From Configuration to Design: Capturing the Intent of User Designers is the title of a paper I recently presented at MCPC 2009 in Helsinki. It details a user trial conducted as part of my PhD research, which sought to understand the extent to which non-professional user-designers are able to engage in design exploration and to communicate design intent. The paper itself, together with the presentation given at the conference, can be downloaded from the Papers and Presentations page of this site. Much of the background argumentation to the study has been made in previous posts, therefore what follows is an edited version of the paper, focusing on the design, conduct and conclusions of the user trial. This first post deals with the design and conduct of the study, a follow-up post will concentrate on the results and conclusions.

Design of the Study

The focus of the trial was the design of a USB memory stick. This was chosen as a relatively simple product whose functionality was easily recognised by those who took part in the study. The trial was intended to investigate two main research questions:
What is the best method for consumers to conduct design exploration?
How well are consumers able to communicate design intent?

It built on the observations of a number of researchers with regard to the way designers and architects use drawing as a way to generate and evaluate design solutions, but sought to place such observations more specifically within a mass customisation scenario. It also sought to understand the practical difficulties of expecting non-designers to use drawing in the same way that trained designers do. The intended outcome was to better understand what future tools will best enable consumer-design, which will form a major part of my future PhD research.

I should make clear from the beginning that within the user trial, neither modelling exercise (in Part I or Part II of the trial) was intended to test or replicate a co-design exercise. Within design research, co-design can be described as a subset of user-centred design (also called participatory design), in which the user takes part in the actual design of the object in question as part of a design team. This contrasts to user-centred design itself, in which the user contributes experience or opinions, but the designer carries out the design task. In both user-centred design and co-design however, it is the designer who is perceived to hold ‘expert’ knowledge and who has the ultimate power of decision, and the user is relegated to the position of contributor. In contrast, this trial sought to investigate a situation in which the participant had the ultimate power of decision over his or her design. In all situations where the participant worked with the CAD operator (described below), care was taken to ensure the CAD operator did not offer opinions as to the value of any design or decision. Advice was given only on the capability of the CAD software to achieve a desired outcome, rather than the value of that outcome. Thus the role of the CAD operator was that of a facilitator between the participant as designer and the requirement to create a 3D CAD model.

Two Methods of Design Exploration

The study was divided into two parts. In part I, the participants were first required to undertake an unobserved drawing exercise, followed by a design modelling exercise with the assistance of a trained CAD software operator. In part II, they were required to choose one of six pre-existing designs which was then modified with the assistance of the CAD operator. Participants were placed in one of two groups; group one conducted part I of the exercise followed by part II, whilst group two conducted the exercise in reverse order.

The first method of design exploration, addressed in part I of the study, could be classified as unconstrained concepting. Participants were free to explore issues of functionality and aesthetics with no constraint other than that the design should be bigger than a minimum volume (the minimum size required for the electronics to fit inside). This meant that the first method was close in scope to the design process of a trained industrial designer. The second method of design exploration, addressed in part II of the study, can be classified as constrained concepting. Participants were able only to modify a pre-existing design within the constraints allowed by the CAD model, the idea being that the task was closer in scope to a MC toolkit experience.

Part I – Sketching Exercise

Participants were briefed as to the task and requirements of the exercise, but conducted the exercise unobserved so that they were in a more natural environment and worked in a less time-constrained manner. Participants reported spending one-and-a-half hours on average on the task, though most reported thinking about the task over a period of days before beginning. Participants were required to complete the sketching exercise within one week of having been briefed.

When being briefed, participants were told to create drawings on an A4 marker pad supplied to them. A number of images were supplied of a ‘minimum volume’ USB memory stick (see the section headed “Concept of the Safe Model” below), and participants were shown how to use these images as underlays over which their own designs could be drawn. Participants were instructed only to use the drawings pads and pens, pencils, markers etc, and specifically not to create designs using a computer. It was also made clear that the purpose of the study was not to judge drawing skill.

Sketch design by one participant in the trial

Participants were instructed to design the body and the cap of the memory stick, and to imagine they were designing a personal product, i.e. not to consider the needs of other consumers. It was stated that participants should not copy an existing design, and that the more personal the design (in terms of either function or aesthetic) the more useful it would be to the research. Participants were also supplied with a sealed envelope of existing USB memory stick images, and told they could open the envelope at any time during the exercise.

Sketch design by a second participant in the trial

Part I – Modelling Exercise

Participants were required to return their drawings by mail such that no verbal explanation could be given about the design. This made sure that their ability to communicate through sketching only was tested. The drawings were used as the basis for construction of a 3D CAD model by a CAD software operator experienced in making industrial design models; the CAD operator was therefore required to ‘read’ the drawings, ‘interpret’ the participant’s design intent, and develop the 2D drawings into a 3D model

Finally, participants were told that whilst they could work on any number of designs, at the end of the exercise they should have one final, favourite design. Participants were instructed to return all drawings, even those of discarded ideas. It was also emphasised that when submitting the final design, participants should consider how well it could be understood by someone looking only at their drawings.

In some instances what was drawn was not physically realisable in 3D. This is sometimes referred to as a ‘failure of reinterpretation’ – where the person making a drawing has failed to reinterpret and understand the logical implication of their own drawing. The image below shows such a failure of reinterpretation: the indentation in the top surface is shown as breaking the side surface in one view but not in another. In these cases the 3D modeller’s task became one of intuitively judging the participant’s intent, rather than trying to faithfully reproduce was what drawn.

Sketch design showing a failure of reinterpretation

3D CAD model from the sketch above

3D models were built using Solidworks 2007 (Service Pack 2.0) CAD software. Solidworks is a hybrid (it allows the use of both solid and surface modelling techniques) parametric CAD modeller, in which features are primarily created from constrained, dimensioned sketches. One of the skills of the CAD operator lies in understanding how to constrain sketches such that dimensions can be altered and the model will update. Sketches which are not appropriately constrained will cause the update to fail, which can then result in significant time spent “debugging” the problems. The software and version were determined specifically by the use of Genoform, an automatic iterative design program that formed part of the study (see the section headed Genoform below).

Participants were invited back to conduct the modelling exercise approximately one week after having submitted their design. They were asked to review the model and to comment specifically on how well it captured their design intent (i.e. did it look the way they expected). Attention was drawn to specific aspects of the model, particularly where the CAD operator had interpreted a difficult-to-understand drawing or feature. Participants were then asked whether there was anything they would wish to change about the model either to improve the design or to correct mistakes in the interpretation of their drawings.

When the CAD model had been modified to a state the participant felt reflected their aspirations for the design, the Genoform software program was used to generate alternative design options. Initially ten options were generated but participants were free to generate more if they wished. Options which were liked or perceived as interesting were imported back into Solidworks; these reimported options were then compared to the originator model. In a majority of cases the participant requested changes to the originator model, based on ideas stimulated by the Genoform options, however in no instance was a Genoform option chosen as a ‘most favoured’ design.

Part II – Modelling Exercise

In this part of the study, participants were shown six pre-designed CAD models. The reasoning behind each design, e.g. why it was a certain size or contained certain features, was explained; the extent to which it might be modified was also made clear. Participants were then asked to choose one of the six models as the basis for the rest of the exercise.

6 pre-designed models presented to the trial participants

Having chosen a model, participants were asked which aspects of its design they wished to change. Where it was possible to modify the model by changing a feature’s dimensions or parameters this change was accepted. However any request which involved adding new features was not accepted. For example: with the ‘grip’ feature on model 1, the number of grip details could be modified, however a similar grip feature could not be transferred to any other model. In such a way participants were deliberately constrained in their ability to influence a given design. The CAD model was again modified by the CAD operator in front of the participant according to his/her instructions.

When the chosen model had been modified to reflect the participant’s intent, the Genoform program was again used to generate alternative designs. Ten options were generated initially but participants were able to request more options. Again, those options felt to be interesting were imported back into Solidworks and compared to the participant’s own modified model. In this exercise Genoform was less able to suggest new ideas or directions, and a majority of participants preferred their own modified model to the Genoform derived options. In those who found the Genoform options useful, none chose a Genoform option as the ‘most favoured’ design, preferring instead to modify their own model further.

Concept of the Safe Model

Most industrial designers understand the safe model (sometimes also called a ‘keep away’ model) concept. It’s used to understand and visualise the minimum possible size of a product, whilst taking account of internal mechanisms and electronics, thickness of materials, tolerances, etc. A safe model of an MP3 player for example, would be created by ‘expanding’ the dimensions of the internal electronics by an amount equal to the thickness of the materials used in the outer casing, plus the distance required between the electronic components and the inside of the casing. It can also incorporate considerations of safety, ergonomics, marketing, etc; thus the safe model for a family car would be affected by the need for crash crumple zones, headroom in the passenger compartment and size of boot. A safe model does not dictate the final design of the product (though it does influence the final design), rather it indicates the absolute minimum volume a product can be when all other requirements are met.

Safe model of a touch-screen mobile phone

Example of a safe model highlighting a potential problem. In such a case the designer would need to redesign the product in this area, or ascertain whether the internal components could be moved

The concept of the safe model was used in two ways in the user trials. Firstly, having calculated a safe model for the USB memory stick, images of this safe model were given to participants during briefing of the drawing exercise. Participants were shown how to use these images as underlays which acted as guides during design. Provided the participant’s drawings were not smaller than the images of the safe model, their design would be realistically manufacturable. The safe model was also used in the two modelling exercises. By modelling the safe model inside Solidworks, any design could be superimposed to check if it satisfied the minimum volume requirements (Figure 4). Furthermore, when setting up the parameters for the operation of the Genoform software, the safe model placed lower limits on the extent to which Genoform could modify the design.

A participant’s ‘final’ design, with the safe model also shown

Genoform

Genoform is an iterative design exploration tool which operates as a plug-in module to Solidworks. It is produced by Genometri, a design technology company which develops specialised software, which was created as a spin-out company from the National University of Singapore. Genoform works by varying the dimensions of a Solidworks sketch; the designer can assign which sketches Genoform can manipulate, which dimensions within those sketches, and the degree to which the dimension can be varied. Thus it is possible (for example) to instruct Genoform to vary a dimension of 10mm by plus or minus 25% (i.e. a range of 7.5mm – 12.5mm). It is also possible to set maximum or minimum values, thus the designer may decide that the 10mm dimension can never be reduced, but can be increased by 45% (i.e. a range of 10mm – 14.5mm). In this way Genoform will run through the structure of a Solidworks CAD model, altering dimensions by a random factor within limits decided by the designer, and creating new iterations of the original CAD model. Genoform will create between one and one thousand variants, as the designer decides. The image below shows variants of a single design created by Genoform. My earlier post goes into more detail about how Genoform works, and you can download a trial copy here.

Design iterations created by Genoform. The original model is in the top left

Trial Participants

Ten participants were recruited from within the postgraduate student body of Loughborough University in the age ranges as shown in the table below. Participants were required to be computer literate as defined by daily engagement with five out of seven of the following activities: web browsing, e-mail, social networking, chat, VOIP (e.g. Skype), Microsoft Office software, other software. Participants were also required to self identify as “being interested in design and new technology”. As such, the profile of participants fitted with the findings of e.g. Bauer et al. (2007) and Füller and Bartl (2007) regarding the types of consumer most likely to engage in mass customisation. Furthermore, the trial excluded participants who had trained or were working as industrial designers.

The results and conclusions of the trial are discussed in the following post.