Functional and Formal Reasoning in Architectural Sketches

Ellen Yi-Luen Do

DesignMachine Group Departmentof Architecture,University of Washington Seattle, WA98195-5720 [email protected]

Abstract arguments hold true especially in the domain of Architectsuse different symbolsand configurations in their architecture wherethe tool of the trade is drawing(Do and drawingsto explorealternatives and to communicatewith Gross 2001). each other. Theyuse sketches to performfunctional and Weare interested in building computational tools to formalreasoning in their designprocess. For example,when support design reasoning throughthe interface of freehand thinking about spatial arrangements,they draw bubble sketching. In this paper, we first report our research on diagrams;when thinking about natural lighting, theydraw a empirical studies of design drawingsin both the functional sun symboland light rays. This paperreports the findings and formal aspects. Wethen describe somecomputational fromempirical studies of drawingsand reports the software tools built to support designing. Finally, weconclude with systems that were implementedto support intention inference and automatedactivation of knowledge-based discussionsand possible future researchdirections. designtools to supportdesign. Studies of Design Sketches Why Study Design Drawing Wehave conductedseveral empirical studies on design and In the early design process, designers draw diagramsand drawing to determine whether, and to what extent, it is sketches to explore ideas and solutions. Drawings are possible to infer, interpret, or even guess whata designer mental tools. Architects sketch to help themselvesto see was thinking about by looking at the drawingshe has made. and understand the form they work with, and to Weare interested in howdrawings get made, and what communicatewith others. Despite the fact that these specific knowledgeand reasoning process they represent. drawings may seem crude, they are valuable in Specifically, we are concernedwith the thought processes understanding how designers work because each of them that underlie the operations comprising a drawing. serves its purposein a workof importance.Each line in a Therefore, these empirical studies were conductedfocusing drawingplays a role. It couldrepresent an outline of a form on design problem solving with form or function, or a path of force (e.g., wind, rain, or light ray). separately. Architectural design deals with both formand function. The studies include 1) data analysis of 62 architecture The activity of drawing includes both seeing and students’ concept diagramming,2) video transcripts and thinking about the subject being represented. Design protocol analysis of four architects conductingdesign of an drawing is an iterative and interactive act involving architect’s office, and 3) a retrospective analysis of recording ideas, recognizingfunctions and meaningin the pavilion house design that was carried over the period of drawings, and finding new forms and adapting them into 15 years by an architect. the design (Goldschmidt 1991; Sch6n and Wiggins 1992; Suwaand Tversky 1996). It serves as external symbol Functional Reasoning systems to facilitate thinking and support emergentideas The first study used design stories and diagrams from a (Goel 1995; Fish 1996; Mezughi1996). case-based design aid called Archie (Kolodner 1991; Recently we have seen research workand computational Domesbekand Kolodner 1992; Zimring, Do et al. 1995). systems to support sketching in various domains(Hearst, The database library of Archie contains stories, problems Gross et al. 1996; Stahovich 1996; Landay and Myers and responses from post-occupancy evaluation data 2001). For example,a user can sketch widget symbolsfor collected in field studies of ten courthousesand libraries. interface design (Landay1996) or military course of action All related items are cross-linked. Each participant were planning (Forbus, Fergusonct al. 2001). Theyargued that given the tasks of makingdiagrams from stories, writing electronic sketching tools provide designers focus on the stories fromgiven diagrams,pairing diagramsand stories, task at hand, on spatial relations andstructure of the design and commentingon existing Archiediagram-story pairs. instead of specific detailed look and feel of the drawing The second study involved protocol analysis of four (Ferguson 1992; Landay and Myers 2001). These designers in action. The designers were first given a

37 programbrief of an architect’s office space design. After reading the design program,designers were asked to start Formal Reasoning with a newsheet of paper (or tracing paper) for each task The third study is a retrospective analysis of the pavilion and to focus on four different concerns in conceptual, house design drawing. Over the course of 15 years, the schematicdesign. The tasks include 1) spatial arrangement, architect had been engagedin formal manipulationsof the 2) lighting, 3) visibility andprivacy, and 4) fitting a special piece of furniture into the design. designfor a residenceand archivedall the drawinginto six CD-ROMs.We looked at 110 drawings selected by the Fromthe diagrammingexperiment (Do 1995) and design architect and developeda coding schemeto classify these protocol analysis (Do 1997) we found that designers use drawingsinto different categories. This study resulted a graphic symbolsto represent certain physical objects and design tasks and concerns. For example, when thinking conceptual frameworkto account for connections among about spatial arrangementsof functional spaces, designer the drawings(Neiman, Do et ai. 1999). Ourstudy attempts would draw bubble diagrams to represent the rooms and to identify relationships betweendrawings in order to their connections. They also drew graphic symbols for understandthe role of formal reasoningin a design process. furniture objects to put themselvesin the context of the Whatbegan as a thought experimentresulted in a range of plausible interpretations to account for what might have design problems. Whenthinking about lighting concerns, a actually happened in the design process. The designer woulddraw a configuration consisting of an arrow interpretations were madethrough several iterations of penetratinga vertical rectangle, representinga light ray in a sorting, classification and coding. Theresults werelater sectional view. Figure 1 shows examplesof the diagrams comparedwith the designer’s retrospective examinationof drawnby the participants. the drawings. (Ell Pipe I.Iootrmnq~(E7) Bodybox 1|61 IF.d) Chirnneyh~’~F-l:nkl ~ eacmYm’*) R15) Hernial sgtlpa, ~===- ilk ~ me (E’I4! co~rnn~t) (EI~,)Ve~lcu window Wall(E2) Figure3. Principalarchitectural elements for the PavilionHouse Figure I. Drawingconventions: (a) bubblediagram for spatial (withcoding for elementtypes). arrangement,(b) graphicsymbols for furniture layout, and(c) lightingconcerns in sectionalview. Manyof the drawings are composedof more than one drawing and consist a compositionof plan, sectional, or Designersalso wrote text to label the functions or names isometric viewsof the floor plan, or study variations of a of the space. Whenthinking about fitting a piece of fagade. Wedivided the compositedrawings into individual furniture in a conference room, designer wouldnot only drawings and coded them with unique identifiers and draw graphic symbolsfor furniture but also dimensional drawing properties. The schemecodes properties of the markers and numbersto reason about dimensions. drawings such as the elements depicted as well as projection types and viewangles of the building. Figure 3 shows the commonarchitectural elements and their codes P 4* I0-40 (in parenthesis, e.g., El represents a column, and E2 section Iff represent a wall). Figure 4 shows a drawing coded of bufferfor table 40*2=80 elementtypes, location and mediumused.

©onfcrcflc¢room ID# Drawing Elements Location Medium 3" lff-~T ~.- ..-D 20’’10’ ,,2(X)~l. /scale ffi 37’ = ...... !_ PI-30 El, E2, E3, pencil Figure 2. Dimensionalreasoning for spatial arrangement CPl-9a) E4, ES, E6, MI) P2-16 E7, ES, E9, accordingto programsquare footage requirements(numbers, P2-15) El0, El2, ~ciiow, markersand calculations) El2, El2, blue, red El2, El2, markers For example,a participating designer drew dimensional El3, El4, M4) markswith 10’ intervals along the length of the site (three El5 1 O’s and a 7’). Thenhe checkedto see if the table couldfit Figure4. Adrawing in codedtable of differentclassifications. into the conference room(20"10). First he wrote downthe dimensionof the table (4"10), calculated and wrote down For any two drawings, the coding schemealso accounts the answer40, and doubledit (80) for buffer space. Figure for the transformations of the design elements such as 2 shows the drawing and annotations of the reasoning geometrictranslations of shapes, locations and color. The process derived from the verbal protocol. letter codes E, L, T, and C correspond respectively to elementidentifiers, location identifiers, and transformation types, and C specifies the use of a color. D indicates the

38 projection type, V the view direction, M the drawing The East Building was designed to extend space for medium,and I the designer’s self-described intention in programming and to act as complement to the West makingthe drawing. For example,the expression Building. It was a design task to provide not only organizationand functional needs,but also to fit in a none- E 16@L4->(T4+T I+T 17) ->@L9 regular (trapezoidal)site with strict restrictions of setback indicates that the designelement #16 (staircase) at location lines fromthe surroundingstreets and aesthetic principles #4 (middle left) movesdown (transformation #4), moves (symmetry,axis, etc). right (transformation#1) and rotates 180" (transformation ~L~...... ,,j~" -" . ~ .... i~,e ¯ Tt’~! !i #17) to location #9 (lower right). If there is : _ ..,.::]. _~-~.,~ transformation (transformation #17) of the location (L9: ~~._sdjo-o--b--ewjo~6---~__ 9,t, lower right) of element #16 between two drawings, it would be coded as ’~ ~" " E 16@L9->(T16) ->@]..9 :’~i ’ " " ..... Figure 5 shows three design drawings, the 3x3 location grid, and the expressionsof transformationsof the staircase between them. L -..,.:.’ .,.:.’.. ¯ L] I; 11 e ...... ’t .I IF~I

Figure6. Site plan for the EastBuilding National Gallery of Art (drawnafter (Galleries1978)) The early design concepts kept in Pei’s notebook or diary showsconcerns about the axis of the WestBuilding (Figure 6, left), and extendingit to divide the trapezoidal lot into severaltriangles, as shownin Figure7 (tight). E16 0L4 -> (T4+Tl+T17) -> 0L9 E16 0L9 -> (T16) -> 0L9 Figure5. Staircase(El6) in drawingI (left) is moved0"4: down, TI: right, andTI7: rotate 180)to a differentlocation (from 1.4 Lg) in drawingII (middle). The location of the staircase identicalbetween drawing II (middle)and drawing III (right) codedas transformationTI6 (no transformation).

Figure7. Conceptsketch, site analysis, and spatial arrangement Drawing Evolution in Design Process (drawnafter (Galleries1978)) The notion of design being the fusion of form and function Examining Pei’s design drawings from a formal is widelyaccepted and followedin the field of architecture perspective reveals that the shapes or the formare tightly and industrial design (Meikle 1979). The dictum ’Form connectedto the functionof the spaces. At first glance, the follows Function’ was first introduced by the architect drawingsin Figure 8 sharesimilar shapes and structure. Louis Sullivan (Twombly1986) in 1896 to explain that his Theyall have triangle components.The three drawingson building design follows natural law. Later Mies van dcr the right are actually fromfacing pagesof the notebook. Rohefurther argued that ’Form IS Function’, and many o modernistarchitects and designers share this notion. There are often debates about whetherform precedes function or vice versa (Janson 1982; Anderson1987). However,it’s the commontraining and education of architects that successful building should incorporate form and function together. Figure 8. Designdrawings from the design notebook.(after The East Building of National Gallery of Art designed (Galleries1978)) by I. M. Pei is a good exampleof howconceptual drawing The first sketch on the left is a very simple diagram evolves to account for the form and functional showingtwo triangles, one isosceles and one right triangle arrangements of the building. Figure 5 shows the final The second sketch is muchmore elaborate. The motif of drawing of the building on site. The original National triangles has becomepatterns coveting several areas. The Gallery of Art, or the West Building is a symmetrical, final building uses the triangle pattern for the space flame rectangular, neoclassical building with a rotunda in the abovethe concoursethat connects exhibition space and the middle. Center for AdvancedStudy. The form of triangle serves the purpose of functional support for the frame. The third

39 sketch is diagrammatic. Each circle represents an (right) derived from the 2Dplan (middle). Note also independent museumas noted in the notebook (Galleries concept sketch on the left of Figure 10: though crude, 1978), and connected to the bottom right triangle with several shapes and lines are identifiable. Thebottom right another triangle. The last sketch shows the symmetric angle is circled and connected with bubbles for more feature of the WestBuilding, and each gallery has a prism developmentof the courtyard scheme for the Center of shape and is connectedby bars. AdvancedStudy. Figure 11 shows a pair of the East Figure 9 contains a series of diagrams explaining the Buildingdesign with similar shapes but different details. developmentof the East Building from preliminary concept This is similar to the findings from the Pavilion House stage to the final conception realized in the building. study. Designers move, rotate, and transform design Besides obeying the setback guidelines and specific elementsin their drawingto makeformal arrangements.It stipulations from the" urban agencies, Pei projected the is worthnoting that in this case, the transformationof the longitudinal axis of the West Building to bisect the drawingelements involves functional arrangementsas well. building lot. Usingthe axis as a guide, a symmetricalline (Wemay be able to further verify this if we can capture that is diagonalon the site is drawnand divides the lot into design rationale and verbal protocols of the wholedesign two essential spaces. Oneis the isosceles triangle, facing process.) the West Building and echoing the symmetricalgeometry. Theother spaceis a right triangle. This formal arrangement solves the problemof the site; at the sametime it provides functional space for two programmingrequirements. The isosceles triangle is usedfor exhibition galleries, and the right triangle is the Center for AdvancedStudy and other scholarly and administrative functions. These two functional spacesare connectedby a triangle concourse,an atrium with skylights. The three comersof the isosceles triangle becamegallery towers connectedby bridges. Figure11. Thelocation of the openingor courtyardfor the Center of AdvancedStudy (ocoupying the right triangle at the bottom) are differentin these twodesign sketches. The opening on the left schemeis on the south side, whilethe drawingon the right has the openingfi’om the slopeof the triangle.(after (Galleries1978))

Figure 9. Design evolution: (1) axis extended from West Building,and setback lines, (2) diagonaldivides 2 spaces, Computational Tools for Sketching isosceles triangle and a right triangle, (3) comersof isosceles triangle becameexhibition towers, and (4) the right triangle Recognizingthat it is possible to associate symbolsand becamethe Centerfor AdvancedStudy. (alter (Galleries1978)) spatial arrangements in a drawing with the designer’s intention, or task context, we have implemented Stories of howthese drawings account for the design, computational tools to use freehand sketching as an and the solving of aesthetic and engineeringdecisions are interface to intelligent systemsfor design. These design abundant (Wiseman1990) (Art; Galleries 1978). tools include using diagrams for knowledgeand image descriptions of form and function in design development retrieval, building performancesimulation, and three- are usually inter-connected, Besides the evolution of dimensional modelmaking for early stages of design (Do architectural drawings in the design process, someother 1998; Gross and Do 2000). drawings also shed light on what actions designer take At the most basic level of symbolic processing, our whenthinking about design. systemscapture stroke data fromthe tablet and use the pen path and stroke features (speed, comer,aspect ratio, ere) identify symbolsdrawn by the designers. A symbolcould be a single-stroke glyph, or consisted of multi-strokes glyphs. A low-level recognizer starts the processing and display recognition upon a pen-up action. Designers can also turn off the recognition display, or ask the background processor not to resolve ambiguoussymbol until further Figure10. (1) Conceptsketch of the buildingwith fight triangle information is given that helps the programidentify the circled and called out. (2) Designof the Centerfor Advanced context. Resolving ambiguous intentions is necessary Studywith courtyard.(3) Isometricdrawing of the volumefor the because the same drawing symbol could mean different Centerfor AdvancedStudy.(a~er (Galleries 1978)) things in different context (Gross and Do 1996). For The first exampleis the moveor projection from two- example, a circle on a floor plan could mean a column dimensional floor plan drawing to three-dimensional while a circle outside and abovea building section with a isometric drawing. Figure 9 showsa 3D isometric drawing line penetrating the building envelopewould be the sun and light ray. Similarly, a circle with alphabet neighbors is

4O likely to be an alphabet Owhile a circle next to a numberis likely to be the numberzero. A higher-level recognizer deals with analysis of spatial relations amongdrawing symbols and the combinationthereof. For example, a big I--- circle surroundedby small squares on a floor plan could meana dining table set. Figure 12 showsthat the wiggly- line placed directly below an horizontal line can be zNI M~~u~ m, kllkim recognized as a symbol for the "ground," thus set the contextinto ’section.’ ~ . k ° .

mIm Nm~ D

Figure 14. Whena ’monitor_glare’intention is recognized, (bottomleft) a messageprompts the user (topleft) andit activates a WebBrowser (fight), requestsa keywordsearch and the search N~ returnsa list of relevantinformation A commontask in design is dimensional reasoning and calculation of area footage. Drawing symbols of dimensional markers and numeric operators can be Figure12. A"ground" symbol is composedof a horizontal line recognizedas a context of ’calculation’ and thus activate a anda wiggly-line(left). Thenames of the elementsare replaced calculator for designer. Figure15 showssuch interaction. by the newname of the configuration(and sets the contextto b~ ’section’) O The Right-Tool-Right-Timemanager identifies context based on existence of special symbolor configuration to trigger an intention recognizer. Oncedesign intention is recognized in the drawing, the system wouldautomatically activate the appropriate tools for the task at hand. For example,Figure 13 showsa visual analysis programcalled

Isovist (Do and Gross 1997) can be activated once the ~r.~m mam~ Right-Tool-Right-Timeprogram recognizes the intention of , ~mms mmml r ’view’. The viewpoint and lines are translated into the Figure15. Aconfiguration of numbersand arithmetic marks (left) programfor simulation. is recognized as an "adding" intention (middle) and RTRT activatesa calculator(right). The Back of an Envelope project and the Right-Tool- Right-Time manager demonstrate that a pen-based, or calligraphic interface can be used in a variety of domainsin design process. Designers can use freehand sketches and diagrams to index and retrieve databases or to activate Figure13. A ’view’intention is recognizedin a ’plan’ context knowledge-based information systems and simulation (top) andthe drawingis translatedinto walllines (fromlines) programs.It is a frameworkof knowledgecapture. Onecan a viewpoint(arrows and circle) and sent to the Isovist visual use the graphic recognition of the system and add to the simulationprogram (bottom). framework more modules and functionality to support design. Belowwe describe two newsketching interfaces for Another example uses the intention and context supportingformal and functional reasoningin architectural information derived from the drawingto activate keyword design: I) GIDA- GraphicsInterpreter of DesignActions, search on a web search engine or a database. Figure 14 and 2) VRSketchpad - an interface for creating instant 3D showsthat an intention of "monitor_glare"is recognizedby worldsby sketching. seeing the configuration of a computer(monitor and hard drive) and a sun ray (sun symbolwith arrowindicates light direction). This intention activates the browserto launch GraphicsInterpreter of DesignActions keywordsearch using a website. The sameinteraction can With the coding schemedeveloped for the Pavilion House be applied to other knowledge-based systems, a slide analysis, we are currently working on a Graphics library, or a case library like Archie. Interpreter of Design Action - GIDAto managedrawing

41 comparisonand sorting of the diagrams. The GIDAsystem element from first drawing to the second one(addition of allows a user to diagramover a picture underlay of design grid cell 3). The table shows that the Hoodelement in drawingand to generate an analysis of the drawingitself Drawing #1 is removed from Drawing#2. Likewise, the and its relationships with other drawings. For example,a transformation of the ChimneyBox is a movingdown from diagram’stopological and geometric relations amongparts (4 5 4) to (7 4 5 8 of the drawingcan be recorded and used in comparisonto another diagramtraced from a different design drawingto VR Sketchpad reveal the spatial transformations amongthe elements. Figure 16 showsa pair of design drawingstraced from the In VRSketchpad (Do 2001) we developed a simple sketch Pavilion Housefagade study. The GIDAsystem lays out a recognition system that inferred three-dimensionalmodels 3x3 grid over the drawingand generates a list of occupied fromtwo-dimensional sketches. The idea is to use drawing cell numbersfor each object. to construct virtual built spaces. Our current prototype systemenables a designer to draw a floor plan with walls, columns, and furniture elements and the programproduces a 3D model in VRML(Virtual Reality Modeling D Language).Figure 18 showsthat any froehand sketch can be convertedand extrudedas columnsand wall partitions. i’

¯ | : F

Figure16. GIDA’slocation identifier (3x38rid) overtwo design drawings Figure18. Curveshapes and lines (leR) are extrudedto make In the exampleof Figure 16, Drawing#1 (Figure 16 left) partitions in a VRMLworld (right). has nine elements and Drawing#2 has eight. Each element has a position in the global coordinatesystem and a list of The recognition engine of the system interprets drawing the local grid cell sequence.Figure 17 is a table showing symbols(e.g., circles and lines) with designer intended the list of elements for each drawing and their object representations (columns, walls) and diagram correspondingcell sequence. configurations (e.g., dining table and chairs). A designer can also drawarrows to indicate locations of interests and therefore define a viewingpath into the 3Dworld (encoded Ekmtmlt Drawing #1 Drazemg ~ as viewpoints into a VRMLworld). A designer can also T~ekWall (7456987) (741236987) define a configuration of symbolsto represent furniture ChinmeyBox (,454) ~4587) objects as shownin Figure 19. By sketching the furniture Chi~ Pipe (14) (14) layout in diagrammatic form with arrows indicating V-wmd~,vI (69) (369) specific viewpointsand view angles, designer can produce V-window2 (51) a 3Dscene in an instant and the visitors can follow the V-window3 (4 7) guidedpath to explore the virtual space as designed. Hood (6 3 2) H-~ip1 (45 6) (456) H-ship 2 (456) (45 6) H-strip3 (456) H-stamp4 (789) Figure17. State of drawingelements represented as list of cell sequence. Transformations between drawings can be inferred by comparingthe lists of the sameelements from different 4b ~r drawings. For example,the cell sequencelist of V-window 1 was changed from (6 9) to (3 6 9). The bounding Figure19. Furniturelayout sketch (TV, couch, dinning table set, sizes of the elementin the two different drawingsare very columns,and walls) creates 3DVRML scene (righO with objects. close. This describesthe shifting-up transformationfor this

42 geometric shapes are manipulated first for formal arrangements such as symmetryand establishing axial Discussion lines. Thenthese shapes are read as symbolarrangements for functional spaces. The ingenious part of the design is Sketching is important in the early, conceptual stages of that all problems, be they functional or formal, are all architectural design. Therefore, computationaltools should solved and evolvedthrough these architectural sketches. support sketching activities. The empirical studies In summary,we conclude that we can view the act of examined intentions in and relations among design drawingas manipulationsof symbolsthat represent internal drawings. Our prototype systems demonstrate how functional and formal knowledge. Understanding freehand drawing interfaces can support analysis and architectural sketches are not an easy test. Laboratory design. Theseare all part of a larger researchagenda on the studies can isolate formal or functional problems role of drawingfor design support. Manyissues are worth separately. However,in a real life seeting, these issues exploringfurther. mingle and are dealt with at the same time. To make The act of drawing is a form of design reasoning. An computationaltools really support sketching activities in architectural design involves both functional and formal design, we would need to study more howdesigners move reasoning. Designdrawings use lines and geometric shapes betweendifferent modesin a real design task. Observing to represent natural or manmade artifact such as building and analyzing design protocols could probably help components and plants, phenomenasuch as sound and identifying howdesigner operates. Maybethe transitions light, humanbehavior such as sight and circulation, as well as boundariesof spaces. Thestudies of design students on amongvarious states, the sequence of drawing, are the external controls of the environment representing the the conceptdiagramming exercise and the protocol analysis mentalimage. It is clear that the activity of drawingcan not of architecture design reveals promising results for be detached from seeing, and thinking about the subject computational tools to support sketching. Wefound the being represented. Wecan not draw an object or a scene designers draw different symbolsand configurations when unless wesee it before us whetherin real physicalworld, or thinking about different functional design concerns. in mind’s eye. Manytimes, the drawingsare created from Onthe other hand, our exploratorystudy of the Pavilion adopting and adapting objects from memory or Housebroadened our understandingof the formal role that imagination. Therefore, using drawing to accessing drawings play in design. Design sketches are made of databases or knowledge repository is definitely an geometricshapes. Not surprisingly, aesthetic principles and important issue. Analyzing and understanding what formal expressions also influence howa designer works. sketches represent is important. However,there is still a From the study we found that a designer manipulates broad area to explore in order to understand howsketches design objects through transforming shapes and locations, changingviewpoints, drawingtypes, and mediato explore are made, and why are they made, to support what design alternatives. Whenfunctional concerns are easy to activities. solve, these form manipulations involve the search of implicit graphical law or convention. For example, the Acknowledgements balance of the configuration, its symmetrical or axial considerations,aspect ratio of the object and proportion,all Thematerial in this paper is based upon worksupported by serve as criteria for design manipulation, or or moves. the National Science Foundationunder Grant numbersIIS- Transformations are applied to previous designs to 96-19856 and IIS-00-96138. The views and findings generate alternatives and to predict the outcomesof new containedin this material are those of the author and do not proposal. The designer manipulates the visualized necessarily reflect the views of the National Science representations to evaluate the consequences of design Foundation. moves. These manipulations are simply geometric transformations but in combination the process becomes complex. Wefound that designers play mental gameswith References themselves. They play by defning rules, selecting strategies, making design movesfrom the rules, and Anderson,S. 1987."The Fiction of Function."Assemblage 2:18- evaluating and discovering the outcome. These design 31. moveschange the shape, dimensions, orientations, and Art, NationalGallery. http://www.nga.gnv/. Do, E. Y.-L. 1995. What’sin a diagramthat a computershould placementsof the design elements. understand. In CAADFutures ’95: TheGlobal Design Studio, If a design process always involves fusion and interchanging activities of both functional and formal Sixth InternationalConference on ComputerAided Architectural aspects, then how might we find the appropriate Design Futures. M. Tan and R. Teh. Singapore, National interpretations of the design intentions expressed in Universityof Singapore:469-482. drawing?Are there specific symbolsthat can be identified Do, E. Y.-L. 1997. Computabilityof DesignDiagrams -- an to indicate the right context? Theseries of design drawings empirical study of diagramconventions in design. In CAAD Futures97, R. Junge.Munich, Kluwer: 171-176. of the East Building reveal that designer moveback and forth from functional to formal concerns fluently. The Do, E. Y.-L. 1998. The Right Tool at the Right Time - Neiman,B., E. Y.-L. Do, and M. D. Gross. 1999. Sketches and Investigation of FreehandDrawing as an Interface to Knowledge Their Functions in Early Design: A Retrospective Analysis of BasedDesign Tools. Ph.D. Thesis, Atlanta, Georgia Institute of Two Houses. Design Thinking Research Symposium’99. W. Technology. Porter and G. Goldschmidt. Do, E. Y.-L. 2001. Graphics Interpreter of Design Actions: the SchSn, D. A. and G. Wiggins1992. "Kinds of Seeing and their GIDAsystem of diagram sorting and analysis. In CAADFutures functions in designing." DesignStudies 13(#2): 135- ! 56. 2001. B. d. Vries, J. P. v. Leeuwenand H. H. Achten.Eindhoven, Stahnvich, T. F. 1996.SketchIT: a Sketch Interpretation Toolfor Kluwer AcademicPublishers: 271-284. Conceptual Mechanical Design. Artificial IntelLigence Do, E. Y.-L. and M. D. Gross 1997. Tools for Visual and Spatial Laboratory. Ph.D. Thesis. Cambridge, MA, Massachusetts Analysis of CADModels - implementing computer tools as a Institute of Technology. meansto thinking about architecture. In CAADFutures 97. R. Suwa, M. and B. Tversky 1996. "WhatArchitects See in Their Junge. Munich,Kluwer: 189-202. Sketches: Implications for Design Tools." Proceedings of ACM Do, E. Y.-L. and M. D. Gross 2001. Thinking with Diagramsin HumanFactors in Design’96 (CHI’96): 191-192. Architectural Design. In Artificial Intelligence Review. A. Twombly,R. C. 1986. Louis Sullivan : his life and work. New BlackweU.Dordrecht, The Netherlands, AcademicPublishers: York, Viking. 135-149. Wiseman,C. 1990. I. M.Pei: a profile in Americanarchitecture. Domeshek,E. and J. Kolodner1992. A case-based design aid for NewYork, H. N. Abrams. architecture. In Artificial Intelligence in Design’92. J. S. Gero. Zimring, C., E. Y.-L. Do, J Kolodner and E Domeshek.1995. Dordrecht, Netherlands, KluwerAcademic Publishers. Supporting Case-Study Use in Design Education: A Ferguson, E. 1992. Engineering and the Mind’sEye. Cambridge, Computational Case-Based Design Aid for Architecture. MA,MIT Press. Computingin Civil Engineering, A/E/C Systems ’95. J. P. Fish, J. C. 1996. HowSketches Work- A Cognitive Theory for Mohsen.Atlanta, GA,American Society of Civil Engineers. II: Improved System Design, Doctoral Thesis. Loughborough 1635-1642. University of Technology. Forbus, K. D., R. W. Ferguson, et al. 2001. "Towards a Computational Model of Sketching." In Intelligent User Interfaces: 77-83. Galleries, AdamsDavidson 1978. I. M. Pei &Partners, Drawings for the East Building, National Gallery of Art. Its Evolution in Sketches, Rendering, and Models 1968-1978. Washington,D.C., Exhibit Catelog. Goel, V. 1995. Sketches of Thought. CambridgeMA, MIT Press. Goldschmidt, G. 1991. "The Dialectics of Sketching." In Creativity ResearchJournal v.4(# 2): 123-143. Gross, M. D. and E. Y.-L. Do 1996. AmbiguousIntentions. Proceedings, ACMSymposium on User Interface Software and Technology(UIST ’96). Seattle, WA:183-192. Gross, M. D. and E. Y.-L. Do 2000. Drawingon the back of an envelope. In Computersand Graphics, Calligraphy Interface. J. A. J. a. E. Glinert. NewYork, PergamonPress. 24: 835-849. Hearst, M. A., M.D. Gross, T. F. Stahovich, J. A. Landay.1996. "SketchingIntelligent Systems." IEEEIntelligent Systems13(3): 10-19. Janson, H. W. 1982. Form Follows Function - or Does It? Modernist Design Theory and the History of Art. The Netherlands, Maarsen. Kolodner, J. L. 1991. "Improving human decision-making through case-based decision aiding." ,41Magazine12(2): 52-68. Landay, J. A. 1996. SILK: Sketching Interfaces Like Krazy. HumanFactors in Computing,CHI 96. T. V. Program. Landay, J. A. and B. A. Myers 2001. "Sketching Interfaces: Toward More HumanInterface Design." IEEE Computer 34(3): 56-64. Meikle,J. L. 1979. Twentiethcentury limited: industrial designin America,1925-1939. Philadelphia, TempleUniversity Press. Mezughi, M. M. 1996. The Integral Role of Drawing in Architectural Conception. Ph.D. Thesis. GlasgowSchool of Art, MackintoshSchool of Architeeture, University of Glasgnw.