Oxman Rivka (1994) Precedents in design: a computational model for the organization of precedent...

Precedents in design: a computational model for the organization of precedent knowledge Rivka E Oxman, Faculty of Architecture and Town Planning, Technion, Israel Institute of Technology, Haifa, Israel, 32000

A computational model for the organization of design precedent knowledge

is developed. The model is composed of distinct chunks of knowledge called design stories. A formalism for the design story is proposed which

represents the linkage between design issue, concept and form in designs. Stories are structured in memory according to a semantic network. The

lexicon of the semantic network acts as a memory index. The memory structure and indexing system are demonstrated to enhance search and to

support cross-contextual browsing and exploration in the precedent library. The approach is demonstrated in a pilot design aid system in the task

domain of early conceptual design in architecture.

Keywords: case-based reasoning, design precedents, memory organization

1 Clerk, R H and Pause, M Pre- cedents in Architecture, Van Nos- trand Reinhold, New York (1985) 20xman, R E 'Pdor knowledge in design, a dynamic knowledge- based model of design and creativity' Design Studies Vol n No 1 (1990) pp 17-28 30xrnan, R E and Oxmen R M 'Refinement and adaptation in design cognition' Design Studies Vol 13 No 2 (1992) pp 117-134

p rior knowledge has been recognized as a significant source of knowledge in the creative process of design. In employing knowledge from prior designs ideas relevant to the current problem may

be accessed from past designs. Since a design precedent is a recognized past design 1 the term provides a convenient reference for this characteris-

tic of the unique knowledge embedded in a known design. The process of the selection of relevant ideas from prior designs in current design situations has been termed, precedent-based design 2'3. In the course of

exploration of design ideas within precedents, designers appear to be able to browse freely and associatively between multiple precedents in order to make relevant connections. Furthermore, browsing enables the discovery of new, often unanticipated, concepts in precedents.

The computation of these properties of design memory suggests a significant direction for the development of computer systems which can

provide aid to designers. Previous designs may be recalled from an organized memory of design precedents in computerized systems in order to support the human designer with their unique ideas and their explicit design concepts. In such systems, the ability to encode, search and extract design knowledge relevant to the problem at hand is highly significant.

0142-694X/94/020141-17 © 1994 Butterworth-Heinemann Ltd 141

A traditional way of formalizing the organizational structure of precedent

knowledge has been through archival categories such as name, historical

period, style or location which can be used as an index. However, in a

design aid system, memory organization should reflect a cognitive model

of design reasoning in order to support the retrieval of knowledge during

design. A cognitive model of design provides for relevant connections to

be made between the current problem and the design rationale and

concepts underlying precedents. An example of such connections in prior

designs are the relationship between the design issues underlying the

precedent design, their affinity for certain design concepts, and their

resolution in the specific design. Furthermore, an organization of memory

according to such categories may support a cognitive mode of browsing.

For example, a design concept presented in one precedent may be directly

connected to a similar concept in another precedent. Such conceptual

connections are usually achieved cognitively by the human designer.

In this work a computational model for the organization of a precedent memory to support design is proposed and illustrated. The term design precedent is employed to designate a recognized, specific design in which

the unique conceptual points and ideas are denoted as distinct knowledge

chunks. The exposition of knowledge by the proposed system is depen-

dent upon the structure of chunks of information and their representation.

Precedent knowledge is here considered as the explication of the relevant insights of particular designs and the appropriate linkages of information

between multiple design precedents. It is this organizational structure

which contributes to the accessing of knowledge relevant to the problem

in hand.

4 Rleetxmk, C K and Schank, R C Inside Case-based Reason- ing Lawrence Erlbaurn Associ- ates, Hillsdale, NJ (1989) 5 Domesllek, E A and Kolod- her, J L 1992. 'A case-based design aid for architecture' in Artifi- cial Intelligence in Design '92 (Ed. J. Gero) Kluwer, Boston (1992) pp 497-516 6 Go~, A, Kolodner, J, Pearce, M, BIIIIngton, R and ZKndng, C 'ARCHIE: A case-based architectural design system' Re- search Report, College of Com- puting, Georgia Institute of Tech- nology, Atlanta (1991)

The computational paradigm known as case-based reasoning (CBR) 4, is

relevant to our work, since it is based on a cognitive model of memory and

provides concepts and methods which reflect the cognitive content of reminding. Current work in case-based reasoning in design 5'6, is directed

towards the development of CBR-based tools for assistance in the

retrieval and presentation of useful information from organized memories

of past experience. Compared with other systems which employ cases for giving advice, warning of a pitfall 5, or to suggest evaluative criteria 6, we

employ the term design precedent here to designate the design case

knowledge which includes the particular conceptual contribution to design which makes a case memorable as a precedent.

In this paper the theoretical and implementational issues for the construc- tion of a design aid system of design precedents are addressed. These

issues include representation of chunks of precedent knowledge; the

142 Design Studies Vol 15 No 2 April 1994

organization of knowledge chunks in a structured memory, and indexing.

We have attempted to develop a structure of memory and an indexing

system which support search and browsing. In the following sections these

issues of knowledge representation, memory organization and indexing

are considered. In the final sections an approach to the implementation of

a precedent-based design aid system in the domain of architectural design

is demonstrated.

1 Chunking knowledge in design precedents One of the distinctive problems in representing designs is the richness and

complexity of their descriptive content. Each design contains many

related pieces of information which are often difficult to describe or to

decompose. Furthermore, not all of the information embedded in com-

plete and exhaustive records of existing designs may be immediately

relevant for aiding in current design problems. An approach which

addresses these unique problems of representation of design knowledge is

to base it upon a decomposition of holistic case knowledge into separate

chunks of design knowledge. One means to decompose case knowledge

into separate and independent representational chunks is the concept of the story which is currently employed in the CBR community 5'7. It

depends upon breaking the mass of information of an entire case into

distinct, relevant, smaller components. The significant contribution of the

story as a representational chunk is that it is a selective presentation from

a case which has a particular lesson to teach. The design story is employed

here as a device to decompose existing descriptions of complex design

precedents into chunks. For each design precedent, many design stories

can be represented.

7 Schank, R C and Ot~xxl , R 'A content theory of memory indexing', Technical Report 2, The Institute for the Learning Scien- ces, Northwestern University, Evanston, IL, (1990) 8 Curtis, W d R 'Virtuosity around a void' in The Architectu. ral Review Architectural Press, London (1984)

A design story is here defined as an annotation of the conceptual design

content which characterizes the uniqueness of a specific design precedent.

Stories organize facts to make a point. A significant story provides explicit

linkages between design issues of the problem, a particular solution concept, and a related form description of an element of the design

solution. These linkages are, in fact, the knowledge content of the

precedent representation. In the research we have developed a formalism

to make these linkages explicit. For example, consider the following story

related to a well known building design, the Staatsgalerie in Stuttgart, Germany: s

The problems posed by the programme included the relationship between the site

and city. The site sloped down to a motorway that cut the old cultural area of the

city in two (issue of urban continuity). The new gallery had to complement the demands f o r . . , an urbanpath-through the scheme. Stirling found tha t . . , the

Design precedent knowledge 143

9 Rlttll, H Second generation design methods, in N Crou (Ed) Developments in Design Metho- do/ogy, Wiley, New York, pp 317-327 10 McCall, M, FIImher, G and Morch, A 'Supporting reflection in action in Janus design environment' in M McCullough end P Purcell (Eds) The Electronic Design Studio, MIT Press, Cambridge, MA (1990)

circular drum, a public space at the heart of the building organization, could also

act as a pivot resolving the varied circulation patterns.

The story illustrates references to significant concepts in the design of an

urban museum. In the example the design issue is 'urban continuity'. The

design concept which was employed to address the design issue is the

principle of 'path-through'. In this case the concept, 'path-through',

describes the ability to pass through the building without entering it. The

ramp within the drum, or central circular courtyard, is the form element

which materializes the solution principle of 'path-through'. This illustrates

the relationship in a typical design story of a design issue, the concept of a

solution principle responding to the issue, and the form element which

materializes the principle in the design. In the following section we

elaborate on issue, concept and form as components of a knowledge

schema for design stories.

2 Issue-concept-form: a tripartite schema for chunks of design knowledge In order to provide a representational formalism for story chunks, a

tripartite representational schema was proposed. The tripartite schema

consists of an issue-concept-form formalism. Each design story is a way of

linking these three components. In a particular design task domain it is

this linkage which constitutes a meaningful chunk of knowledge. The

implication of this tripartite schema for memory organization, indexing,

and the search will be elaborated in subsequent sections. In the following

sections each of the three components is described.

Design issue Design issue is a point related to the design task which is deliberated by

the designer. Such points may be formulated by the programmatic

statement, the intrinsic problems of the domain, or by the designer

himself. The characterization of design as the deliberation of issues was promulgated by Rittel 9'1°. It offers a convenient term to identify particular

points in design problems. Each story generally contains one issue. For

example, in the story presented above, the issue of 'urban community'

through the project site was a programmatic objective of the design, provision for which had to be achieved in the final design. As illustrated

below, in any precedent there are generally many stories, each one containing an issue.

Design concept The design concept is the formulation of a design idea in relation to an issue. It is a form of ideation related to the design task. For example, the


Figure 1 Three components

of story chunks in precedents

precedent

story

E! design issue I

design concept

• design form

• • 9 •

q! story

~receoent

development of ideas related to a particular issue such as the provision of urban continuity in a building design may take the form of a solution

principle, such as 'path through', i.e. providing a path through the building mass. Rather than a specific design solution, the concept describes the principle of the physical separation of enclosed building

space for public circulation. This concept can be realized in various ways in building design. Another example of another design concept which is

found in the same museum design (Staatsgalerie, Stuttgart) is 'focal space', that is, the employment of a spatial focus, such as a central court,

in order to provide orientation in the building.

Design form

The form is the specific design artifact which materializes the solution principle. For example, in the Stuttgart museum the design concept, 'path-through', is realized by the public ramp passing through the drum,

or open central courtyard. It is important to note that this is the specific element of the total building design which directly relates to the design

issue.

The three components of the design precedent are illustrated diagramati- cally in Figure 1. The large black circles are design issues. These are linked to the small squares which are design concepts. These are, in turn, linked to form representations in the precedent. Figure 2 illustrates the schema in the museum example discussed above.


story

I orientation foc~ story

Figure 2 Three components in

two stories of an architectural

precedent

2.1 A formalism for the representation of design stories A formalism which makes the linkage between issue-concept-form explicit

in design stories was investigated. A preliminary version of a formalism

similar to the universal context frame proposed by Schank 7 has been

developed. It contains a frame, slots and fillers, with slots predefined and

fillers to be filled. Fillers employ the domain categories which underlie

each story. The formalism encodes the particular linkage of components

which are the essence of the story. The essence is a particular exemplary

point contained within the story and worthy of storage in memory. Each

design precedent may contain many such stories.

Stories may be classified into several types according to their content. An

example of one significant type of story is a design conflict. Design conflict

stories are particularly memorable chunks of knowledge in design prece-

dents, since they describe conflicts between objectives and constraints which require trade-offs, or innovations. In the design literature, there are

frequent references to design conflicts, such as conflicts between the

design problem type and the context of the environment.

In a conflict story that illustrates interactions between problem and

context, significant issues may be derived from objectives of the context as well as from objectives of the problem itself. For example, the urban site

of the Stuttgart museum had as a programmatic site objective 'through circulation'. This conflicts with the objective of closed circulation or 'no

pass through' in the design problem which indicates the requirement for the control of entry and exit in the museum building type.

The extended formalism for a story is illustrated below. The formalism may be interpreted in the following way: the issue-problem type and


issue-context type pair represent the problem specification. The concept-

problem and concept-context pair represent the solution principle; while

the form-problem and form-context pair are the details of the form

solution. The slots for issues, concepts and form are organized according to two categories: problem-type and context-type. In architectural design,

problem-type describes the building type; and context-type describes the site. There is an additional category for classifying the story type.

Problem type Context type Story type Building type Site type Story type

Issue Issue Issue

Concept Solution principle Solution principle

form Form element Form element

The applications of the formalism to represent the specific textural story

mentioned above is illustrated below.

Problem type Context type Story type

Museum Urban site Conflict

Issue Control Urban continuity

Concept Path through Path through

form ramp ramp

3 Memory organization and indexing In the following section an organizational model of a memory structure for precedent knowledge and indexing are proposed. The organization of

memory is based on the modules of precedent knowledge which have been presented above. There are design stories clustered into design precedents, each of which contains a substructure of issue-concept and form. This organization will be shown to contribute to an indexing system

which supports issue and concept-related search in the design precedent memory. Furthermore, this approach enhances the capability of browsing within the memory and enables cross-contextual indexing. The development of these cognitive capabilities of conceptual search, and cross- contextual reminding derive from the story-based memory structure.


issue

concept

precedent d ~[It.

,dent

Figure 3 Structure of memory

network precedent

~recedent

11 Barduz, T and Zeld, I 'DE- JAVU' in Artificial Intelligence in Des/gn '92, (Ed. J. Gem) Kluwer, Boston (1992) pp 477-496 12 ~ , p and RNcNmrger, M 'A design dependent approach to integrated structural design' in J.S. Gero (Ed) Artificial Intelli- gence in Design '91 ButtenMtorth- Heinemann, London, 10p 171-187 13 noemmn, u, O~ro, J end Oxman, R E 'What's in a case?' in CAAD Futures '91 Computer Aided Architectural Design - Education, Research, Applica- t/on. (Ed. G. Schrnitt) Vieweg, Wiesbaden, Germany (1992) pp 285-300

In distinction to current case-based design systems, memory is here

organized by the knowledge chunks rather than by complete designs.

Clusters of stories are grouped into design precedent cases. Issues, and

concepts are organized into a semantic net of domain concepts. Concep-

tual links in the semantic network can connect different stories from

different precedents. From any node of issue or concept in the net, related

stories can be retrieved which in turn, call up their precedents. Precedents

then may provide additional references to other related stories. Figure 3

illustrates the memory network.

3.1 Indexing Indexing is a complementary issue to memory organization. Indexing, or the labelling of chunks in cases, contributes to a particular functioning of

retrieval of relevant knowledge from memory. Indexing is, therefore, the functional key to the exploitation of memory structure. Various

approaches to indexing categories have been developed in case-based design systems. Among them are taxonomies of mechanical design 11, typological cate~ories, or principles of component parts12'6; and function- behaviour-structure generalizations 13. Design cases are unique in requir-

ing a complex representational formalism which may include the problem


14 I~Aleeu, R HYPERTEXT: Theory Into Pract/ce Intellect Ltd, London (1989)

description and the situational aspects of problem context, as well as the

design description. Current approaches to case indexing lack sufficient

expressive power to describe the significant relationships in design cases

such as the relationship between the problem type, problem context and

the complexity of real world design situations and solutions. A design

indexing system should, therefore, include case descriptions as well as

reference to problem type and context.

We have stated that decomposing designs into stories as separate chunks

offers a method to deal with the complexity and richness of design

description. In the memory, we have treated each story as an elemental

case. Indexing here becomes story indexing rather than case indexing.

Having developed an explicit internal structure of story representation

which provides for a mapping between the issues, concepts, and the form

description, these assumptions provide a basis for indexing.

We index stories as the main memory elements and employ the fillers of

the representation, the domain vocabulary of concepts and issues, as the

lexicon of the indexing system. As compared to indexing of whole cases by

nominal lexical categories, in the proposed approach the context of the

knowledge chunk becomes the lexical system. By employing this lexical

system linkages between precedents can be established through matching

of issues and design concepts. Two kinds of lexical vocabularies have

already been distinguished in the story formalism above: situation descrip-

tors (problem and context issues) and solution descriptors (problem and

context concepts and form). We shall now demonstrate how these lexical

categories function within the memory structure to support search and

browsing.

3 . 2 Search and browsing It is important to note the distinction in the use of the terms, search and

browsing, as employed here. The following definitions have been prop-

osed for search, browsing, and exploration 14. In search an explicit goal is

known beforehand. In browsing information is sought without the estab-

lishment of an a priori goal. Browsing is explorative within certain overall

boundaries. Exploration is a form of search for new information where

the information sought is unknown beforehand, but anticipated.

Two kinds of indices can be identified: search indices and browsing indices. The search index is the fillers in the story formalism. These

indices can match to all stories which have concepts and forms relevant to

stated issues. Search is based on matching design issues to find relevant solution principles and design solutions. While search brings us directly to


Figure 4 Browsing in a net-

work of design components

relevant stories, browsing operates in an associative fashion. In browsing,

the indexing system is exploited associatively by the user by activating the

linkages between design concepts within the semantic network. During

browsing, linkages between similar concepts can be activated in order to

find alternative approaches to form which materialize the same design

concept in diverse stories and precedents. The form elements of a solution

are also a lexical index which can be employed to search for other design

precedents employing the same form device, or for browsing.

It appears that matching by concept, or form, indices can potentially

contribute to the rich process of retrieval of design ideas which is characteristic of the human designer's cognitive capabilities. Browsing

through activating the conceptual network of a design domain is illus-

trated in Figure 4.

3.3 C r o s s - c o n t e x t u a l i n d e x i n g in s e a r c h a n d b r o w s i n g

Cross-contextual reminding is a significant phenomenon in design 2. To be reminded of relevant precedents in design requires the recognition of

similarities between current design situations and prior design solutions.

Such matching appears to be dependent upon abstractions, such as the conceptual principles of design solutions which we have identified. Since

these concepts are abstractions they may function as the mechanisms which can transcend specific design domain boundaries. In fact, we may postulate that design concepts appear to be general within a domain and


independent of specific design tasks. Therefore, the body of high-level

concepts in a design domain may support cross-contextual indexing.

Knowledge expansion through experience appears to strengthen these

capabilities in the human designer. Experience enhances the ability to generalize and to establish analogies, both of which are present in the

human designer's ability to make analogical references.

Design stories may be considered as cross-contextual when their design

issues, concepts or forms are similar and shared by other design stories in

different precedents from different design domains. A simple example of

cross-contextual relevance may be illustrated from architectural design. In

museum design, among key factors as interpreted from a functional point

of view are the organization and character of displays, the relationship of

display space to the circulation system, and the control of movement. As

such, the museum may have relevance for other functional building types

such as shopping centres, in which similar issues and concepts are part of

their typological definition.

The memory structure of the system described below is essentially an

associational network of the concepts of the task domain. Terms in the

design concept vocabularies may be common to many stories. This

enables linkages and the operation of cross-contextual search and associa-

tive browsing. Cross-contextual linkages operate through the form-

concept-issue vocabularies to link to relevant design concepts in other

domains. Cross-contextual indexing is illustrated in the examples below.

4 A content-based approach to design knowledge acquisition One approach to the acquisition and formalization of such design

knowledge is to investigate how descriptions are employed to present

precedent knowledge in traditional media. While the holistic design is

presented by graphical illustrations, conceptual points are generally

presented in textual descriptions in an explanatory form. These annotated

sections of text contain chunks of conceptual knowledge which characte-

rize the uniqueness and relevance of a design. These textual annotations are not exhaustive records of a past design, but illustrate generalizations

which underlie a conceptual point, or a significant idea, manifested in a

particular design precedent.

A method for acquiring knowledge from textual annotations has been developed by Schank 7 in the area of social advice. In our work, we have

employed the approach by analysing key words in descriptive and critical

writings relative to the design task. We investigated textual annotations of


15 MItchqdl, W J and McCuI- Iough, M Digital Design Media Van Nostrand Reinhold, New York (1991)

design precedents in current urban art museum designs. Each story was

selected for its description of noteworthy conceptual points, design

rationale, etc. in relation to its relevance to conceptual design of this

building type. Through these annotations the linkage between form,

concept and issue was formulated in each of the design stories which they

describe.

Conceptual descriptions are domain-dependent and generally derived

from the concept vocabularly of the design task. Through analysis of the

textual content of writings on museum designs, stories begin to form a

lexical vocabulary of forms, concepts and issues related to this design task. For example, the vocabulary employed in the stories relates to the

spatial-organizational task in museum design. These may be common to

stories of other design precedents which deal with organizational issues.

5 PRECEDENTS: a precedent-based computer library of designs PRECEDENTS is a library of significant design precedents which pro-

vides relevant ideas for current design. The library is intended to support

the initial stages of architectural design through structured knowledge

from past experience. It is a pilot precedent-based design aid system

which supports search and browsing through cross-contextual indexing by

design issues and design concepts. The domain is the spatial-

organizational task of early architectural design through which a design

concept is established. This is a concept-rich task referred to in the design

literature as concept formation in the design partis. It provides a particu-

larly relevant case study for the potential of the story approach to design

precedent libraries.

The subdomain is the design of urban art museums. A content analysis

was performed in selected works in the architectural literature in order to

identify issues and concepts. The design precedent library contains

museum design precedents, and certain related building types which help

to illustrate the operation of cross-contextual search in a future larger

precedent-base of designs. These examples were selected as representa- tive of current significant designs, and for the range of issues and concepts which they represented.

The pilot system PRECEDENTS is currently developed in a prototype

Hypermedia system for experimentation and demonstration. Hypermedia is widely recognized as a medium supporting associative networks of knowledge 15. Browsing through a structured or open net of information


"1'~ ~ f i l z ~. St~t tB~zt w~i~k w ~ ~.vaxfcc

des iSt .of t~mx~c:t']~d ~ w e ~ ~ o ~ y Lw o ] ~ z d ~ . l x o , ~ ~ t x a t ~iz~ pu 61i e ~ , e e ~ t ~ t w

t ~ ~. : d e : to ]~ov6~ ~ x m-I~a ~ mmtimmiE,]t ~ w e . ~:~ t]w o t ~ ~ z ~j.~la~ • ~ a . ~ d

ao tak:tm. • ~<l~e~mdi~t~z~ l i~ wv /~ cut tb~g

gt~ttS~xt. ~ StJ .dk~ ,~cd~tect )

Figure 5 Story cards

can be supported by the medium. HyperCard was employed as the

software in the current application.

5.1 System organization and operation The system contains three kinds of cards. The issue card, the story card

and the precedent card. To initiate a session in the system, the user

presents a description of the design problem and situation by filling out an

issue card. The card contains scrolling fields in which the user highlights

the appropriate terms for indexing. Generally, the user will fill in the issue

category for problem and context, and search for concepts and forms. The

system searches for relevant stories. These are presented to the user as

story cards (Figure 5).

The user can select the interesting concepts among the stories which were

presented to him for further browsing. If the user wishes to investigate

how related concepts were realized in other building types he may browse

through the library by employing the 'concept ' linking button.

The story cards contain a textual description of the selected story and a

graphical description of a specific related form element. Also included are

link buttons to the precedent card (Figure 6) and other design story cards.

Precedent cards contain the holistic description of the precedent and a set

of buttons. The buttons provide for the linkage between a precedent card

and story cards. The user can investigate how a particular concept was

realized in a specific precedent. He may explore other concepts in that

particular precedent by employing the buttons of the precedent-card. The

user is free to browse within the network of issues, concepts and related

forms.


Figure 6 Precedent card story card

5.2 Search and browsing in PRECEDENTS Browsing in PRECEDENTS can be demonstrated through the example of

the Staatsgalerie in Stuttgart, Germany which was discussed previously.

For example, search by the issue of urban continuity in urban art

museums selects the Staatsgalerie as a relevant precedent. Through

browsing the related concept, 'path-through', of this building a cross-

contextual linkage can be established. This concept occurs in two different

precedents. It appears in the Staatsgalerie by Stirling, and in the well

known Visual Arts Center at Harvard by Le Corbusier. It is found that

this similar design concept is realized in the two buildings by different

form elements. In the Stirling example, by a ramp within the central

court, and by a central walkway in the building by Le Corbusier.

The linkage which has been discovered may help in the discovery of another design concept associated with the form element in the new

precedent. Exposition in the Visual Arts Center by Le Corbusier is provided by an elevated walkway through which the interior functioning

of the building is 'exhibited' in the walkways. Similarly, the concept of 'orientation' indexes a story about the ramp and central court of another


Figure

linkages

7 Cross-contextual

exposition

|

path through

path through

Visual -Arts Center (VAC)

walk-way Le-Corbusier

focal space

Staatsgalerie Museum James Stifling

orientation

|

focal space

1 ramp

Guggenheim Frank Lloyd Wright

precedent, the famous Guggenheim Museum in New York. Cross-

contextual browsing in the conceptual network is illustrated in Figure 7.

Figure 8. illustrates how the knowledge of the system can be expanded. In the example, the concept of 'decorative emphasis' exemplified in the entrance of the Staatsgalerie cues a decorative entrance example in a

different building type, a famous Viennese shop design. This adds another branch to the memory network which was previously illustrated in Figure

7.

6 Conclusions An approach to design aid systems through computer-based libraries of

design precedents has been described. In encoding design knowledge within the system, the device of the design story was proposed as an individual descriptive chunk which is part of the conceptual uniqueness of

a design precedent. It was employed to decompose existing descriptions of complex design precedents into independent chunks of knowledge. The design story was also developed as an effective formalism for encoding relevant design ideas. It enables the representation of complex and rich knowledge by breaking designs into tractable chunks of information. The tripartite scheme of issue-concept-form appears to be an effective

formalism for design knowledge in early conceptual design. It is suitable in early architectural design and may also be relevant to other domains.

The development of domain concept vocabularies in design stories provided issue and concept vocabularies as a basis for the formalization of


e x p o s i t i o n

through

space

o r i e n t a t i o n

focal space

Figure 8 Network expansion

p a t h through

Q

walk-way

Visual-Arts Center Le-Corbusier

Staatsgalerie Museum James Stirling

Guggenheim Frank Lloyd Wright

0e¢o'ya:ti0 nl

• - d e e o r a t t v e . . . . . . . . . . . . . . . • - o : - • . . . . . . . . . . . . . . . . . . . Je.W.el|eJ:y .~o(e, • e.mpn.asls . . . . . . . . . . . . Hans -Hollei-n. ,

design knowledge in a particular domain. A promising approach to the

acquisition of design vocabularies which was presented in our research is

based upon knowledge acquisition through the content analysis of written

descriptions of designs in the literature. Further research in domain

content analysis in various domains is required to develop these vocabu-

laries.

A memory structured around design stories rather than design cases has

been proposed as a means to support browsing and explorative modes of

search. Stories are structured in memory according to a semantic network•

The lexicon of issue-concept-form in the semantic network acts as a

memory index.

The application of Hypermedia to the development of precedent-based

libraries explored the potential of this technology to support browsing and

exploration in libraries. In contrast to a system with highly structured

networks of objects and linkages, we experimented with indexing in order

to support associational search and browsing. The pilot system has

provided a means to explore certain of the theoretical assumptions. A

large-scale system is required to validate these assumptions.


In encoding the relevant conceptual knowledge of prior designs and making the system responsive to the cognitive nature of design thinking, our approach to design support through libraries of design ideas appears to be of promise as one the possible paradigms of computer-based design aid.

7 Acknowledgments This research was supported by the Technion V.P.R. Fund through the E. Deutsch Fund for CAD-CAM research. The author gratefully acknow- ledges the assistance.


Oxman Rivka (1994) Precedents in design: a computational model for the organization of precedent...

Documents

Transcript of Oxman Rivka (1994) Precedents in design: a computational model for the organization of precedent...