficult. as shown in Table 5. x2 ) = 1 e((x1 −1)2 +(x2 −1. 2] for x2 . 1.2. Bimodal examination To further explore PREFMAP and MNL models. but would obscure the slopes and curvatures of the surrounding design space. This information was then used in Equation (3) to create a ‘predicted’ model of preference.1. PREFMAP For the PREFMAP query.3. Figure 6.1. Thus. we defined an example bimodal ‘goal’preference function as Downloaded by [University of Michigan] at 14:13 03 October 2011 f (x1 .Journal of Engineering Design 637 precisely the most preferred option. as was done in the unimodal case.2) and the other at (x1 . we ranked the 25 design points by scaling the function values at the discrete points to be integers between 1 and 9.2)2 ) . 3. was examined within the intervals [−2.

1 PREFMAP solution b-values for bimodal problem.2.14 b4 0. but only at the boundaries of the design space. PREFMAP interpretation of proposed bimodal preference model. Contour separation = 0. −2). and a point of second greatest preference at (x1 . b0 −0.638 J. Conjoint analysis The unimodal formulation of the MNL model only accounted for main effects and not interaction effects. b1 0 b2 −0. x2 ) = (−2. as shown here. Bounded scenarios allow this bimodality to be indicated. .67 b5 3. x2 ) = (2. The main effects are influenced by each attribute independent of each other attribute. This bimodal function examination illustrates the case of saddle points in PREFMAP models. 2).21 b3 −0.2. Table 6. Table 5. Values used to answer bimodal PREFMAP survey. The b-values associated with Equation 2 are shown in Table 6. Kelly et al.2. Figure 7. The R 2 value for this model was 0. it would be impossible for a saddle point model to yield a bimodal solution in an unbounded scenario.42. This interesting caveat of the PREFMAP model actually provides valuable insight into understanding the bimodal function. 3. x2 x1 −2 −1 0 1 2 −2 4 6 2 1 1 −1 5 8 3 1 1 0 2 3 2 4 3 1 1 1 3 9 5 2 1 1 2 4 3 Downloaded by [University of Michigan] at 14:13 03 October 2011 Figure 7. Clearly.75 PREFMAP yielded a saddle point model showing greatest preference at (x1 .C.

The location of these two locally optimal points are (x1 . This model has an Rcox = 0. as in the unimodal example.12 Table 8.20 β15 −0. Again.34 Downloaded by [University of Michigan] at 14:13 03 October 2011 β11op β12op β13op β14op β15op 0.29 0.03 −0. Also the random error term was included.74 −0.29 1. These points are in modest agreement with the actual optimal points on the real bimodal surface. Again.09 β13 −0.06 −1.14 0. This model also provides a great deal of gradient data that can be used in instances where the optimal points are infeasible in a larger design problem. β0 −17. Using cubic splines to fit this data. MNL model part-worths for proposed bimodal model.12 β21 −0. causing ‘improper’ choice selections. Contour separation = 0. The main effects coefficients are presented in Table 7 and the interaction effects are shown in Table 8. βmn11 βmn12 0. the resulting model was developed (see Figure 8). 1. −1.14 βmn15 −0.36 βmn14 −0. MNL model interpretation of proposed bimodal preference model with interaction terms. Due to the nature of the bimodal design space in our example. main effects.20 −0.24 βmn13 0. and was added to the value derived in Equation (11). which link one attribute to another.13 0. . β11 −0.19 639 β25 −0.41 0. as we would expect based upon the example function.10 β22 0. In this figure.03 0.07 β14 0.02.33 Equation (9) accounts for interaction terms. x2 ) = 2 (−1. x2 ) = (1.28 1.36 −0.Journal of Engineering Design Table 7.16 β24 0.02. Figure 8. the distribution term was characterised by μ = 0 and σ = 200.50. interaction terms.16) and (x1 .38 MNL model part-worths for proposed bimodal model.34 −0. Note that a part-worth value is again included for the no-choice option.08). it is appropriate to use this formulation of the MNL model because interaction effects should be significant. we see that the model predicts two points of local optimality.21 0.18 −0.07 −0.20 β23 −0. the survey questions were answered by using the goal model to inform which discrete choice should be selected.02 −0.10 β12 0.08.32 −1.33 −0.29 −0.

the PREFMAP model generates a saddle point surface. conjoint analysis was used to examine the design trade-offs between shape preference and engineering functionality in the design of cola bottles. the variables were continuous. 4. and subjected to prescribed end conditions. in France. However. we discretised the design space with five possible values for R2 and R4. 4. The other three points were fixed parameters during optimisation. and the data were analysed using both commercial software (Sawtooth. both conjoint analysis and PREFMAP were used to determine the same design trade-offs in the case of bottled water designs. conjoint analysis MNL) and newly developed software (Matlab program for PREFMAP analysis). In the conjoint analysis. Kelly et al. The bottle shape used for this study was defined by a spline fit through five points.25 mm. On the surface. Participants for both studies were university-aged students from the Ecole Centrale de Nantes. points R2 and R4 in Figure 9. In the case of bimodality. In the engineering analysis. It is worth recalling that the PREFMAP model required 25 integer inputs between 1 and 9. Two of the five points were considered variable. such a statement seems true. Values for R2 and R4 were constrained between 25 and 50 mm. Downloaded by [University of Michigan] at 14:13 03 October 2011 4. thus creating an efficient survey design. Much effort is put forth in creating unique and appealing bottle designs (Grimm 2000.640 J. this statement neglects an important issue: data collection. The data were analysed using Sawtooth Software to obtain part-worths for each variable and level of the two design variables. yielding 25 possible designs (uniform design). This highlights an inherent limitation of the PREFMAP model. This second study was done to reduce the impact that currently dominant shapes in the cola bottle sector have on the results. The design space was discretised into a 5-by-5 grid of equally spaced points. Cola bottle case study Branding through shape is important to the beverage industry. it is tempting to conclude that using discrete choice analysis to develop an MNL model will provide us with a better model of the design space. Linking preference and engineering Now that we have described some techniques to model preference data. .1. A Matlab program was developed to collect data from the participants. Each respondent answered a survey consisting of 16 questions. while the MNL model generates a bimodal surface. However. Lamons 2001.1. we will use them to explore shape preference for the design of a particular product. Each individual received a unique survey. Point R1 was set for a perfectly vertical end condition. In the first study. In general. and provided sufficient shape differentiation. spaced at an increment of 6. while R5 was set with an end condition to create an angle of 20 with the horizontal. thus creating a design space with 25 different designs. as shown in Figure 10.1. Two studies will be described that attempt to understand shape preference as it relates to bottle design.C. In the second study. Vanderbilt 2001). Preference assessment The conjoint analysis survey was administered to 39 college-age individuals from the Ecole Centrale de Nantes. France. The unimodal and bimodal results illustrate the capacity of the PREFMAP model and the MNL model to reproduce a predefined preference function using each model’s query tool. whereas the MNL model required 640 choices to be made from a set of 3200. the participants from the first study are not the same as those in the second. and each question offered the respondent four shapes and the no-choice option to choose from.

The internal gauge pressure for this experiment was chosen to be 300 kPa (60 psi). I would choose none of these Figure 10. The .2.Journal of Engineering Design 641 R1 R2 Downloaded by [University of Michigan] at 14:13 03 October 2011 R3 R4 R5 Figure 9. Engineering model From an engineering viewpoint. 4. Screenshot of survey tool. Parametrised bottle shape.1. we desired the bottle shape that used the least amount of material to hold the desired amount of fluid and resisted the internal pressure without plastic deformation.

In this problem. A Pareto optimal point represents a single design in R2 and R4 that is non-dominated. f1 is the shape preference function.3 Youngs modulus 1. Its material properties are shown in Table 9. R4). R4) − σmax ≤ 0 (12) Here w1 is an objective weighting. f (R2. f2 is the material volume calculation.40.25 GPa Downloaded by [University of Michigan] at 14:13 03 October 2011 analysis model was built using the finite element package ANSYS (Moaveni 2003). the message conveyed by the shape). The cap section was given a double wall thickness to prevent a high level of stress in that area (McEvoy et al. is presented in Figure 11 along with the optimal shape. the wall thickness here was also increased slightly to accommodate increased stress. R4) = (32. to solve the optimisation. The maximum von Mises stress within the bottle was calculated to ensure that the bottle would avoid exceeding the material’s tensile strength. available through the Matlab optimisation function. Interaction terms were considered significant according to the ‘2 log likelihood test’ and included in the model (Orme 1999). An axisymmetric solid model was created with a spline shape as previously described. g1 (R2. g1 is the maximum von Mises stress in the bottle and R2 and R4 are the shape variables. Combining the data The preference model. Kelly et al. 1998). The 2 optimal design was (R2. The convex hull of the Pareto frontier set was calculated by varying w1 between 0 and 1.01. The shape is similar to that of cola and other soda bottles in the market. so it appears flat to the user in a side view and is therefore consistent with the figures shown to respondents in the conjoint survey.2. This spline shape was given a uniform wall thickness treated as a design variable. the result suggests that participants may prefer this particular shape for a cola bottle specifically because they have . In this work. it is axisymmetric.642 Table 9. R4) + (1 − w1 )f2 (R2. along with splines fit to a discrete set of potential bottle designs. The Pareto frontier identifies the set of all Pareto optimal points in the design space. 4. to generate this contour plot. While this bottom section of the bottle is not flat. from the stand point of semantics (i. R4) = w1 f1 (R2.6). The values of the main effect and interaction effect part-worths are in Tables 10 and 11. J.t. we use a sequential quadratic programming approach. we solved Equation (12) using different values of w1 between 0 and 1 at intervals of 0. Tensile strength 25 MPa Poissons ratio 0. A linear multi-objective formulation converted to a scalar substitution was used to find the optimal designs using single objective optimisation techniques. The bottle’s bottom section was designed according to an available patent since this is typically the critically stressed location of bottle designs (Rashid 2001).e. the optimal design point may also change due to different weightings for the two competing objectives.3.C. obtained through survey data. wall thickness was fixed at 1 mm to simplify the calculation and to make the trade-offs between the two objective functions clearer. The results of the conjoint study suggest that individuals gravitate towards a shape that they are familiar with. min s. To develop the convex hull of the Pareto frontier. fmincon. Cola bottles are typically blow moulded from polyethylene (PET). As w1 changes. Material properties of PET cola bottle. PET was selected in this design problem. therefore. This model has an Rcox = 0. An MNL model that included interaction effects was used. In fact.1. 31.

The optimal bottle design has (R2.43 −0.47 β13 0.44 β14 −0.54 βmn15 −1.51 β23 0. and a wall thickness of 0.19 −0. This is shown in Figure 12.14 β25 −0. the values of R2 and R4 will be minimised to further reduce the amount of material used to make the bottle.04 β15 −0.84 0. One may argue that .06 −0.43 −0.74 βmn13 0.15 0. Table 10. Further. The maximum von Mises stress for the bottles occurred in roughly the same place on the bottle’s bottom. Contour separation = 0.11 0. and most preferred shape. the constant 1 mm wall thickness assumption made in the broader optimisation is justified.73 β21 0. MNL model describing preference for cola bottle shape. β11 −0.34 −0. main effects. subject to the stress constraint.82 βmn14 −0. R4) = (25. From the engineering perspective. 25).41 0. More importantly. the wall thickness should be as small as possible to reduce material volume.91 β11op β12op β13op β14op β15op 0.02 0.24 0. MNL model part-worths for preference survey.07 −0.Journal of Engineering Design 643 Downloaded by [University of Michigan] at 14:13 03 October 2011 Figure 11.86 0.48 −0. β0 −0. Therefore. interaction effects.23 −0.32 −0. no bottle design will fail with a wall thickness of 1 mm. which shows monotonic decrease towards (R2. These two objectives are shown to compete.01. R4) = (25. The convex hull of the Pareto solutions are shown in Figure 13 and are also plotted on the individual objective surfaces in Figure 14 to visualise the trade-off between maximising preference and minimising material volume.76 Table 11.78 −0.11 β22 0.11 encountered it as a cola bottle shape so often previously: This shape means cola bottle to these respondents.28 β24 −0. Note that this figure is presented with a wall thickness of 1 mm to show the general trend.27 0. βmn11 βmn12 0.98 mm.66 −0. 25).15 β12 0.82 MNL model part-worths for preference survey.44 0.

constraints restricting the interior volume of acceptable bottle designs may change the optimal design. Namely. the simplified model exposes the asserted quantification of design trade-offs between shape preference and engineering functionality. Figure 13. Kelly et al. Monotonic surface representing bottle weight.C. More refined engineering models are certainly possible.644 J. Contour separation = 5e − 6. however. Downloaded by [University of Michigan] at 14:13 03 October 2011 Figure 12. Bottled water case study In the examination of cola bottles. 4.2. we encountered issues that confounded our study. users appeared to have a distinct notion of what cola bottles should look like based upon market . Pareto frontier of solutions in the cola bottle design problem. This is true.

Respondents then rated each of these bottle design on a scale from 1 to 9 (least to worst. its centre of gravity is low.16 β21 −0. i. once again. this shape may connotate a level of functionality. The discrete choice data were analysed. This may indicate that. We also queried users with the PREFMAP evaluation tool. beyond a particular visual attraction.e.62. 4.20 β12 0.2. This result helped us identify that the MNL model can be useful in determining shape expectations of consumers. the ‘waist’ of the bottle may be taken to indicate the proper location to place one’s hand. β11 −0. The ‘weight’ of the top versus the bottom of the bottle may assure users that the bottle will not tip over easily. We asked them to express their preference for the shapes of bottled water designs. However. saturation with cola bottles whose shapes are similar to that of Coca Cola™ bottles. This model has an Rcox = 0. at four levels). in this instance.28 MNL model part-worths for bottle water preference survey.17 β22 0. That is to say. The results of this study. using an MNL model with interaction effects. For instance.1.85 .54 β14 0. The design space was partitioned in the same way as in the cola bottle study using a uniform design with 25 designs. users were quite adept at identifying a preference for a known shape.74 β24 −0. The design space was discretised into a uniform design with 16 separate designs (two attributes. participants had an affinity for the bottle shape often associated with Coca Cola™. In an attempt to limit a preconceived notion of an appropriate shape.35 β15 −1. Pareto optimal solutions of cola bottle design plotted on preference (a) and engineering (b) models. indicate that.61 β25 −0. as before. we again conducted a bottle study. 40 college age students for Ecole Centrale de Nantes were queried about their preferred shape for a water bottle using both a discrete choice survey and PREFMAP survey. This was captured by the MNL model and gives us some confidence that such a model is valuable for understanding shape preference in other contextual instances. Preference assessment For the water bottle study.Journal of Engineering Design 645 (a) (b) Downloaded by [University of Michigan] at 14:13 03 October 2011 Figure 14.47 β13 0. respectively) based on how Table 12. β0 −0.90 β23 0. Tables 12 and 13 show the associated beta values and Figure 15 shows 2 how the preference model varies with R2 and R4. we provided users with a different context for their evaluations. We believed that such an alteration of the experiment would help identify a different preference space that would not be so predictable. given a specific context. main effects.

20 −0.67 0. and most preferred shape.90 βmn15 −0.42 −1. PREFMAP model describing preference for bottled water shape. Figure 16.68 βmn13 0.53 0. .14 0.15 −0. and most preferred shape.02 −0.55 0. interaction effects.19 βmn12 1.C.65 0.07 −0. Contour separation = 0.87 1.81 −0. Kelly et al.15 0.26 −1. Contour separation = 1.31 1.646 J.02. MNL model describing preference for bottled water shape. βmn11 β11op β12op β13op β14op β15op 1.62 −1.04 −1.56 −0.28 1.59 Downloaded by [University of Michigan] at 14:13 03 October 2011 Figure 15. MNL model part-worths for bottled water preference survey. Table 13.63 βmn14 −2.36 0.

015 b2 −0. relate to each other in regards to this bottle design.2. What’s more. and engineering space with Pareto optimal design points shown on them to understand how varying R2 and R4 will affect the design. Therefore. and engineering space. indicating that it will have a colocation with the engineering optimal solution. This PREFMAP information was averaged over the entire population of respondents. 25). We notice that there is not a colocation of the optimally preferred and engineering optimal design and we can tell by looking at the preference space. that there is a nearly 1:1 ratio of preference between R2 and R4.08 647 Downloaded by [University of Michigan] at 14:13 03 October 2011 well they liked them for a bottled water design. R4) = (25. It is interesting to note that both the MNL and PREFMAP models suggest a rotated preference space. b1 −0. in both cases.0002 PREFMAP solution b-values for bottled water study. we can see how the two models. Combining the data The engineering analysis for this bottle is exactly the same as the cola bottle design. this trend indicates.0049 b3 −0. Figure 17. Pareto frontier of solutions in the bottled water design problem using MNL for preference. Thus. Thus. The PREFMAP space of shape preference is shown in Figure 16 and the b-values are shown in Table 14. In Figure 18. The R 2 value for this model was 0. only an optimal point for all weightings. 4. In the PREFMAP scenario. Notice that the PREFMAP solution is at the boundary.0047 b4 0. .008 b5 8. b0 0. thus an inherent interaction between the R2 and R4 variables is found. Using this data. there is no Pareto frontier developed from the combined optimisation. (R2. PREFMAP and MNL. people appear to prefer R2 and R4 that are of equal proportions.72.Journal of Engineering Design Table 14. we present the MNL determined preference space. we are not surprised to notice that the Pareto curve developed by combining the MNL shape preference results and the engineering objective provides information that is quite similar to that of the cola bottle (Figure 17). the optimal engineering design is colocated with the preferred shape design. figures that distinctly different shapes will be produced by the two differing objectives.2.

models. Designers and engineers currently make informed decisions regarding product form. and experiments to elicit preference must be conducted carefully. This information can come from marketing demands. several concepts are devised and evaluated until one design concept is agreed upon. or a suggestion of how to use the object. Doing so allowed a relatively easy generation and interpretation of results. then it is likely that different results would be found. the definition of the survey must be made by applying design of experiments techniques. In the presented study. this information can be mathematically incorporated into the design optimisation process during the final stages of design.3. Kelly et al. In this case. Information regarding shape preference could be effectively utilised during this detailed design phase. in order to limit the size of the survey and also to improve the confidence of the estimates. (a) (b) Downloaded by [University of Michigan] at 14:13 03 October 2011 Figure 18. the shape may actually provide an affordance for the user. This study can only speak to the impact of these two variables on shape preference. using MNL. We also note that this study limited users to a selection between two variables. Two factorial designs must be proposed: one to define the designs (configurations). Finalising the form of a product occurs in the detailed design phase of the design process.648 J. engineering specifications and designer interpretation of user wants. and many papers in discrete choice analysis deal with the definition of the second factorial design (Huber and Zwerina 1996).C. we can use that information along with an engineering model to learn about the trade-offs between the goals of minimising material used. Pareto optimal solutions of bottle water design. Meaningful quantification of a product’s shape preference is possible using standard methods from psychology and marketing. and one to define which configurations will be integrated in the choice-set proposed to each user. subject to the constraints imposed by the bottle’s architecture as explained previously. Discussion This study indicates that users may have a particular preference for the shape of bottles based on an association with historically well-marketed bottle shapes. this study does indicate that we can discover distinct relationships between the variables R2 and R4. Further. The methods have limitations. If the bottle were parametrised in a different manner. we used two variables (or attributes) to define the variations in a particular product offering. Further. plotted on preference (a) and engineering (b) 4. Prior to that. To properly create the proposed surveys. However. A more complex design model may describe the product with more variables. and how they are related to shape preference by using both PREFMAP and discrete choice analysis. the amount of data needed for statistical validity of the MNL model would increase significantly. and maximising shape preference. Several tens of variables can be considered in discrete . R2 and R4. The first factorial design must be generally balanced and orthogonal (or D-optimal). With a quantitative model of shape preference.

Coxon. Intuition of industrial designers need not be the only informative evidence to support a particular stylistic desire.. Conjoint analysis in marketing: new developments with implications for research and practice.P. London: Heinemann.Y. Chang. 64 (8). some important product design considerations remain unexamined in this work.. . and Srinivasan.M. A primary one is devising appropriate ways to identify the right attributes of a design and their associated design variables. 1972. Ensuring that the proper design variables are identified will increase the effectiveness of determining shape preference.. et al. Other methods may be also studied for improved efficiency and accuracy. is. International Journal of Human–Computer Studies. organise it into a model and incorporate it within a physics-based design optimisation formalism. concerning vehicle headlight form preference. 937–968. V. In the same way. We examined PREFMAP and MNL as the preference modelling techniques.T. 1990. The experimental design is optimised according to the form of the model.D.. J. To reduce the number of coefficients. A quantification of shape preference allows it to be included along with engineering attributes to explore products that are optimal in a multidisciplinary design sense.. for example. Bauerly. Germany: Breitkopf & Härtel.E. Balancing these trade-offs is still a decision that the designer must ultimately make. 2006. Acknowledgements The authors are grateful to the US National Science Foundation Graduate Student Research Fellowship. An example with 14 variables. The survey had in this case 64 designs. How to use PREFMAP and PREFMAP-2: programs which relate preference data to multidimensional scaling solutions. 1982. 5. Fechner. Logistic regression using SAS® : theory and application. Journal of Marketing. Leipzig. J. involved in a choice set of 72 questions. Both techniques are capable of providing insights into shape preference through their mathematical models.J. form and function have distinct trade-offs that meaningfully affect each other. 1995. Green. Vorschule der Aesthetik. Unpublished manuscript. G. Computational modeling and experimental investigation of effects of compositional elements on interface and design aesthetics. 1999. 24 (8).. Conclusions We proposed and implemented a procedure to collect information regarding shape preference. our findings showed that there is a dislocation between what is most appealing to users and what is most sound technically. 2007).. A. based on which the models will be built. In the study presented. 1897. Murray Hill. However. With the PREFMAP model. proposed in (Swamy et al. Perception. and Liu. Green. a classical way is to carry out a principal component analysis on the design variables and to regress preference on the two or three first principal components (McEwan 1996). the increase in the number of variables leads to an increase in the number of coefficients to estimate in the model. NJ: Bell Telephone Labs. the D-optimality being generally used (Mitchell 1974). asked to 18 users.D. M. 54 (4). References Allison. and Carroll. C. Cary. P. presumably of quality higher than without the trade-offs quantification.Journal of Engineering Design 649 Downloaded by [University of Michigan] at 14:13 03 October 2011 choice analysis. 3–19. P.. NC: SAS Publishing. The user’s guide to multidimensional scaling. In the example of bottle design. All that glitters: a review of psychological research on the aesthetics of the golden section. and strictly meeting a particular shape preference can have a detrimental effect on product performance. specifically exploring tradeoffs between form and function. partial least-square regression (PLS) can be used with several tens of variables. The respondent fatigue is of course the main limitation to the size of the problem considered. 670–682. which in turn can be incorporated within a mathematical optimisation framework.

Journal of Mechanical Design. Ergonomics. Louviere. American Demographics. Inc. 17 (4).M. 609–626. MacDonald. 2001. 20–22 June. 2007. 2004.. P..com [Accessed 12 November 2008]. 15 (1). 2000. Marketing Science. A. 2000. Holden. Journal of Marketing Research. D.M. D. Stated choice methods: analysis and applications. Michalek. T. Park. Lidwell. 71–102. Norman.M. E. International Journal of Industrial Ergonomics. MA: Rockport Publishers.C. 33 (3). NV. Using conjoint analysis to elicit preferences for health care. J. Nagamachi.F. Thesis (PhD). A. Finite element analysis: theory and application with ANSYS. 2003.. Boston: Newnes-Butterworths. S. 217–233. Kansei engineering: a new ergonomic consumer-oriented technology for product development. Journal of Engineering Design. Gloucester. The construction of preferences for crux and sentinel product attributes. and Butler... University of Michigan. 2005. 2003a. 56–73. 1293–1305. J. K. Data Handling in Science and Technology. CBC user manual. 1273–1292..J.. A logit model of brand choice calibrated on scanner data. Vol... 5–7 November.sawtoothsoftware. Upper Saddle River.. H.. and Papalambros. P.. Sawtooth Software. version 2. UK: Cambridge University Press.J. 2006. 22 (1). P.382. M... Michalek. A short course in industrial design. A. 131 (12).. and Wind. and Boatwright. Plastic container having base with annular wall and method of making the same.F. Las Vegas. Rashid. New York: Cambridge University Press.. US Patent 6. Austria. Petiot...P. The nature and aesthetics of design.. Computer construction of “D-optimal” first-order designs. Armstrong. Huber.. Drink me. Hensher. M. and Little. 307–317. M. Orme. 2009. 2005. 2001. E. Grimm. 2003b. J. Tjalve. Multiagent shape grammar implementation: automatically generating form concepts according to a preference function. D. 19 (2). 061001-1–061001-10. 197–206. Liu. In: Proceedings of Virtual Concept – 2003.. Principles of optimal design: modeling and computation. 82–101.. 1974. M. and Zwerina. J.. Product design: a vectors field-based approach for preference modelling. 46 (13–14). Coca Cola. Preference mapping for product optimization. W. Vienna. A learning from the past a foundation for the future (Special publication of papers presented at the CAAD futures 2005 conference).. WA. and Wilde. S. France: ESTIA.. Feinberg. Orsborn.R.. Simulation of the stretch blow molding process of PET bottles. R. Advances in Polymer Technology. 17 (3).Y. J. P.A. Absolut fortune: one liquor’s success story.D. Park. 79–80. The design of everyday things. and Papalambros. Vanderbilt. Demand modeling using discrete choice analysis – Part 1 and Part 2. 1530–1533. 121007-1–121007-10. Petiot. S. Martens and A. Biarritz. In: Proceedings of the ASME 2007 international design engineering technical conferences & computers and information in engineering conference. 1996. J. J. 22 (2). Pye. 45 (3). 1981. F. and Papalambros. A marketing audit using a conceptual model of consumer behavior: application and evaluation. eds... P. Green. A quantification of proportionality aesthetics in morphological design. and Cagan. 2001. The Journal of Marketing. and Crawford. K. 339–352. Liu. 20 (6). P. United States.. International Design.. C. 2002. 2000. Journal of engineering design. 211–220. Technometrics. International Journal of Industrial Ergonomics. 42–62. 16. Ryan. Y. B.. 1996.J. S.. Measurement of headlight form preference using choice based conjoint analysis. Y. 46 (13-14).A. 1998. Quantifying aesthetic form preference in a utility function. New York: Basic Books. and Grognet. Krieger..S.. Ann Arbor. S. Kansei engineering and comfort. 1995. J.. 12 (1). and Chablat. Swamy. 44–49. Hermes: a computational tool for proportionality studies in design. Available from: http://www.. Journal of Engineering Design. Marketing News. Linking marketing and engineering product design decisions via analytical target cascading*. Interfaces. 1978.. Economou. France. Subjective evaluation of forms in an immersive environment. B.. 6. Biarritz. Vienna: IRIS-ISIS Publications. Journal of Product Innovation Management. Ergonomics. 2009. J. 1999. 2000. T. Guadagni. and Cagan.. P. J. Michalek. British Medical Journal. Brown.176... J. The importance of utility balance in efficient choice designs. 203–238. J. Taubman College of Architecture and Urban Planning. Journal of Mechanical Design. Nagamachi. Universal principles of design. J. Downloaded by [University of Michigan] at 14:13 03 October 2011 .D. S. Papalambros. A. 1997. New York: American Society of Mechanical Engineers.. 8. NJ: Prentice Hall. 31 (3). J... 2004. 320 (7248). 2003. 48 (3). 0 [online]. McEvoy.. 3–11. J. 1979. 99–108. Orsborn. New York: Herbert. 2001. 2007. A. H. Bottled up. Sequim. and Hauser.J. Tybout.. 2 (3). R.. 62–63.Y. Thirty years of conjoint analysis: reflections and prospects.. MacDonald.. S. In: B. Cagan.. Aesthetic intelligence: optimizing user-centred design. 35 (4). 2001. D.. and Farrar.G. The aesthetic and the ethic dimensions of human factors and design.F. Orsborn. 2nd ed.C. Cambridge. Gonzalez. Kelly et al. 2003. D. Y. 37–45. Moaveni. Engineering aesthetics and aesthetic ergonomics: theoretical foundations and a dual-process research methodology.650 J. J.. Mitchell. 131 (6)... Lamons. McEwan. 2003. and Swait.. 1983.

Sign up to vote on this title
UsefulNot useful