You are on page 1of 15

Automation in Construction 8 Ž1999.

395–409

Performance-based design
Yehuda E. Kalay
Department of Architecture, UniÕersity of California, Berkeley, CA 94720, USA

Abstract

Even before Louis Sullivan coined the phrase ‘Form Follows Function,’ architectural researchers have sought, to no
avail, a causal relationship between these two primary constituents of the building enterprise. This paper attempts to explain
why this quest has been futile, and proposes a performance-based design paradigm, instead of the prevailing process-based
paradigms. It suggests that the driving force behind any design activity is the desire to achieve a qualitative solution for a
particular combination of form and function in a specific context. Furthermore, it suggests that quality can only be
determined by a multi-criteria, multi-disciplinary performance eÕaluation, which comprises a weighted sum of several
satisfactionrbehavior functions. The paper develops a performance-based design methodology and demonstrates its
application in an experimental, knowledge-based CAD system. q 1999 Elsevier Science B.V. All rights reserved.

Keywords: Design methods; Design process; Paradigms of design; Design knowledge; Performance evaluation

1. Introduction understand how designers do what they do when


they design. This understanding would lead, it was
The quest for understanding how humans perform hoped, to the development of methods and tools that
complex cognitive activities, such as architectural can help architects and engineers consistently and
ˆ of
and engineering design has been the raison d’etre reliably achieve desired high-quality results. Many
design methods research for the past four decades. approaches have been tried, including psychological,
Behind this quest stands the need to improve the philosophical, and engineering research methods
quality of the built environment, as well as the w1,3,12,18x.
processes of its procurement Ždesign, construction, For the most part, this endeavor has been guided
and management.. Why, then, after four decades of by the Aristotelian notion that design is a process
diligent research and development, we find that that seeks a convergence of form and function: a
buildings are far from perfect in their ability to physical means that can support certain human needs
satisfy all the physical, social, cultural, and eco- or activities, subject to certain conditions and con-
nomic needs of the people who are affected by straints. Following Louis Sullivan’s proclamation that
them? Why, in fact, the more we know about the ‘Form Follows Function’ w27x, most architectural
built environment, the less satisfied we are with our design methods researchers sought a processes-based,
creations? causal relationship between form and function. At
In their quest to affect such desired improve- the core of this quest lay three assumptions: Ž1. that
ments, design method researchers have sought to a physical system’s significant geometrical Žand ma-

0926-5805r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 6 - 5 8 0 5 Ž 9 8 . 0 0 0 8 6 - 7
396 Y.E. Kalayr Automation in Construction 8 (1999) 395–409

terial. properties have some function, or utility; Ž2. coexistence emerges w1x. Thus, while some forms
that one form is more suitable to fulfilling that and functions do exist at the outset of the design
function than other, alternative, forms; and Ž3. that process, neither can be considered the basis for
finding a causal relationship between form and func- seeking the other. Moreover, the existence of intu-
tion will lead to the development of a method, which itive leaps introduces discontinuity in the causality-
can be applied with some assurance of success in based search process, destroying any hope of devel-
every case where a form must be produced that will oping a coherent method that is based on any
optimally facilitate and support a given set of func- monotonous theory.
tional needs. If the prevailing paradigms cannot explain how
Over the past four decades, two different architects work, then how can they form a basis for
paradigms of design have emerged, representing two the development of design tools that purport to assist
fundamentally different approaches to explaining the them? Indeed, practice has shown that current design
causal relationship between form and function. The tools Žwhich are predominantly based on one of the
first, attributed to Simon w24x, attempted to explain two paradigms. force architects into a methodologi-
the process of design as a unique instance of general cal ‘straight jacket’ which they use only when forced
problem-solÕing. It postulated that the designer start to Žwitness, for instance, the limited success of pre-
with the sought function Ži.e., the desired behavior fabricated building systems..
of the system., which is often represented as a set of This paper proposes an alternative approach to the
goals and constraints. The designer then attempt to understanding of the process of architectural design.
discover a form that will support the desired func- It suggests that the quest for design tools must begin
tion, using deductiÕe search strategies. not by exploring how architects design, but rather by
The other paradigm, called puzzle-making, has asking what they do when they design. An account
emerged from the work of researchers like Alexan- of what architects do would stand a better chance to
der w4x and Archea w6x. It postulated that designers be accepted by architects, because it will not purport
begin with a kit of forms Žthat include materials as to describe how each individual pursues the design
well as shape., which are modified and adapted process. On the other hand, this changed research
according to certain rules until they achieve some agenda raises the question: how will such an account
desired functional qualities. This paradigm is based help bring about the sought improvement in the
on inductiÕe reasoning, and has been modeled with design process or its products? The answer lies in
the aid of analogical inferencing methods Žmeta- performance eÕaluation.
phors, symbols, and case studies.. The notion of performance is derived from the
While logically consistent and computationally argument that the relationship between form and
convenient, neither of these two paradigms, nor their function is context-based, rather than causality-
many derivatives and permutations, has gained much based. That is, the performance of a proposed design
favor with architects themselves. When presented solution can only be determined by an interpretive,
with these theoretical paradigms, or better yet—when judgmental evaluation, which considers the form
examined ethnographically under actual conditions Žand other physical attributes. of the proposed solu-
w13x, most architects would not agree that their own tion, the functional objectives Žgoals. it attempts to
design process resemble either one of the two achieve, and the circumstances under which the two
paradigms w22x. They would argue that design, espe- come together. Hence, performance-based design
cially architectural design, is a serendipitous, recognizes that different forms can successfully
‘wicked’ process, replete with uncertainty and dis- achieve similar functions, and that different functions
covery Žoften referred to as the intuitiÕe leap . w21x. can often be afforded by similar forms. In addition, it
Instead of the well-behaved theoretical process, accounts for performance variances of the same
which begins with a statement of forms or functions, formrfunction combinations within different con-
architectural design often begins with an incubation, texts.
introspective phase, followed by iterative refinement The viability of the proposed paradigm, from a
of both form and function until some harmonious practical implementation point of view, depends on
Y.E. Kalayr Automation in Construction 8 (1999) 395–409 397

our ability to represent explicitly, then reason about, 2.1. Problem-solÕing


the desirability of a particular combination of
form–function-context. Such representation differs Problem-solving is a general theory that attempts
from common evaluation and simulation procedures, to explain the cognitive process of creative thinking.
in that it must account for judgment, preferences, as It was first formalized by Simon, Newell and Shaw
well as trade-offs and other subjective measures of in the late 1950s, and implemented in a computer
satisfaction. program called GPS ŽGeneral Problem Solver.. Prob-
In the following, we develop the argument for lem-solving assumes that the desired effects of some
performance-based design, then, introduce a specific intellectual effort can be stated in the form of con-
representation of performance. An experimental straints and goals at the outset of the quest for a
framework that implements both the paradigm and solution to achieve them. To find the solution, the
the performance representation measures serves to problem solver uses a variety of search strategies to
illustrate the theoretical concepts. Initially, we will generate successive candidate solutions and test them
use the terms ‘form,’ ‘function,’ ‘context,’ and ‘per- against the stated goals, until one is found that meets
formance’ loosely, relying on the reader’s intuitive them. The goals, thus, ‘guide’ the search for a
understanding of their meaning. We will define these solution right from the beginning of the problem-
terms more precisely in the second part of the paper, solving process. Problem-solving assumes that set-
where such rigor is needed. ting goals Ži.e., knowing what should be accom-
plished. can be separated from the process of finding
a solution that meets them, and that such knowledge
2. Causality-based design paradigms can be acquired through an independent inquiry
Žanalysis., which should be completed before the
The notion that Form follows Function is derived search for a solution has been initiated w1x. For
from the assumption that a system’s significant geo- example, using this approach, selecting a structural
metrical Žand material. 1 properties have some func- system to span some opening will generally follow
tional utility, and that one form is more suitable for after an analysis of forces, cost, and other character-
fulfilling that function than other, alternative forms. istics of the structure have been determined.
This notion and its inverse Žfunction is derived from Since the characteristics of the problem, accord-
form., has guided architects and engineers for mil- ing to the problem-solving paradigm, are known
lennia. Among its other achievements, this notion prior to commencing the search for the solution
has provided a convenient causal relationship be- itself, its proponents hold that the search for a
tween form and function, the two pillars of architec- ‘satisficing’ 2 solution is goal-directed, and there-
tural design, hence, for developing theories and fore, that means–ends analysis can be employed to
methods intended to assist architects in performing guide the search towards finding the desired solution.
their increasingly more demanding task of finding Thus, the skills that are employed when following
the ‘right’ form–function combination. the problem-solving paradigm are mainly analytical:
Many formal theories that were forwarded over the ability to compare the current ‘state’ of the
the years to explain what architects do have been designed artifact to its desired ‘state’ Žin terms of its
based on this logical foundation. They can be classi- expected utility and behavior., and the ability to
fied into two general groups w2x: Ž1. those that follow draw operational conclusions from this comparison,
Simon’s w24x Problem-solÕing paradigm, and Ž2. so that the differences can be reduced.
those that follow what Archea w6x called Puzzle-mak- Such goal-driven approaches have been computa-
ing. tionally represented as deductive, backward-rea-
soning search strategies, where operators are applied

1 2
In the following, the term FORM will be used to refer to all Meaning ‘good enough.’ The term was coined by Herbert
the physical attributes of objects, including their material composi- Simon in his book Sciences of the Artificial, MIT Press, 1969, pp.
tion, surface finish, etc. 35–36.
398 Y.E. Kalayr Automation in Construction 8 (1999) 395–409

to the goal statement in order to convert it into a set of a given set of components, following a given set
of sub-goals that are easier to solve. This method is of combinatorial rules. Since architects cannot in-
applied recursively until a set of sub-goals that can vent information from scratch in every case, they
be solved without further reduction is found w19x. rely on design ‘cases,’ either from the architect’s
Examples of tools based on this paradigm include own experience or from the experience of the profes-
space allocation programs w7,25x, and a large number sion at large, to provide them with a rich pool of
of evaluation programs, such as way-finding and empirically validated information which has been
energy w15,26x. refined through many years of practice and has
gained society’s or the profession’s approval. This
2.2. Puzzle-making
information comes in the form of proven solutions
The assumption that, in architecture, the charac- w4x, architectural styles, celebrated buildings, estab-
teristics of the desired solution can be formulated lished metaphorical relationships, and recognized
prior to and independently of the search for the symbolisms w29x. How architects adapt this body of
solution that satisfies them was rejected by critics knowledge to the particular problem at hand is not
like Archea w6x and Bijl w8x. They argued that such known—it is the essence of architecture’s celebrated
knowledge cannot exist prior to the search itself, ‘intuitive leap’ and creativity.
since the sought solution is unique, and the process Therefore, rather than rely on a goal-driven strat-
of finding it is characterized by discoÕery and has to egy, the puzzle-making paradigm relies on adapta-
contend with uncertainty. Kim w17x and others have tion of precedents, symbols, and metaphors. The
argued that the brief architects are given by their main skills employed when following this paradigm
clients, which often constitutes the basis for the are synthetic: the ability to compose given parts into
design goals, is much too vague, in most cases, to a new, unique, whole. Such data-driÕen approaches
form a complete goal statement. Rather than use the have been computationally represented as forward-
client’s definition of the desired effects of the sought reasoning search strategies: operators are applied to
building as a complete problem definition, architects the current state of the problem with the aim of
can only use them as a starting point and a catalyst transforming it according to pre-set rules. Example
for the design process, something that provides a of tools based on this approach include generative
sense of direction and a sounding board for potential expert systems, shape grammars, and case-base de-
resolutions. They suggest, instead, that architects sign systems w10,14,20x.
must gradually develop the statement of goals as
they proceed with the design process itself. The
additional information needed to complete the goal 3. Other kinds of relationships between form and
statement must either be inÕented as part of the function
search process, or adapted from generalized prece-
dents, prototypes, and other relevant past experiences
Žso-called ‘design cases’.. Since the relationship be- In this paper, we argue that the relationship be-
tween Form and Function is much more complicated
tween the newly invented information, as well as the
than implied by the causality-based notion of ‘Form
precedents, to the particular needs of the problem
Follows Function,’ and its inverse. Indeed, a particu-
can be discovered only as the problem becomes
lar form often affords many different functions, and
clearer, the adaptation itself is problem-specific and
a similar function is often afforded by many differ-
cannot be accomplished prior to engaging in the
ent forms. The following examples will serve to
search process itself.
illustrate this argument.
Design, according to this view, is a process of
discoÕery, which generates new insights into the
problem. The design search process may, therefore, 3.1. Many forms, same function
be compared to puzzle-making—the search for the
most appropriate effects that can be attained in unique The over-simplicity of the notion ‘Form Follows
spatio-temporal situations through the manipulation Function’ is evidenced by the multitude of different
Y.E. Kalayr Automation in Construction 8 (1999) 395–409 399

forms that essentially were designed to support simi- Andel w5x observed that playgrounds for children
lar functions. Chairs provide one of the best exam- between the ages three and seven perform best if the
ples of different forms that were developed to sup- activities they afford are less structured, in terms of
port exactly the same function Žsitting.. Design com- the equipment they contain. For instance, placing an
petitions, where competitors must respond to the old fire engine in a playground Ža form. will direct
same set of functional requirements within the same the children’s activities towards particular play pat-
context, provide additional evidence that in architec- terns. Furthermore, van Andel observed that this
ture, form does not necessarily follow function. Each particular form tends to create gender-biased play
and every competitor will, invariably, produce a very patterns, which appeal more to boys than to girls. On
different form for exactly the same function. the other hand, a playground that consists mostly of
Some scholars have tried to explain this apparent a sandbox, some rocks, and a few trees or bushes
lack of causality by arguing that, typically, the func- affords less restricted play patterns, and is equally
tional requirements of a building do not tightly con- accessible to both boys and girls. He attributes this
strain its form, thus, leaving the architect with much performance to the creative imagination of the chil-
room to entertain ‘styles’ and other ‘nonpractical’ dren, who can adapt the existing, generic forms into
considerations. Herbert Simon, for example, has de- particular needs, such playing games like ‘house,’
fined style as ‘one way of doing things,’ chosen ‘cops and robbers,’ or the landing of an alien space-
from a number of alternative ways w23x. Since design ship.
problems generally do not have unique or optimal Another example of architectural multi-purpose
solutions, says Simon, style can be used to select a Ži.e., functional. spaces has been described by Eliza-
solution from among several functionally equiÕalent beth Cromley in her paper on the history and evolu-
alternatives, just as any other criteria can. He offers tion of modern bedrooms w11x. In addition to provid-
the following analogy: ing a place for sleeping, bedrooms, through the
‘‘Mushrooms can be found in many places in the seventeenth century, also functioned as parlors, din-
forest, and the time it takes us to fill a sack with ing rooms, and as places for entertaining guests. In
them may not depend much on the direction we the eighteenth century, the function of bedrooms
wonder. We may feel free, then, to exercise some became more focused, as a place for sleeping and
choice of path, and even to introduce additional dressing, for quiet retirement, and for socializing
choice criteria . . . over and above the pragmatic one with close friends and family members. In the nine-
of bringing back a full sack Žof mushrooms.’’. teenth century, bedrooms became a place to occupy
Most architects, however, would reject this notion only at night. In the 20th century, the definition of
that form is the result of less ‘practical’ functional their function was broadened again, especially as far
considerations than other aspects of the building, and as children’s bedrooms are concerned. Today, such
therefore, an afterthought, something to be contem- functions include sleeping, doing homework, read-
plated only when all the other ‘important’ aspects of ing, and playing with friends. Bedrooms for the
the design have been dealt with. Rather, they would adults Žthe so-called ‘Master Bedroom’., have turned
argue, that it is something a competent architect will into ‘suites,’ which include full bathrooms, dressing
consider before, during, and after the development rooms, and walk-in closets. They often serve as
of solutions satisfying the functional needs. More- home-offices, gyms, and entertainment centers.
over, the two issues cannot be separated, since each The ability of the same form to afford different
one informs the other, and influences its develop- functions is further demonstrated by what we now
ment. call adaptiÕe re-use. The term designates the conver-
sion of older buildings to meet modern needs. It is
3.2. Many functions, same form rooted in the economic realities of the late 20th
century, and the growing need for urban renewal and
The notion that a given form can support many rehabilitation. This trend is characterized by corpora-
different functions is demonstrated well by designs tions, shops, and even residential units moving into
of playgrounds, parks, and civic plazas. Joost van older buildings in the core of cities. Rather than tear
400 Y.E. Kalayr Automation in Construction 8 (1999) 395–409

down a building which may have some historical or


cultural significance, new tenants may rehabilitate it
while preserving its character. A typical case in point
is Hayes Hall, in Buffalo, NY. Built circa 1865, this
landmark building served as a poorhouse and a
lunatic asylum until 1893, when it became a county
hospital. In 1909, it was acquired by the University
of Buffalo, and served as the office of the president
until the new campus was built in 1968, when it
became the School of Architecture and Planning of
the State University of New York at Buffalo.

3.3. Other kinds of form–function relationships Fig. 1. Performance, as a measure of the confluence of Form,
Function and Context.

Peter Eisneman’s structuralist approach to build-


ings, which derives from his own interpretation of serving, measuring, and interpreting this behavior,
Noam Chomsky’s linguistic theories Žas well as we can assess the performance of the solution.
Jacques Derrida and other philosophers., demon- Performance evaluation is intended, therefore, to as-
strates well the complexity of the relationships be- sess the desirability of the behavior of the confluence
tween form and function, as depicted in his design of of the form, function and context. It may reveal, for
House X, which is based on a series of geometrical example, that a particular form is capable of support-
transformations on a cube. ing a certain functional need in a particular context,
in which case, it will be deemed ‘successful.’ On the
3.4. The importance of context other hand, it may reveal a need to modify the form
to meet the desired function in the particular context,
The form of a building also depends upon the or to modify the desired function to meet the ones
physical, cultural, social, and other contexts in which afforded by that form in that particular context.
it is embedded, at least as much as it depends on the Designing, accordingly, can be considered as an
function it must serve. The form of the Sydney opera iterative process of exploration, where desired func-
house is an example of a form derived from the tional traits are defined, forms are proposed, and a
physical context of the building Žthe Sydney harbor., process of evaluation is used to determine the desir-
as much as from its function Ža symphony hall.. ability of the confluence of forms and functions
Likewise, the shape of Le Corbusier’s Ronchamp within the given context ŽFig. 2.. The process termi-
Chapel has been derived from its spiritual context, as nates when the designer finds a form that fulfills the
much as from its functional and physical site consid- function, or is satisfied by the functionalities af-
erations; and Gerrit Rietveld’s colorful Schroder ¨ forded by the chosen form, within the given context.
House in Utrecht, The Netherlands Ž1931., has been We call this condition functional adequacy: the
shaped as much by the neoclassicist cultural ideas of instance when form and function come together to
the De Stijl movement to which he belonged, to- achieve acceptable performance within a given con-
gether with painters like Theo van Doesburg and Piet text w9x.
Mondrian, as much as by functional requirements.
4.1. The notion of performance

4. Performance-based design We suggest that this description of design leads to


a different paradigm than either problem-solving or
The position taken in this paper is that Form, puzzle-making. We call it performance-based de-
Function and Context combine to determine the sign. As stated earlier, we consider performance to
behaÕior of the proposed solution ŽFig. 1.. By ob- be a measure of the desirability of the confluence
Y.E. Kalayr Automation in Construction 8 (1999) 395–409 401

Fig. 2. Design as a bi-directional exploration of a Form–Function-Context composition.

form and function within a given context Žwhich, in they measure the behavior of some aspects of the
turn, we call ‘behaÕior’.. ‘Desirability,’ however, is designed system, such as cost, or noise level. On the
a fuzzy and subjective measure. To deal with this other, they measure the degree of satisfaction each
fuzziness, we offer the concept of satisfaction func- behavior value elicits in the client. Each point on
tions. Satisfaction curves were first introduced by every curve denotes the performance of the form–
Kunz and Rittel in the 1970s, and used by Mahdavi function-context combination with regard to some
in his SEMPER programme ŽAutomation in Con- measure Že.g., cost..
struction 6Ž2.: 353–373.. These are mappings that The curves demonstrate several phenomena com-
express the specific relationship between the behav- monly associated with satisfaction. Fig. 3b, for ex-
ior of a system and the subjective measure of its ample, demonstrates that the client may generally be
desirability under specific circumstances. Fig. 3 de- satisfied with the behavior of the system, until its
picts several typical satisfaction curves: on one axis, behavior in some area reaches a certain threshold.

Fig. 3. Some typical satisfaction curves.


402 Y.E. Kalayr Automation in Construction 8 (1999) 395–409

Then, satisfaction diminishes, but the change from afford are expressed as numerical values, each of
100% Žcompletely satisfied. to 0% Žnot satisfied., is which expresses the client’s satisfaction with respect
gradual. The curves allow for such notions as ‘quite to one specific behavior. To aggregate the separate
satisfied,’ ‘more or less satisfied,’ or ‘barely satis- satisfaction curves into one composite measure of
fied,’ to be expressed. The slopes of the curves allow performance, we can add them up. But since differ-
us to express the rate of change: the steeper the ent behaviors weigh differently in the overall perfor-
slope, the more abrupt the change, which means that mance measure, we must first assign to each of them
once the threshold has been reached, a small change a relative weight. This method is well-established,
in the system’s behavior will result in satisfaction or and has been used by other researchers to develop
dissatisfaction. On the other hand, a shallow slope aggregates of multi-criteria evaluations w30x. The
indicates a wider latitude in satisfying the client, composite result of the summation of weighted, nor-
which allows more room for trade-offs with other malized satisfactions is presented to the client as the
satisfaction curves that may need to be modified. overall performance of a given design solution.
The satisfaction curves must, of course, be set by
the client, or by the designer. They are unary func-
tions, in the sense that each curve pertains to satis- 4.2. Trade-offs
faction derived from one behavior only. This makes
it possible to set them individually. For example, the Trade-offs are the hallmark of every design activ-
client may state that his budget for building a single ity. Typically, all the functional needs of a building
family house is US$300,000. Using the satisfaction cannot be satisfied by any one design solution. The
curve depicted in Fig. 3c, he may develop a function achievement of certain needs often must come at the
of the kind shown in Fig. 4. It shows that the client expense of other needs. For instance, eliminating
will be most satisfied if the building costs windows on the west side of a building to save
US$300,000. He will not be satisfied at all if the energy might also deprive the inhabitants of a fabu-
building costs over US$315,000, or less than lous view. Hence, the degree of satisfying some
US$270,000. The curve also shows that in the vicin- needs may have to be compromised, so that others
ity of US$300,000, say "US$3,000, his satisfaction are also satisfied. But how much should any one
is virtually unchanged. The different slopes of the need to be compromised? The satisfaction functions
rising and diminishing parts of the curve show that also facilitate this often difficult decision-making
there is more latitude in satisfying the client’s bud- process, in three ways:
getary needs under US$300,000 than there is over 1. by explicitly showing how well any one need is
US$300,000. being satisfied, as a percentage between full and
Similar satisfaction functions can be developed zero satisfaction;
for each aspect of the building. The mappings they 2. by expressing the tolerance for satisfying the
expressed need, in terms of the steepness of the
curve; and
3. by prioritizing the relative importance of each
need, in terms of the weight assigned to it.
Using these three measures, it is possible to iden-
tify needs that are not being satisfied, and those that
are over-satisfied. It is possible, therefore, to seek a
design solution that better achieves the under-satis-
fied needs, while achieving less-well the over-satis-
fied needs. In fact, an algorithm can be developed
that provides hints to the designer, indicating possi-
ble trade-offs. It first identifies the under-satisfied
needs, then the over-satisfied ones. Among the
Fig. 4. A satisfaction curve expressing building cost behavior. over-satisfied needs, it would suggest that those of
Y.E. Kalayr Automation in Construction 8 (1999) 395–409 403

Fig. 5. The general multidisciplinary, collaborative design environment.

lower importance Žas expressed by their associated algorithm cannot tell which specific need ought to be
weights. would be candidates for reduced-satisfac- compromised to achieve another need. Such advise
tion. It will also indicate how much latitude exists in could be added through a knowledge base, which
reducing their satisfaction levels. stores rules about the relationships between the vari-
Given that the inter-relationships between the dif- ous needs. It might also store specific suggestions for
ferent needs are not obvious, for the most part, the improving under-satisfied needs. Nonetheless, only a

Fig. 6. The structure of an Intelligent Design Assistant ŽIDeA..


404 Y.E. Kalayr Automation in Construction 8 (1999) 395–409

complete new design solution can, in general, make making design changes to achieve the sought degree
all the necessary adjustments. of satisfaction, in case it has identified under-satis-
fied needs.
The Fenestration IDeA has been developed by
5. A case study Gustavo Llavaneras, as part of a larger research
project, which aims to develop a multidisciplinary,
To test the validity of the proposed paradigm, we collaborative design environment w16x. This environ-
have implemented it in a test program that operates ment comprises several components, including a
in the domain of windows. The program, called The Project Database ŽPDB. for storing the evolving,
Fenestration IDeA, is a design agent capable of three project-specific design information, and several Ob-
actions: Ž1. it provides a Žsimple. environment for jects Database ŽODBs. that store object-specific, but
developing a set of functional requirements Žneeds. project-independent data ŽFig. 5.. Intelligent Design
and design solutions that attempt to meet the stated Assistants ŽIDeAs. are the means used to interact
needs; Ž2. it evaluates the performance of proposed with these databases, while using their expertise in
solutions Žin terms of fenestration only., using five different fields to actively assist the designers. The
criteria Ždaylighting, sound transmission, ventilation, IDeAs may also call upon external evaluation tools,
views, and budget.; and Ž3. it provides advise for and may be composed of other, more specialized

Fig. 7. Setting desired satisfaction levels.


Y.E. Kalayr Automation in Construction 8 (1999) 395–409 405

IDeAs. Each IDeA is a goal-based agent, comprising and private residences, as well as on the locality in
an ActionrDecision system which stores its task- which they are being built. Hence, the IDeA first
specific rules, a task-specific database, and a group asks the designer to choose the domain of his work
of performance predictors and evaluators ŽFig. 6.. Žschools, office buildings, residences, etc.., and the
The Fenestration IDeA implements the proposed location of the project. These inputs are used to
paradigm, in a simplified manner. It represents ex- select the pertinent knowledge bases, and represent,
plicitly function and form, as well as the context of in their own right, the context of the project.
the particular design project. Function is limited to Once the Context and the Function have been
the five criteria listed earlier. The designer can set specified, the Fenestration IDeA provides the de-
the desired satisfaction levels for each function, us- signer with the means to design a room with its
ing sliders ŽFig. 7.. The IDeA verifies that the windows ŽFig. 8.. Again, assistance is provided in
satisfaction levels set by the designer are within terms of verifying code compliance for minimal
acceptable building code limits, if such codes exist dimensions, as well as other aspects Že.g., if the
Že.g., for daylighting and ventilation.. ‘Acceptable designer has indicated that a wall is not an external
limits’ are, in turn, dependent upon the overall func- wall, the IDeA will not let him put a window in that
tion of the design: they differ for classrooms, offices, wall..

Fig. 8. Designing the room and its windows.


406 Y.E. Kalayr Automation in Construction 8 (1999) 395–409

Fig. 9. Linear piece-wise approximated noise-reduction satisfaction function.

Once the Form has, thus, been specified, the and


Fenestration IDeA performs the necessary tests to
y2
predict the behavior of the form–function-context
combination, and assess its performance. For the MLS Ž DesL y MaxL . y DesL Ž 100 y MLS.
sake of simplicity, the Fenestration IDeA approxi-
mates the satisfaction curves in a piece-wise linear
s
ž Ž DesL y MaxL. /
form. For instance, the noise reduction satisfaction Ž 100 y MLS.
curve is approximated using the functions depicted
in Fig. 9, as expressed by the following equations:
q
ž DesL y MaxL / x.

The overall performance is calculated and pre-


sented in numerical and visual forms, as depicted in
°0 if x - MinL ¶ Fig. 10.
y1~ if MinL F x F DesL • The Fenestration IDeA has not yet progressed to
f Ž x. s ,
y2 if DesL - x F MaxL the advise-giving level. It is envisioned, however,
0
¢ if x ) MaxL
ß that such advise will be provided using the method
outlined earlier: the system will identify the least-
satisfied functions, and the ones that are well-satis-
with fied yet have some latitude in lowering their level of
satisfaction. Then, using the task-specific knowl-
MLS Ž MinLy DesL . y MinL Ž MLS y 100 . edge-bases available to it, the system could identify
y1 s
ž Ž MinL y DesL. / strategies for satisfying the under-satisfied functions.
For example, if the noise-reduction function is not
satisfied, but there is some room for reducing the
MLS y 100
satisfaction of the budget requirement, the Fenestra-
q ž MinLy DesL / x
tion IDeA may suggest using a more expensive
Y.E. Kalayr Automation in Construction 8 (1999) 395–409 407

Fig. 10. Several different ways for presenting the overall performance.
408 Y.E. Kalayr Automation in Construction 8 (1999) 395–409

triple-glazed window, instead of the selected two- the computation of performance, satisfaction func-
pane window. Likewise, if the noise reduction func- tions were introduced. These functions allow for
tion is satisfied, but the budget has been exceeded, it mapping a given behavior onto measures of satisfac-
would recommend trying a cheaper two-pane glazing tion. They also facilitate trade-offs, a necessary
instead of the selected three-pane. means to improve the overall performance of a sys-
tem by sacrificing the degree of satisfaction from
some parts of the system in favor of others.
6. Conclusion The proposed design paradigm fits well within
our view that computers ought to be partners in the
The development of computational tools that can design process, tools the designer can draw upon
truly assist humans in performing complex activities when developing forms, specifying functions, and
such as architectural design relies upon developing a interpreting their confluence w28x. This approach,
deep understanding of the process that is to be along with the proposed paradigm, have been tested
assisted, and on casting this understanding into a through the development of an experimental system
model that can be represented explicitly Žand thus intended to support the design of windows in a
can be translated into a computer program.. Having building. The so-called Fenestration IDeA has been
identified the two main characteristics of architecture implemented in Visual Basic 4.0. While it is not yet
as Form and Function, the search for formal theories complete, we believe it already demonstrates well
that can explain the process of design tended to the issues underlying the Performance-based design
converge on causality-based paradigms. Hence, the paradigm.
attractivity of statements such as ‘Form follows
Function.’ This statement provided a convenient log-
ical foundation for design theories, much like other Acknowledgements
causalities have formed the foundation of many engi-
neering and practically all scientific paradigms. The author wishes to thank Gustavo Llavaneras, a
Many architects found this logically-convenient PhD student in the Department of Architecture at UC
statement inadequate to describe what their experi- Berkeley, whose dissertation work helped developed
ences taught them, for it failed to account for the the paradigm, and who has been implementing it
discontinuity in the relationship between form and through the Fenestration IDeA that was presented in
function, which architects call ‘the intuitive leap.’ the paper. Thanks are also due to Professor Carlo
This leap occurs when architects, engaged in the ´
Sequin from the Department of Computer Science at
search for a form that will facilitate some desired Berkeley, for his insight and assistance in developing
function, actually find the ‘right’ form. The paradigm the Satisfaction Curves.
presented in this paper attempts to recognize this
experience, and use it as a basis for an alternative
formal model of design, which can be implemented References
by computational means. It does not attempt to
formalize the intuitive leap itself, only to accommo- w1x O. Akin, How do architects design?, in: Latombe ŽEd..,
date it in the model. This accommodation takes the Artificial Intelligence and Pattern Recognition in Computer-
form of contextuality: the convergence of form and Aided Design, IFIP, North-Holland, New York, NY, 1978.
w2x Y. Aksoylu, Two different systematic approaches to design,
function in a particular context. The paradigm strives Technical report, University of California, Berkeley, CA,
to eliminate the precedence of either form or func- 1982.
tion and, hence, of the causal relationship between w3x C. Alexander, Notes on the Synthesis of Form, Harvard
the two. To compensate, it develops the notion of Univ. Press, Cambridge, MA, 1964.
w4x C. Alexander, S. Ishikawa, M. Silverstein, M. Jacobson, I.
performance, as a means for interpreting and deter-
Fiksdahl-King, S. Angel, A Pattern Language, Oxford Univ.
mining the confluence of the two entities. Press, 1977.
Performance is a measure of the desirability of the w5x J. van Andel, Expert systems in environmental psychology,
predicted behavior of a design solution. To facilitate JAPS 10 conference, Delft, The Netherlands, 1988.
Y.E. Kalayr Automation in Construction 8 (1999) 395–409 409

w6x J. Archea, Puzzle-making: what architects do when no one is w18x J.C. Jones, Design Methods, Wiley, London, 1980.
looking, in: Y.E. Kalay ŽEd.., Computability of Design, w19x J. Laird, P. Rosenbloom, A. Newell, Universal Subgoaling
Wiley-Interscience, New York, 1987. and Chunking, Kluwer Academic Publ., Boston, 1986.
w7x G.C. Armour, E.S. Buffa, A heuristic algorithm and simula- w20x R. Oxman, Multiple operative and interactive modes in
tion approach to relative location of facilities, Management knowledge-based design systems, in: Y.E. Kalay ŽEd.., Eval-
Science 9 Ž2. Ž1968. 294–309. uating and Predicting Design Performance, Wiley-Intersci-
w8x A. Bijl, An approach to design theory, in: Yoshikawa, War- ence, New York, 1992.
man ŽEds.., Design Theory in CAD, North-Holland, Amster- w21x R.B. Norman, Intuitive design and computation, in: Y.E.
dam, 1987. Kalay ŽEd.., Computability of Design, Wiley-Interscience,
w9x G. Carrara, Y.E. Kalay, G. Novembri, Knowledge-based New York, 1987.
computational support for architectural design, Automation w22x H.W. Rittel, M.M. Webber, Planning problems are wicked
in Construction 3 Ž2–3. Ž1994. 123–142. problems, in: Cross ŽEd.., Developments in Design Method-
w10x R.D. Coyne, M.A. Rosenman, A.D. Radford, M. Balachan- ology, Wiley, New York, NY, 1984.
dran, J.S. Gero, Knowledge-based Design Systems, w23x H.A. Simon, Style in design, in: Eastman ŽEd.., Spatial
Addison-Wesley, Reading, MA, 1990. Synthesis in Computer-Aided Design, Wiley, NY, 1975.
w11x E.C. Cromley, Sleeping around: a history of American beds w24x H.A. Simon, Models of Thought, Yale Univ. Press, New
and bedrooms, Journal of Design History 3 Ž1. Ž1990. 1–17. Haven, CT, 1979.
w12x N. Cross, The Automated Architect, Pion Press, London, w25x E. Shaviv, D. Gali, A model for space allocation in complex
1977. buildings, Build International 7 Ž6. Ž1974. 493–518.
w13x D. Cuff, Architecture: The Story of Practice, MIT Press, w26x E. Shaviv, Y.E. Kalay, Combined procedural and heuristic
Cambridge, MA, 1991. method to energy-conscious building design and evaluation,
w14x U. Flemming, Case-based design in the SEED system, in: G. in: Y.E. Kalay ŽEd.., Evaluating and Predicting Design Per-
Carrara, Y.E. Kalay ŽEds.., Knowledge-based Computer- formance, Wiley-Interscience, New York, 1992.
Aided Architectural Design, Elsevier, Amsterdam, 1994. w27x L.H. Sullivan, Kindergarten chats on architecture, in: C.F.
w15x M.D. Gross, C. Zimring, Predicting way-finding behavior in Bragdon ŽEd.., Education and Democracy, Scarab Fraternity
buildings: a schema-based approach, in: Y.E. Kalay ŽEd.., Press, 1934.
Evaluating and Predicting Design Performance, Wiley-Inter- w28x L.M. Swerdloff, Y.E. Kalay, A partnership approach to
science, New York, 1992. computer-aided design, in: Y.E. Kalay ŽEd.., Computability
w16x Y.E. Kalay, L. Khemlani, J.W. Choi, An integrated model to of Design, Wiley, New York, 1987.
support collaborative multi-disciplinary design of buildings, w29x R. Venturi, D. Scott-Brown, S. Izenour, Learning from Las
in: O. Akin ŽEd.., Proceedings of the Descriptive Models of Vegas, MIT Press, Cambridge, MA, 1972.
Design, Istanbul, Turkey, 1996. w30x A. Wizel, R. Becker, Integration of performance evaluation
w17x M.K. Kim, Development of machine intelligence for infer- in computer-aided design, in: Y.E. Kalay ŽEd.., Evaluating
ence of design intent implicit in design specifications, in: and Predicting Design Performance, Wiley-Interscience, New
Y.E. Kalay ŽEd.., Computability of Design, Wiley-Intersci- York, 1992.
ence, New York, 1987.

You might also like