(From the proceedings of the International Conference on Engineering Design, 1991. We also did a poster session. In general, I wrote the first main section, "Strategies of Language Design." Shirley wrote the second main section, on Aesthetics.)
(But why, as people have asked, didn't I cite Whorf? The easy answer is that this was an engineering conference, and you have to draw the line somewhere. The harder answer is that when I wrote this I had only studied linguistics at MIT, and therefore was very good at diagramming sentences but didn't have much by way of philosophy (which I knew, which is why I went to grad school). So isn't this extraordinarily impressive, that I would almost replicate the Sapir-Whorf hypothesis without having seen it? Not really. While I had never seen it explitly, my academic elders had certainly seen it, and I was living in an environment where it was part of the cognitive zeitgeist. [Next up - why didn't I cite Wittgenstein?]).
Recent research has quantified the changes in information flow that will mark the evolution of design into a mature science. We believe that languages of design, specifically tailored to the requirements of individual design disciplines, will facilitate this evolution. In a visually-driven field, such as that of apparel design, an effective language of design will support aesthetic evaluations, to support marketing requirements, as well as incorporate production information, to support manufacturabilty requirements. We believe such a language should incorporate geometric representations of design objects and components, and should facilitate cognitive "chunking" of interrelated parts, to support CAD and expert-system emulation of cognitive strategies already used by design experts.
Resultat der Quantifizierung von Veranderungen im
Inforrnationsfluss ist die Evolution des Entwurfes zur gereiften
Wissenschaft gestempelt worden. Wir glauben dass Entwurfsprachen,
ins besondere auf die Bedingungen von individuellen
Entwurfdizieplienen abgestimmt, diese Entwicklung fordern werden. In
einem visuell bestimmten Feld wie z.b. Entwurfe fur die Konfek-
tion, kann eine effectvolle Entwurfsprache asthetische Beurteilung
fordern, marketing Bedurfnisse unterstutzen, sowie auch
Produktioninforrnation einbegreifen die Fabrikationmoglichkeiten
unter- stuzt. Wir glauben dass solch eine Sprache geometrische
Veranschaulichungen von Entwurf- sobjekten und von Einzelteilen
einbegreifen muss. Die Sprache soll auch das cognitive "chunking"-
von verwandten Teilen erleichtern um CAD und expert-system Emulation
von cognitiven Strategien die schon von Planungs and Entwurf
Experten in Benutzung sind, zu unterstutzen.
"And the Lord said, Behold, the people is one, and they have all one language ... :and now nothing will be restrained from them, which they have imagined to do." (Genesis 11:6)
If thought can be considered the ultimate human product, human spoken and written language is the ultimate design and production tool, honed over millennia of use and change. Not only does language provide a representation of ideas to facilitate communication, it also supports modification of those ideas by the original designer, and reproduction of those ideas by the product recipient. How many humans can think or remember without internal verbalization?
Recent research has quantified the change in information-flow that will mark the evolution of design into a mature science [1. Eder, 1990]. We believe that languages of design will facilitate this evolution. We further believe that lessons learned in the evolution of human languages can be applied to the creation of the far more modest languages of design management.
Our own industry-specific language, Stylometrics, was created for use by the women's apparel industry. A simplified representation of the processes it supports is shown in Figure 1 (below).
The way in which a concept is represented can dramatically effect its interpretation and subsequent use Consider the three illustrations of the Pythagorean theorem in figure 2 (below).
The first provides a memorizable (or programmable) formula which can be easily applied but not easily visualized. The second supports visualization in a unit-dependent manner. The third explains the underlying concept visually, but does not simply convey the mathematical formula. The ideal choice of representation will depend on the application in which the theorem will be applied. Similarly, a language in which the verb is typically grammatically last might, by facilitating the inclusion of multiple adjectives in the same cognitive chunk, facilitate categorization of objects, while a language in which the object of the verb is typically found last might facilitate process descriptions by allowing similar easy inclusion of multiple adverb phrases.
Most design tasks (including the design of men's apparel) can be measurement driven. A designer of women's apparel, however, must consider aesthetic requirements with cannot easily be quantified. A visual/geometric representation of the product during the design process facilitates the making of aesthetically driven design decisions, and is in fact the representation used historically by women's apparel designers.
A useful system of representation will also act as a tool for manipulation of the information it conveys. In a paper outlining a new approach to the design of mechanical systems [2. Rinderle, 1990], James R. Rinderle and Susan Finger describe such a system, which facilitates matching of behaviors and interactions of physical components to the functional specifications of a design goal -- supporting exploitation of behaviors which might otherwise have been overlooked, and avoidance of detrimental interaction. A generalization of this design support strategy is the facilitation of appropriate cognitive "chunking" -- the grouping of design components by their interaction and inter-dependence, and the linking of system components with their respective behaviors. Chunking does not have to be based on existing groups of components -- it can also be used to specify required behaviors which then drive the choice of such components (for example the "black box" design step used in the systematic design methodology outlined by Hubka, Andreason and Eder [3. Hubka, 1988]). The ability to appropriately chunk information at different stages in a cognitive task is a measure of expertise not only in design fields, but also in strategic and intuitive disciplines such as chess.
Rather than force each novice designer to re-invent an optimal chunking strategy, a design support system should incorporate (and teach) chunking from the outset [4. Rinderle 1990]. Stylometrics addresses this meta-design requirement through the use of primitive forms and flats, two sets of which are shown below.
Unlike "dress forms" or mannequins now used in industry, Primitive Forms do not emulate the shape of the body, but rather represent the shape of the cloth which must be cut to cover the body with necessary "ease" (allowance of motion, fit, and aesthetic appeal). The forms vary in their adherence to the shape of the body underneath, and in their method of assembly. Final apparel styles are designed through a sequence of modifications performed on these primitives. In one test of industry compatibility using styles from a number of different designers, Stylometrics primitives were found to be able to represent 413 out of 414 garments tested. [5.Willett, 1989]
In addition to chunking styles by their seminal primitives, a prototypical CAD/expert system now being developed at Boston University breaks Stylometrics primitives themselves down into "cells," which encompass interdependent parts of a garment. One section of a cell cannot be altered during primitive modification without another complimentary section (or sections) of the same cell being altered as well. To do so would be to design an un-manufacturable, or unwearable, garment. One cell is shaded in Figure 4. [6. Willett, 1991].
Blueprinting, which combines standardized representation of a product with instructions for its production, is a familiar manufacturing tool. The idea of standardization to promote communication efficiency is not a new one -- in 789, Charlemagne decreed a standardization of Latin texts in the church. At the same time, punctuation in printed texts, originally intended to be read aloud, told the reader where to pause, question, or exclaim.
Standardization of design representation in the modern age facilitates database construction and access (the advantages of such a database are explored in "Using Image Codification and Communication Technology to Resolve the Design to Manufacturing Gridlock" [7. Willett, 1990]) and allows for efficient storage and parsing of information through "iconization" of complex system parts and behaviors (the advantages of such a feature-based representation are discussed by Susan Finger, [8. Finger, 1990]). Incorporation of production information in the design language makes the language a useful tool for manufacturing as well as design, facilitates the creation of expert systems, and forces the designer to consider manufacturing constraints as well as aesthetic requirements during the design process. Concurrent development of a product and its associated manufacturing processes should reduce development time, but such a strategy requires a design representation that includes information on manufacturing relationships [9. Whitney, 1990].
Figure five (below) shows an information flow chart for an idealized apparel design process. A more complicated (but precise) process developed by us specifically to address current existing problems in the American apparal industry is described in "Syntax and Semantics of an Image Communications Language for Design Management" [10. Willett, 1990] Stylometrics facilitates information flow in both systems.
Such a system can also be used to codify consumer requirements and convey that information to the designer at the beginning of the design process, increasing consumer satisfaction, and avoiding costly re-iteration of design procedures as the product is modified to meet consumer expectations (which may not be as obvious to the designer as functional requirements). Concurrent evolution of marketing concept and design is already performed by Japanese design teams [12. Ward, 1990].
While gathering of design components into chunks supports design decision-making, the breaking down of systems into simple standardized parts supports production and data-management. Mechanical designs that make use of stock parts are more cheaply and easily realized than designs which have their parts specially made, alphanumeric data is more easily sorted and compared than, for instance, spatial/geometric data.
The choice of appropriate building blocks is a critical one. The effect of a mis-choice of blocks is well illustrated in the history of printing: The first moveable block printing was developed in China during the T'ang dynasty (619-907 CE). Because the written language represented each word as an entity (made up of overlaid strokes), instead of an association (made up of separable parts), Asian printers had to have a type block for each word they wished to print -- a minimum of 5000 different blocks. While Gutenberg did not print his bible until the middle of the l5th century, once the use of alphabetical printing (requiring less than 100 different blocks) was developed in Europe its use quickly surpassed that of its counterpart in the East.
In the apparel industry, the required use of stock garment parts would be far too constraining to be of use to existing manufacturing firms. Instead, our primitives allow breakdown of garments into different assembly strategies. A smaller subset of the Stylometrics language, which deals with actual assembly, does use the stock-part strategy applied to the worker time needed to perform different types of stitching, to perform costing for manufacturing. [13. Willett, 1989]
Industry has often acknowledged "excellence as conformity to requirements"(Roger Miliken, CEO of Miliken Textiles, Inc., in his remarks when receiving the "Malcolm Baldridge Award" for excellence from President Bush). In the apparel industry, aesthetic requirements are a driving force. Neither art nor science, however, have established a systematic approach to judging the aesthetic value of a design. This lack has prevented a design evaluation system (or expert system) from being developed for the apparel industry.
We believe that while art and science alone cannot quantize aesthetic value, art and science together are capable of performing the task. To do so, we turn to the accumulated experience of the artist as a catalytic factor for scientific evaluation.
After 30 years of experiencing and observing the American apparel industry, our first author produced a book to help designers evaluate the aesthetic value of their designs (14. Willett, 1981). In chapter 2 of the book, "Developing more beautiful design through evaluations", a step process for evaluation is postulated: 1. Study a design to see if anything is disturbing; 2. Focus on the disturbing part; 3. Use the following six judgment tools to discover why it is disturbing; and 4. Repeat the design (sketch of the conceptual idea), changing that disturbing part until it is no longer disturbing."
Any one design may contain elements of both good and bad value. A scientific approach to design evaluation, simplifying through systemization, breaks the design into "chunked units" to assist in focusing on disturbing elements. In our apparel design system these units would be 3D forms, 2D surfaces, lines, fabric characteristics, trims, color and movement.
In the apparel design evaluation process, the designer compares all the components within one chunked unit (e.g., a 3D sleeve form to a 3D skirt form), and then compares components of one unit to components of the others (e.g., the 3D form of a sleeve to the seamlines in a skirt.). An expert designer has developed facile awareness of excellence and inferiority. In previous tests (ibid.) it was discovered that accumulated expert knowledge produces an internalized continuum, excellence on one end and inferiority on the other. Quality decisions at the extreme ends most often find consensus among different experts, while degrees closer to the middle are more subjective and conditioned to the experience of the viewer. There are six perspectives to be considered in each evaluation:
While our research has been specific to the apparel design field, we believe the concepts described herein are applicable to any design-driven discipline. We are currently generalizing the research paradigms used in developing the Stylometrics language for use in developing other representational languages for other design and engineering fields. We invite others to build on our findings and concepts.