Design sketches and sketch design tools

Preparing to load PDF file. please wait...

0 of 0
100%
Design sketches and sketch design tools

Transcript Of Design sketches and sketch design tools

Knowledge-Based Systems 18 (2005) 383–405

www.elsevier.com/locate/knosys

Design sketches and sketch design tools
Ellen Yi-Luen Do
Computational Design Laboratory, School of Architecture, Carnegie Mellon University, Margaret Morrison Carnegie Hall 403, 5000 Forbes Ave, 201 CFA, Pittsburgh, PA 15213, USA Received 20 May 2005; accepted 15 July 2005 Available online 11 August 2005
Abstract
In early stages of design architects often use sketching and diagramming to perform formal and functional reasoning. Design sketches are an external representation aid for visualization and evaluation of the spatial arrangements of artifacts. Symbols and configurations are used in design sketches to define context and object arrangements. This paper argues the need to study design drawing, reports the findings from empirical studies of design drawings, and describes the software systems implemented to support intention inference and automated activation of knowledge-based design tools to support design. q 2005 Published by Elsevier B.V.
Keywords: Architectural objects; Empirical studies; Knowledge-based design tools

1. Introduction—motivation
Designers sketch. The ubiquitous cocktail napkin diagrams and tracing paper sketches in architectural folios, exhibits, and their working environment—studios and offices alike, demonstrates the important roles of drawing artifacts and sketching activities in design. Designing is a knowledge-based activity. This creative act involves analyzing, synthesizing and integrating information such as production requirements and constraints. Diagrams and sketches are symbolic representation that designers developed for problem solving and spatial reasoning for their specific domain or discipline. Functional and formal design intentions are often embedded in architectural sketches. Can then, a computer recognize and understand these intentions from designer’s sketches and further provide feedback or assistance to the designer? Can a computer provide appropriate knowledge-based design tools at the right time for the tasks at hand?
We are interested in building computational tools to support design reasoning through the interface of freehand sketching. Therefore, it is important to approach the issues from two distinct perspectives: (1) study of design sketches
E-mail address: [email protected]
0950-7051/$ - see front matter q 2005 Published by Elsevier B.V. doi:10.1016/j.knosys.2005.07.001

and (2) implementation of sketch design tools. We focus on the designer’s drawing conventions that a computer might understand, and the relationship between drawing and design intent. We describe a computer-based freehand sketching environment for design that tries to deliver the right tools at the right time by interpreting a designer’s drawings. Rather than asking the designer to find and select tools for specific design tasks, we explore the idea of automatically invoking various computational tools based on the designer’s drawing. Developing such a system creates both conceptual and technical problems. If a sketching environment is to present the designers with an appropriate tool, it must recognize what the designer is doing. Conceptually, this requires that there is a link between overt behavior—the marks that the designer makes on the paper or in the computer—and the designer’s information needs. Hence, the first part of the paper discusses design studies to understand if there are drawing conventions shared by architectural designers that might be incorporated into a sketching environment. It is argued that designers do share drawing conventions, both at the level of the individual glyph and at the level of the combination of symbols, and that these conventions map onto the tasks that the designers are engaged in. The second part of the paper discusses the technical implementation of the design environment. This environment encompasses several sketch design tools that allows the designers to keep sketches in ambiguous, ‘sketchy,’ states. It operates on complex

384

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

systems of symbols within sketches, including the topological relationships among symbols. It provides access to a wide range of additional information tools.
The rest of the paper is organized as follows. In Section 2 we argue the importance of sketching activities in conceptual design stages. Section 3 reports empirical studies conducted to investigate how functional and formal intentions are embedded in design sketches and Section 4 presents case studies of architectural design drawings. Section 5 presents freehand sketching software tools implemented to support various knowledge activities in design. Finally, we conclude with discussions and possible future research directions.
2. Why study design drawing
The act of drawing is important not only as a vehicle for communication with others, it actually helps designers see and understand the forms they work with [1]. In the early design process, architects sketch to help themselves to see, to reason, and understand the form they work with, and to explore functional ideas and solutions. Designers— especially architects—are trained to use paper and pencil when developing conceptual designs. Architects are visually oriented and are taught to think graphically [2,3]. They draw to develop ideas graphically, and in the process of drawing, designers communicate their thinking. These drawings are valuable in understanding how designers work. Each drawing marks on paper could represent an outline for a space (room, courtyard) or object (furniture, wall and column) or a path of force (e.g. wind, rain, light ray or circulation).
‘Drawing’ in this paper is used to mean the freehand diagrams and sketches designers draw and use in their early design stages. Designers use the term ‘drawing’ and ‘sketches’ generally to represent all the kinds of marks they make when designing: ranging from concept sketches to construction drawing. Here, in particular, we are referring to freehand drawings instead of drafting or construction drawings. Drawing serves as external symbol systems to facilitate thinking and support emergent ideas. Fish argues that sketches help designers attend to thought and help trigger short-term memory [4]. Goel argues that drawings are used as ‘external symbol systems’ to represent real world artifacts which can be manipulated and reasoned with [5]. Mezughi argues that sketching is ‘the principal means of visualising design solutions and crystallizing the thinking process’ [6].
We also use the term ‘drawing’ to refer to the making of drawing marks. Designers use the act of drawing to help them discover and explore ideas. The activity of drawing includes both seeing and thinking about the subject being represented. Design drawing is an iterative and interactive act involving recording ideas, recognizing functions and meaning in the drawings. It also involves

finding new forms and adapting them into the design. Designers ‘see’ and then ‘move’ the manipulated design objects in their drawings [7]. Schon described the kinds of seeing and their functions as (1) literal visual apprehension of marks on a page, (2) appreciative judgments of quality, and (3) apprehension of spatial gestalts. Goldschmidt views sketching as an operation of design moves and arguments, an ‘oscillation of arguments’ that brings about a gradual transformation of images [8]. She argued that a designer interacts with drawing with ‘seeing as’ and ‘seeing that’ reasoning modalities. Suwa and Tversky claim that designers draw to externalize their concepts and that drawings provide visual cues for revision and refinement of ideas [9].
Recently, we have seen research work and computational systems to support sketching in various domains. Hearst, Landay, Gross and Stahovich outlined sketching systems that support design of user interface, web page, and mechanical device [10]. Landay’s system recognizes interface designer’s sketches of widget symbols and supports interaction behaviors [11]. In the similar spirit, a military course of action planning can be supported with a simple sketch system [12]. They argued that electronic sketching tools provide designers focus on the task at hand, on spatial relations and structure of the design instead of specific detailed look and feel of the drawing [11,13]. These arguments hold true especially in the domain of architecture where the tool of the trade is drawing [14]. Herbert argues that drawing are more than just a convenient strategy for solving design problems, suggesting that they are ‘the designer’s principal means of thinking’ (p. 1), and that a designer ‘must interact with the drawing’ (p. 121) [15].
3. Empirical studies of design sketches
To determine whether, and to what extent, it is possible to infer, interpret, or even guess what a designer was thinking about by looking at the drawing she has made, we have conducted several empirical studies on design and drawing. We are interested in how drawings get made, and what specific knowledge and reasoning process they represent. Specifically, we are concerned with the thought processes that underlie the operations comprising a drawing. These controlled studies attempt to answer two questions: (1) What different activities and drawings consistently appear in the conceptual stages of design? and (2) How are symbols and drawing conventions associated with different design activities.
The studies include (1) data analysis of 62 architecture students’ concept diagramming, (2) video transcripts and protocol analysis of four architects conducting design of an architect’s office. The hypothesis is that designers represent architectural concepts in a consistent and conventional way, using a limited set of diagrammatic elements. In order to

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

385

generalize how different designers actually use design diagrams in different contexts and check for consistency, the first empirical study was designed to see how diagrams are related to certain design concerns (represented as problems, responses, and stories from an architecture case library). The experiment asked design students to both draw and interpret a variety of diagrams representing architectural issues derived from post-occupancy data. The purpose of the experiment is to test the feasibility of using diagrams as a way to access design information. The second design drawing experiment was set up to more generally investigate the relations between drawings, their configurations and design intentions in the process of design. The hypothesis is that designers use common drawing conventions when they design. Moreover, the conventions used will depend on the design concerns being considered. If this is true, the drawing conventions used in design concept experiment (first study) would be consistent with the sketches used in a design process (second study). The goals of the empirical studies are thus (1) to verify the findings about drawing conventions, (2) to further investigate design context and intentions as they are reflected in drawing conventions, and (3) to establish a record of the experiment to demonstrate the feasibility of making drawing convention computable.
3.1. Archie diagram story experiment and graphic symbol conventions
The first study used design stories and diagrams from a case-based design aid called Archie ([16–18]). The database library of Archie contains case stories, problems and responses from post-occupancy evaluation data collected in field studies of ten courthouses and libraries. All related items are cross-linked. These case stories include textual descriptions associated with pictures, videos, or diagrams.
This study focuses on diagrams that explain design issues of environmental forces, building components, and human responses. The term ‘diagram’ is used here to mean a drawing that uses geometric elements to abstractly represent natural and artificial phenomena such as sound, light, wind, and rain; building components such as walls, windows, doors and furniture; and human behavior such as sight, perception of privacy, and circulation, as well as territorial boundaries of spaces or rooms.

Table 1 Groups and their respective test variations

Group
A B C D

Task sequence
1–2–3–4 1–2–3–4 2–1–3–4 2–1–3–4

Story titles
Yes No Yes No

Participants
17 17 11 17

has six sub-tasks, and each sub-task used a different story. Each participant was given four tasks (Table 2). In each task, the first three sub-tasks were concerned with architecturally problematic situations (problems), and the second three sub-tasks with possible design responses to an architectural problem (responses). The time spent by participants on all tasks ranged from 30 min to 1 hour.
The first task (1) asked participants to illustrate given text describing an architectural story. For example, Fig. 1 shows the text from a (response) sub-task of task 1:

Fig. 1. Example material for task 1, diagram-making task.
The second task (2) asked participants to write a brief text description for a given architectural diagram. Fig. 2 below shows a (problem) diagram for task 2 with its title.
The third task (3) asked participants to match six diagrams with six textual descriptions. Two sets of three diagrams and three text paragraphs were arranged in rows and the participants were asked to pair them. Fig. 3 shows one example page.
The fourth and final task (4) asked participants to simply comment on six diagram–text pairs taken from the Archie case library. For example, Fig. 4 shows a (response) diagram–text pair from this task.
3.1.2. Experiment results This study revealed four significant features of design
diagram making [19]. First, participants used only a limited set of symbols to represent design concepts. Second, concepts suggest sectional or plan preferences. Third, keywords were used as text labels in plan diagrams. Fourth, there is consistent interpretation of diagrams among designers.

3.1.1. Experiment sequences and test materials Sixty-two undergraduate design students participated in
the experiment. They were divided into four groups, each of which had a different task sequence (Table 1). The tasks included (1) making diagrams from stories, (2) writing stories from given diagrams, (3) pairing diagrams and stories, and (4) commenting on existing Archie diagram– story pairs.
Twenty-four Archie text-diagram story pairs were selected for the four tasks of this experiment. Each task

Table 2 Standard task sequence in a test (for group A and B)

Task sequence
Pre-test questionnaire Task 1 Task 2 Task 3 Task 4 Post-test questionnaire

Tasks
Background information Illustrating texts Interpreting diagrams Pairing diagrams and texts Commenting archie diagram–text pairs Comments and suggestions about the experiment

386

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

Fig. 2. Example material for task 2, a diagram illustrating architectural problem with a text title.
3.1.2.1. Graphic symbols represent design concepts. From the experiment, we found that designers use graphic symbols to represent certain physical objects, design tasks and concerns. The participants chose primitives from a limited universe of geometric shapes and symbols to draw their diagrams, and composed these in highly conventional ways. They predominantly used lines, ovals and blobs, and rectangles. The basic drawing elements—the lines, arrows, and geometric shapes—the ‘primitives’ are used in diagrams in a variety of different domains, and are not limited to architectural design. When primitives are combined, they can form symbols to represent architectural objects such as walls and windows or to illustrate natural phenomena such as sun and human figures. Fig. 5 shows graphic symbols used by the participants to represent human figure, sun with sun rays, walls and windows.

Fig. 6 shows a pattern of preference for plan and section diagrams was found from the experiment results. Functional problems suggest the use of plans, whereas lighting or sight line problems suggest the use of sections. Problems 1, 3, and response 1 dealt with functional placement of spaces, and used plan-based representation. Problem 2, and response 3 concerning natural lighting lead to the use of sectional diagrams with symbols for light and sectional walls. Response 2 was about visual contact to the children’s area, which involved both the arrangement of functional spaces and consideration of visual sigh lines, and hence resulted in a mix of plan and section diagrams.
3.1.2.3. Keywords as text labels in plan diagrams. From the experiment we found that participants were more likely to include keywords from the given text as labels in plan diagrams than in sections. The descriptions about noise and privacy resulted in plan diagrams produced a higher percentage of diagrams with label. The tasks that involved lighting and sight lines produced a higher percentage of section drawings without labels. Functional spaces were mostly labeled by placing text inside a shape (oval, blob, or rectangle), occasionally beside the shape with an arrow or line pointing to the shape. In contrast, concept labels such as ‘supervise,’ or labels of material, such as ‘walls,’ ‘frosted glass’ were usually placed beside the shapes with or without a pointer (line or arrows). Fig. 7 shows diagrams using keyword such as map room (problem 1), courtroom and public area (response 1) as labels. No surprisingly, labels in the diagrams were used by the participants in writing the interpretive texts.

3.1.2.2. Design concepts suggest sectional or plan preferences. Results from task 1 showed that participants seem to share a preference for using plans or sections to illustrate certain architectural concepts. For example, most participants chose a plan view to illustrate relations between different functional spaces and acoustics, but chose a sectional view to illustrate lighting conditions and sight lines. Table 3 below shows that (except for task 5) each of the six sub-tasks in task 1 resulted in a strong preference for either plan or section diagrams.

3.1.2.4. Consistent diagram interpretation among designers. The results from all tasks demonstrated that participants shared similar and consistent interpretations of the diagrams. For example, when given diagrams consisting of bubbles with labels and arrow lines (about the relation of noise and accessibility to a music room and reading area), participants introduced the concept of separation, writing that accessibility leads to adjacency of the two areas and noting that concerns about noise transmission should be addressed by a soundproof buffer. Participants performed

Fig. 3. Example page for pairing test, task 3, diagrams on top row, and texts on bottom row with different sequence.

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

387

3.2. Design context and intentions

To understand more about the role of drawing in design, one needs to look also at sketches made in the design process, not only at diagrams. From the previous study, we have identified several drawing conventions with respect to specific design context and intentions. For example, attention to lighting is often indicated by arrows, attention to functional arrangements with bubble diagrams. However, this previous study only deals with diagrams drawn to illustrate given design concerns, not those drawn in the context of a real design task. This raises an obvious next question: do designers use the same drawing conventions when thinking about design concerns when they design? In order to answer this question, we conducted protocol studies in which participants were asked to perform design tasks while attending to different design concerns [20].

Fig. 4. Example material for task 4, an architectural response diagram–text set from Archie.
the diagram–text pairing exercises (task 3) with ease and most participants agreed with the Archie diagram–text pairs (task 4). They interpreted other designer’s diagrams or texts as they were intended. This suggests that designers mostly agree with each other’s diagrams.
The experiment results support the hypothesis that diagrams are used in a consistent way to represent different issues in architecture. They shared an understanding of the definition of diagram, only used a small set of symbols in their drawings and arranged them in conventional and consistent ways.

3.2.1. Design drawing experiment setups, and tasks Two architectural instructors and two senior design
students participated in this experiment. Each of them was given a design brief for an architect’s office and a sequence of four tasks, where each task asked the participant to focus on a particular issue. The experiment was set up to obtain the designers’ drawings made in response to each individual task. Video recording was introduced to record actions taken in design, and the relations to the verbal descriptions of design intentions. Participants were asked to ‘think aloud’ describing what they were thinking when they design. The design program was to design an office for an architecture firm in a 70 ft by 25 ft one-story warehouse, providing workspace for 3 architects, 3 CAD operators, 2 contract draftsmen, a secretary and 1–2 student interns.

Fig. 5. Graphic symbols for sun, people, walls and windows.

Table 3 Number of plan and section diagrams drawn in task 1

#

Summary of text

Plan

(%)

1

P1

Map room location

58

94

2

P2

Natural lighting create glare

3

5

3

P3

Fun children’s section

45

73

4

R1

Thick walls to reduce noise

40

65

5

R2

Visual contact to children’s area

25

40

6

R3

Glass windows for lighting

7

11

#, Sub-task number, P1, problem 1, R1, response 1, total participantsZ62 (100%).

Section

(%)

0

0

52

84

14

23

8

8

31

50

52

84

Other

(%)

4

6

7

11

3

5

14

23

6

10

3

5

388

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

Fig. 6. Diagram examples showing view preference for descriptions, problem 1—plan representations, problem 2—sectional diagrams and problem 3—plan diagrams. Response 1—plan view, response 2—both plan and section and response 3—sectional representation.

The office was to be designed to have space for work groups, a meeting room, a small kitchenette, a bathroom, and a chief architect’s private office, a general affairs section, storage space, printing and plotting area.
After reading the design program, designers were to start with a new sheet of paper (or tracing paper) for each task and to focus on four different concerns in conceptual, schematic design. The tasks include (1) spatial arrangement, (2) lighting, (3) visibility and privacy, and (4) fitting a special piece of furniture into the design.
Task 1 was to make a conceptual, schematic design for the above program and pay particular attention on zoning arrangements. The instruction sheet included suggestions to consider where to put the lobby, chief architect’s office, meeting room, and the different work group spaces. Task 2 asked the designer to pay particular attention to lighting issues with the suggestions of considering the window location on the wall and introduce skylight into the office if plausible. Task 3 focused on the visibility and privacy concerns between different spaces. For example, one could

consider making each work group space to command a fine view to the exterior, or an easy access to the meeting area. Task 4 asked the designer to make sure the meeting room is large enough to accommodate the chief architect’s favorite meeting table (4 ft!10 ft), as well as making sure the designers’ work space would have at least 800 ft2. Similar to the previous study, a pre-experiment questionnaire was included in the instruction sheets to collect basic background information such as their design experience, and the use of freehand drawing in communication or design process. A post-experiment questionnaire was given to solicit suggestions and evaluations of the task difficulties and their personalized symbols or short hands for design. Participants were asked to complete the first tasks in 10, and 5 min for each successive one. The two students followed the instructions and performed the experiment within 30 min. However, the two instructors simply ignored the time constraints and spent about half an hour each to perform the first task. Therefore, the time spent by participants for the entire design experiment ranged from

Fig. 7. Example of diagrams using keyword given in the text to label a space. Top: ‘map’ in map room location problem. Bottom: ‘P’ and ‘C’ or ‘public’ and ‘court’ for architectural response to reduce noise between courtrooms and public areas.

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

389

Fig. 8. Top: primitives (drawing elements) used in drawing included arrows, lines, hatches and simple geometric shapes. Bottom: symbols (architectural objects) such as walls, windows, and stairs are formed by combination of primitives.

30 min to 11⁄2 h. All of the participants told the observer that they enjoyed the design experiment. The two participating instructors even commented that they were ready to make detailed design and physical models after they completed the design experiment.
3.2.2. Findings from design sessions The four designers all have at least 3 years of design
studio education in college with some experience working at architecture firms. A detail description of the profile of participating designers, detail analysis of the design sessions and the sample transcripts of the protocol can be found in [21]. Below we briefly describe the experiment results.

3.2.2.2. Designers draw furniture in space and to put themselves in context. Another finding of the experiment is that designers draw simple shapes representing furniture in order to put themselves in the right context to think about design. The construction of a space includes architectural elements such as walls and windows. However, additional symbols of furniture often are arranged in a configuration to represent how the space will look and feel. For example, top row of Fig. 9 shows variations of a conference table for the conference room in task 4 and bottom row shows different furniture placed in space for a lobby and office space. Designers used these symbols as a scale indicator and to see and test if the size and dimension of the space is appropriate.

3.2.2.1. Designers share drawing conventions: symbols and their configurations. The conclusion from the experiment sessions and the post-experiment questionnaires is that designers use drawing symbols in their design in a consistent fashion. Participating designers chose primitives from a limited universe of geometric shapes and symbols in their drawings, and composed them in highly conventional ways, just as the results of the previous study. Fig. 8 shows that primitives such as ovals and rectangles were drawn with varying size and aspect ratios, and combined together as symbols to represent architectural objects. For example, a symbol for direction North was composed with an arrow and a letter N; lines were composed to indicate walls and windows; and circle with lines or a blob was drawn to indicate a person.

3.2.2.3. Lighting concerns are portrayed in sectional view with light rays. The result of the experiment suggests that different design concerns tend to be correlated with either sectional or plan representations. The experiment asked the designers to focus on separate issues such as zoning, lighting, visibility and dimensioning. As in our previous study of diagram making, participating designers seem to share a preference for using plan or section to illustrate certain architectural concepts. For example, participants used conventional symbols for the sun, light rays, windows and walls; some drawings also included symbols for persons and computers. Fig. 10 shows lighting included in a sectional view, with light from the sun passing through the building envelope. The changed direction arrows

Fig. 9. Symbols give clues for context. Top row: symbols for the conference space including tables and chairs. Bottom row: furniture drawn for office space arrangement.

390

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

Fig. 10. Drawings illustrate lighting by using light lines that penetrate building envelope from windows and roof.

Fig. 11. Diagrams for layout and zoning included overtracing (circle) to select a space, labels inside the shapes, or lines drew out to indicate space use, and putting emphasis through hatching.

indicate reflecting light from a roof skylight (second to right) and in an interior lighting fixture design (right).
3.2.2.4. Attention and focus can be identified through labels and overtracing. The design drawings from the experiment exhibited that plan drawings have more text labels than sections. The data also revealed that participants frequently included key words from the design concepts as labels in their drawings. Designers wrote text to label the functions or names of the space. Fig. 11 shows examples of the labels for functional spaces written inside a containing shape (oval, blob, or rectangle), or occasionally put beside the shape with an arrow or line from the label pointing to the shape it identified. The video protocols of the empirical studies showed that designers constantly engaged in overtracing, repeated outlining a particular shape or area of the drawing. This overtracing, or redrawing, serves as an act of selection, that draws attention to an element, refines a shape, adds detail to the drawing, or explain something. Designers also used hatching to distinguish a particular space from others.

3.2.2.5. Designers perform dimensional reasoning through figure calculations. When thinking about allocating objects or spaces within required dimensions, designers wrote down numbers besides the drawing to reason about scale and calculate sizes. For example, a participating designer drew dimensional marks with 100 intervals along the length of the site (three 100s and a 70). Then he checked to see if the table could fit into the conference room (20!10). First he wrote down the dimension of the table (4!10), calculated and wrote down the answer 40, and doubled it (80) for buffer space. Designer wrote down a sequence of numbers to label dimensions, to calculate the square footage requirement, and to reason about what the width and length of a space should be. Fig. 12 shows the drawing and annotations of the reasoning process derived from the verbal protocol.
The results from this experiment verified the hypothesis that graphic conventions also appear in a design process. In both empirical studies, we found that novice and experienced designers consistently using diagrams and symbol configurations in their drawing to help them thinking about the design context. For example, when thinking about

Fig. 12. Dimensional reasoning for spatial arrangement according to program square footage requirements (numbers, markers and calculations).

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

391

architects that successful building should incorporate form and function together.

4.1. The case of Pei’s National Gallery of Art

Fig. 13. Site plan for the east building, National Gallery of Art (drawn after ([26])).
spatial arrangements of functional spaces, designer would draw bubble diagrams to represent the rooms and their connections. They also drew graphic symbols for furniture objects to put themselves in the context of the design problems. When thinking about lighting concerns, a designer would draw a configuration consisting of an arrow penetrating a vertical wall with window, representing a light ray in a sectional view.
4. Case studies of design drawings
In the field of architecture and industrial design the notion of design being the integration of form and function is widely accepted and followed [22]. Architect Louis Sullivan introduced the dictum ‘Form follows Function’ [23] in 1896 to explain that his building design follows natural law. Later Frank Lloyd Wright further argued that “form and function are one”[42]. Many modernist architects and designers share this notion. There are often debates about whether form precedes function or vice versa [24,25]. However, it’s in the common training and education of

The east building of National Gallery of Art designed by I. M. Pei is a good example of how conceptual drawing evolves to account for the form and functional arrangements of the building. The east building was designed to extend space for programming and to act as complement to the west building. It was a design task to provide not only organization and functional needs, but also to fit in a none-regular (trapezoidal) site with strict restrictions of setback lines from the surrounding streets and aesthetic principles (symmetry, axis, etc). Fig. 13 shows the spatial relationships of the National Gallery of Art’s original west building and the east building on site.
The early design concepts kept in Pei’s notebook or diary shows concerns about the axis of the west building (Fig. 14, left), and extending it to divide the trapezoidal lot into several triangles, as shown in Fig. 14 (right). This could be interpreted as design reasoning concerning only the form factors.
Examining Pei’s design drawings from a formal perspective reveals that the shapes or the form are tightly connected to the function of the spaces. At first glance, the drawings in Fig. 15 share similar shapes and structure. They all have triangle components. The three drawings on the right are actually from facing pages of the notebook. It appeared that these explorations of geometrical formal arrangements were not just useful construct to respond to symmetry and site constraints, these shapes also served as tokens representing functional concerns that can be manipulated and moved. Functions were embedded in the graphic symbols of these drawings.
The first sketch on the left is a very simple diagram showing two triangles, one isosceles and one right triangle.

Fig. 14. Concept sketch, site analysis, and spatial arrangement (drawn after pei [26]). Fig. 15. Design drawings from the design notebook (after pei [26]).

392

E. Y-L. Do / Knowledge-Based Systems 18 (2005) 383–405

The second sketch is much more elaborate. The motif of triangles has become patterns covering several areas. The final building uses the triangle pattern for the space frame above the concourse that connects exhibition space and the Center for Advanced Study. The lower triangle shows a courtyard arrangement. The form of triangle serves the purpose of functional support for the frame. The third sketch is diagrammatic. Each circle represents an independent museum as noted in the notebook [26], and connected to the bottom right triangle with another triangle. The last sketch shows the symmetric feature of the west building. Each gallery has a prism shape and is connected by bars to the other two galleries. These bars seemed to represent container for circulation traffic or corridor. The opening courtyard in the lower triangle seemed to be a simplified representation of the second sketch.
This series of diagrams explain the development of the east building from preliminary concept stage to the final building realization. This formal arrangement solves the problem of the site; at the same time it provides functional space for two programming requirements. The isosceles triangle is used for exhibition galleries, and the right triangle is the Center for Advanced Study and other scholarly and administrative functions. These two functional spaces are connected by a triangle concourse, an atrium with skylights. The three corners of the isosceles triangle became gallery towers connected by bridges.
Stories of how these drawings account for the design, and the solving of aesthetic and engineering decisions are abundant [26,27]. The descriptions of form and function in design development are usually inter-connected. Besides the evolution of architectural drawings in the design process, some other drawings also shed light on what actions designer take when thinking about design. One example is the move or projection from two-dimensional floor plan drawing to three-dimensional isometric drawing. Fig. 16 shows a 3D isometric drawing derived from the 2D plan.
Pei’s sketches also showed the development of the design involved trying out alternatives such as changing the arrangements and locations of the design components. Fig. 17 shows a pair of the east building design with similar shapes but different details. An architectural component, the opening courtyard for the Center of Advanced Study

Fig. 17. The location of the opening or courtyard for the Center of Advanced Study (occupying the right triangle at the bottom) are different in these two design sketches. The opening on the left scheme is on the south side, while the drawing on the right has the opening from the slope of the triangle. (after pei [26]).
appeared in different locations in the two drawings. This is not surprising. Designers often move, rotate, and transform design elements in their drawing to make formal arrangements. It is worth noting that in this case, the transformation of the drawing elements involves functional arrangements as well. Any move of functional space would result a change of circulation pattern and spatial relations, and hence induce a sequence of design moves and reasoning.
4.2. Architect’s design sketches
Our retrospective analysis and story telling of Pei’s design process with drawing may be compelling. However, was his use of sketching in functional and formal reasoning a personal style? Is there a convention of operation for design drawing that designers follow? Below we looked at several famous architects’ drawings to investigate how different drawings are used in design.
Fig. 18. Sketch (plan and 3D) of the house of the Silver Wedding, Pompeii, 1911 (after Le Corbusiner [36]).

Fig. 16. (1) Design of the Center for Advanced Study with courtyard. (3) Isometric drawing of the volume for the Center for Advanced Study (after pei [26]).

Fig. 19. Variations on the House Forms (after Le Corbusiner [37]).
DesignDesignersDiagramsSymbolsFig