See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/30871789

From Design Concepts to Design Descriptions

Article  in  International Journal of Architectural Computing · September 2008

DOI: 10.1260/1478-0771.6.3.335 · Source: OAI

CITATIONS READS

2 126

1 author:

Sotirios Kotsopoulos
Massachusetts Institute of Technology
40 PUBLICATIONS   76 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The Connected Sustainable Home View project

All content following this page was uploaded by Sotirios Kotsopoulos on 13 March 2015.

The user has requested enhancement of the downloaded file.

From Design Concepts to
Design Descriptions
Sotirios D. Kotsopoulos

international journal of architectural computing issue 03, volume 06 335

 From Design Concepts to Design Descriptions
Sotirios D. Kotsopoulos

The paper examines the process of articulation and

development of design concepts from a computational
standpoint.The context of the research is the
architectural studio and the process of designing from
scratch.The scope of the research is educational.
Shape grammar formalism is used in a retrospective
analysis, to show how the concept of “porosity” was
used by architect Steven Holl and his team in
designing Simmons Hall, a 350-unit student residence,
at MIT.

336
1. INTRODUCTION
“I depend entirely on concept diagrams, I consider them my secret
weapon.They allow me to move afresh from one project to the next,
from one site to the next.” [1]
Architect Steven Holl acknowledges his dependence on open-ended
conceptual frames rather than on the existing building morphologies and
typologies. Holl disregards any fixed architectural vocabulary in favor of an
“initial concept” capturing the essence of design possibilities that he
considers unique for each project. For Holl, a concept more than just a
verbally expressed idea sets a “manifold relation” among the site, the
geometry, the program, the circumstance, the materials etc.
For many architects and designers, design concepts play a key role in the
development of innovative design solutions. Even though designers do not
make a sharp distinction between the process of production and the
process of interpretation of designs, an “intended” interpretation usually
guides their actions in the studio. Early conceptual schemes are used to
frame a particular design approach. In this paper, a method for generating
architectural form from design concepts is suggested.The method is based
on visual productions rules that generate design descriptions.The
production rules are expressed by means of shape grammar formalism.The
presented paradigm is a retrospective demonstration of how porosity, a
concept transferred from medicine, biology and organic chemistry, was used
by architect Holl and his team in designing Simmons Hall, a 350-unit student
residence, at the Massachusetts Institute of Technology.
The novel aspect of the paper is to show how a design concept can be
treated by formal-generative means to produce design descriptions “from
scratch”. Background assumption and motivation of this study was that a
design concept is at its root a course of action meant to be performed by
the designers in the studio. Setting a design concept is equal to
distinguishing the identity of particular design events. It is proposed that
there is no more to framing a design concept than there is to grasping a
number of grammatical transformation rules. But, setting forth a design
concept is not equal to laying down instructions. It depends on “framing
ideas” of particular design activities and coping with spatial relations and
their features.The paper shows that the productive contribution of early
conceptual schemes in design can be enhanced by formal-generative means,
in three ways: First, by making their description explicit; second, by leading
to the implementation of computational devices with strong generative
capacity; and third, by making them available for future reference.The
descriptive task involves the mapping of the actions implied by a design
concept with the aid of rule schemata and rules.The productive task
involves the implementation of the rules in shape grammars and computer
programs.The reference task involves the assemblage of custom
computational tools and data structures that can be retrieved by future

From Design Concepts to Design Descriptions 337

 users. In extension of the above, it is proposed that design concepts can
mediate between intuition and computation. Against the temptation of
“developing a computer program first, and then see what happens”, a
conceptual scheme assists in framing patterns of activity within a specific
context.This is important in architectural design, where the distinction
between abstract problem solving methods and case-specific techniques
cannot be as concrete as in other branches of engineering. Intuitions about
associations with often ill-defined, but familiar, concepts have to be
consulted frequently to assure that one is not dealing with fake issues.

2. BACKGROUND
Engineers, design theorists, and researchers of Artificial Intelligence have
thoroughly examined the use of concepts in design. David Ullman, for
example, considers the formation of design concepts in the context of
mechanical engineering, in designing or re-designing devices with specific
functionality. In recent years, analogous views became increasingly popular
among architects and designers. Ullman [2] defines a design concept as “an
idea that is sufficiently developed to evaluate the principles that govern its
behavior”. A key feature of Ullman’s approach is the generation and
evaluation of multiple concepts for the same design task. Concept
generation involves two steps: a) functional decomposition, and b) concept
generation from functions. Functional decomposition involves breaking
down the needed function of a device, as finely as possible and with as few
assumptions about form as possible. Concept generation happens through
the listing of several conceptual ideas for each function. Conceptual ideas
derive from the designer’s own expertise, enhanced through pattern search,
reference books, brainstorming etc.
Donald Schön [3] proposes the displacement of concepts as a unifying
principle in terms of which innovation in technical discovery and in theories
may be explained as manifestation of a single process. Schön’s view evolves
in relation to metaphor, analogy and comparison.The displacement of
concepts has a “radical” character, in that old concepts can be used as a
projective model for new situations and a “conservative” character, in that
old assumptions may be transposed in a covert or non-critical way to fresh
contexts. Schön [4] approaches design as a situated activity in which
designers seek to comprehend and “frame” a problem. In their effort, the
designers initiate a reflective conversation with the problem, involving action
and reflection on the consequences.This reflective, bi-directional process,
leads to the formation of new meanings and to the reframing of the
problems.
John Gero examines the formation of design concepts in the making of
design descriptions by means of computational models developed in
Artificial Intelligence. Gero [5] uses implemented examples drawn from the
genetic engineering of evolutionary systems to show that the formation of

338 Sotirios D. Kotsopoulos

 design concepts is based on the emergence of patterns in the available
design representations. A key feature of Gero’s approach is that the
emergent patterns form the basis of new concepts: they are memorized to
become a repertoire of known patterns that remain available for future use.
Along the same lines, Suwa, Gero, and Purcell [6] examine how early
sketches are essential for the formation of design concepts.The researchers
use protocol analysis in studying the design process of a practicing architect,
to show that the early sketches serve as a record and as a provider of
visual cues for association of non-visual information, but also as a physical
setting in which design thoughts are constructed.The authors propose that
the inspection of early sketches forms the basis for the invention of novel
design issues. New concepts emerge out of inspection of the existing
information, or as a result of action upon previously produced visual
material. Novel design issues may include design goals derived from explicit
knowledge, extensions of previous goals, goals aiming to resolve specific
conflicts, or unsupported goals.
Like in Schön [3], [4], Gero [5], and Suwa, Gero and Purcell [6] the
present study refers to creative design from “scratch”, as opposed to re-
design.The focus is on architectural design, as opposed to design in
mechanical engineering. Unlike Gero [5] and, Suwa, Gero and Purcell [6] this
study is based on the observation that architects introduce concepts that
do not necessarily emerge out of the existing design representations. On
the contrary, they can be imports from domains foreign to design, like
biology, chemistry, physics, mathematics, music, etc. Just as proposed in
Schön [3], these concepts are transposed in a non-critical way to design
contexts, with a view to focus the designers’ attention on particular features
of a problem and to propagate a course of action. In this paper, shape
formalism is used to model this action.
Shape formalism, as defined in Stiny and Gipps [7] and developed in
consecutive papers, extends the formalism of production systems and
generative grammars in modeling the interaction of spatial forms. Points,
lines, planes and solids are classified in sets according to their spatial
dimension, they are ordered with the part relation (≤) and they are
organized into shape algebras, where operations like addition, subtraction
and product can be used to perform spatial calculations. For example, the
shape algebra U12 contains 1-dimensional elements – lines – that are
manipulated on the 2-dimensional plane.

Figure 1. Examples of forms in the

algebra U12

Accordingly, the shape algebra U13 contains lines that are manipulated in the
3-dimensional space and the shape algebra U33 contains solids that are

From Design Concepts to Design Descriptions 339

 manipulated in 3-dimensional space. Product algebras can be formed as
combinations of shape algebras to provide more inclusive environments for
calculation. For example, the product algebra U13×U33 contains lines and
solids, manipulated in 3-dimensional space.

 Figure 2. Examples of forms in U13,

U33 and U13×U33 algebras (the 3-d
space is in this case virtual)

Within the context of shape formalism, the spatial elements are composed
with the aid of production rules.The rules can be used for computer
implementation or to construct computational systems of generative and
explanatory capacity known as shape grammars. A comprehensive
presentation of shape formalism is presented in Stiny [8]. A discussion on
the dual character (creative–expressive) of spatial rule systems can be found
in Knight [9], while a discussion on how grammars may incorporate physical
design-reasoning can be found in Sass [10]. Representative experiments in
the use of rules in design synthesis can be found in Celani [11], Duarte [12],
Kotsopoulos [13], and in Knight and Sass [14]. But, the strengths of formal
composition have not been adequately explored in designing from “scratch”.
This paper is an attempt to compensate this shortage. It shows how rules
can be useful to describe the regular and productive early design processes
and their suppositions.

3. METHODOLOGY
A design problem is described in terms of observation, past experience and
on the basis of properties that are empirically ascertained. However, the
scheme that provides the means to move from known to novel solutions
cannot be contrived in terms of the existing representations alone. A
designer has to provide new hypotheses that establish new productive
connections among the available data and an interpretation for the network
of their relationships. Hypotheses are employed in science and in design
with different objectives at view. March [15] observes that a scientific
hypothesis is of universal character: it seeks to predict all future
occurrences of a phenomenon and to give account for its possible causes. A
design hypothesis, on the contrary, is of existential character: it intends to
produce at least one successful solution in response to a problem. Science
aims at general laws while design at case specific results. Accordingly, a
scientific hypothesis aims at being predictive while a design hypothesis at
being productive.

340 Sotirios D. Kotsopoulos

 Hypotheses are associated with the introduction of concepts. A concept
singles out a property, a relation, or an action by setting out a name, or a
scheme. For example:
Pore = minute opening.
A concept can be introduced contextually by a list of synonyms that explain
it.The explanation becomes in this case a creative medium as it may suggest
new meanings. Design concepts are introduced contextually and in parallel
to a course of action that is organized and explained in terms of them.The
meaning of a design concept becomes its use: interpreting the output of the
action confers meaning on the concept.
The designers proceed from tentative constructions to definite ones.
They make their way towards case specific results by gradually adapting
their general conceptual schemes to the given contexts.This progression
can obtain formal expression.Within a formal system, it is analogous to
moving from rule schemata to rules. In general, a rule of the form x → y
denotes the action “x is substituted by y”. It specifies that provided a
condition x, a conclusion y is generated.That is, an objective is accomplished
(y is produced) provided that a condition is satisfied (x is found).Table 1
shows how a subtraction of one prismatic solid from another can be
expressed by a parametric rule in the shape algebra U33.

Table 1. Example of a shape

rule: expression, condition and rule expression
conclusion x→ y

rule condition
x

rule conclusion
y

When the rules of a formal system are potentially infinite it is impossible to

state all of them one by one.Therefore, they are indicated by one or more
syntactic statements. Such statements may be seen as rules with an empty
class of premises that are able to introduce other rules.The expression
(∀x) (∀y) g(x)→ g(y) denotes a rule schema. Rule schemata determine rules
each time the variables x, y are substituted by specific instances. An
assignment g, determined by a predicate, states the attributes of x and y in a
general way. In a rule referring to spatial forms the assignment g may specify
a certain subclass of shapes, i.e.:
g: “x, y are prisms ∧ the faces of x, y are rectangles”
The expression (∀x) (∀y) g(x)→ g(y) is noted g(x)→ g(y) or simply x → y,
while the assignment g is usually provided as a short verbal description. Rule
schemata are formal generalizations of rules.They specify a certain

From Design Concepts to Design Descriptions 341

 treatment for an entire family of forms, instead of just for specific instances.
For example, the shape rule illustrated in Table 1 becomes the expression of
a rule schema after providing the above mentioned assignment g, which
specifies a general family of shapes that this rule can apply.
Two are the key aspects of its application, “recognition” and
“modification”. As shown in Stiny [8], a shape rule schema applies on a
description C in two steps: First, a transformation t “recognizes” some part
of C geometrically similar to the shape g(x). Second, the same
transformation t is used to “modify” C: it substitutes t(g(x)) with t(g(y)).
Concisely: C' = [C – t(g(x))] + t(g(y)).

C t(g(x))<C C-t(g(x))
 Figure 3. Example of applying the
rule of Table 1, in the shape algebra
U33

Figure 4 presents a sequence of shapes produced after applying the rule

schema of the Table 1, on an initial shape C, in the shape algebra U33. Such a
sequence of shapes is called a derivation.

C
 Figure 4. Example of a possible
derivation in the shape algebra U33

⇒ ⇒ ⇒ ⇒ …

When rules are used in the execution of predetermined tasks they are
reduced to “instructions”. But, in the context of a non-deterministic design
process rule setting is a creative activity. Setting rules is equal to identifying
particular kinds of action, usually of repetitive character.This makes rules
ideal for the framing of design concepts.There is no more to framing a
design concept than there is to grasping a number of grammatical rules.
Nevertheless, in the context of a non-deterministic design process
possessing a number of rules is not equal to obeying to instructions. It
requires dealing with relations and features. Lacking the ability to make
relational judgments prevents one from understanding how something can
be “part of” something else and makes the grasping of a rule, or a concept,
impossible. First, “awareness” of what stands as a complete x – that is, what
we see on the left hand side of the rule, is required. Second, “judgment”, the
ability to use one’s awareness of x, to distinguish a location to “match” x in
a display at a given time, is necessary. Since different matching locations are
usually presented, judgment is required in distinguishing where and when

342 Sotirios D. Kotsopoulos

 t(x) becomes part of C in an evolving design process.Third, coping with the
recognition of matching locations that have the feature to be unfinished or
incomplete xs – sets of maximal elements that can be embedded on it,
without forming a complete x – is necessary.The recognition of incomplete
xs emphasizes the distinction of a “characteristic pattern” from an arbitrary
one.While any arbitrary x allows the recognition of individual instances t(x)
in a description, an arbitrary x does not prompt the viewer to distinguish
incomplete xs, which are not presented as wholes. On the contrary, a
characteristic pattern – that is, a pattern singled out by name – provides
better basis for the distinction of locations of incomplete matching.This
observation allows for the distinction of places that have the feature to be
parts of a characteristic pattern, and those which do not. Accordingly, a rule
schema provides better means for treating incomplete xs, since it refers to
a family of x-like forms rather than a specific x instance.

4.THE SIMMONS HALL PARADIGM

The Simmons Hall student dormitory is part of a strip of future buildings
forming the Vassar Street edge along the Briggs Athletic Field at MIT, in
Cambridge, Massachusetts.This strip of buildings was seen by Holl’s team as
an opportunity to propose a new type of student living. Instead of the
typical brick urban wall the strip was envisioned as a “porous membrane”
including a number of individual dormitories, which function as the
boundary of the residential city fabric.The dormitories were envisioned to
be “permeable” to allow maximum visual penetration. At the same time,
each dormitory was designed as “an individual house with a particular
identity” [17]. Simmons Hall is a 350 bed residence, designed as a vertical
slice of a city, 10 stories high and 382 feet long, including facilities such as a
125-seat theater and a night café.The corridors connecting the rooms were
metaphorically approached by the designers as “streets”, operating as public
places. In designing Simmons Hall, the features of “pores” and porous
materials were approached as tectonic possibilities.The porosity concept
formed a “porous” morphology through the application of a series of
operations. Accordingly, the overall building mass has five large scale
recesses, while a system of vertical sponge-form cavities creates vertical
porosity, allowing natural light and air to circulate in the interior.The
building facades are composed by 5538 operable windows.

Figure 5. Erecting Simmons Hall
student dormitory

From Design Concepts to Design Descriptions 343

 The discussion of the Simmons Hall paradigm emerged out of three
interviews with the project architect Timothy Bade, two meetings with the
architect Steven Holl and a public lecture that Holl gave at Columbia
University, in February 12, 2003.The illustrations include descriptions such
as sketches, working drawings and models from the various stages of the
design process.The presented visual material was also exhibited at the
opening of Simmons Hall, in February 2003.
In the next section formal-generative methods are outlined that depict
how the concept of “porosity” was transposed in a tectonic-urban context
to guide the production of a sponge-like morphology for Simmons Hall.
Parametric rule schemata are used to map the patterns of development in
the working drawings and the models.The analysis is retrospective.The rule
schemata aim at framing the design concept rather than describing the
actual studio techniques.The adopted approach is favored for its
explanatory and productive merits: it makes the design concept transparent
by establishing links between “words” and “actions”, and it provides a basis
for further computational exploration.

4.1. Porosity
Pore (from Greek πóρoζ) means “a minute opening”. Porosity or “the state
of being porous” in medicine and the study of plants and animals indicates
the existence of small openings. In biology and in organic chemistry porosity
is defined as: “the attribute of an organic body to have a large number of
small openings and passages that allow matter to pass through”.The forms,
the sizes and the distribution of pores are arbitrary.Their functionality is
associated with circulation and filtration with respect to the external
environment.The concept of porosity was transposed in a tectonic/urban
context to guide the production of a porous morphology for Simmons Hall.
This brings into mind the principle of “concept displacement” as described
in Schön [3]. Holl [16] mentions the influence of Marleau-Ponty stating that
environments include patterns or “lines of force” and possibly meanings.
Holl’s conceptual approach to design is addressed in his public talks [17]:
“Within the phenomena of experience in a build construction, the
organizing idea is a hidden thread connecting dispersed parts with exact
intention”. More specifically, Holl [16] phrased the working hypothesis for
Simmons Hall as follows: “What if one aspect of a site – porosity – becomes
a concept? Porosity can be a new type of being…We hope to develop the
possibility of a collection of things held together in a new way”.Table 2
presents the synonyms used by Holl’s team in organizing a contextual
definition for porosity.

344 Sotirios D. Kotsopoulos


Table 2: Contextual definition porosity
of the design concept of porosity porous, permeable honeycomb
by Steven Holl Architects, New screen, net riddle, sponge
York, NY pore opening, hole
aperture, passageway cribiformity
sieve-like, sieve pervious
unrestricted

The porosity concept was part of a wider hypothesis, the “permeability

hypothesis”, conjecturing that a porous morphology would have positive
effects at an urban and building scale: better air and light circulation, better
accessibility and visibility at an urban scale, and better communication
between interior and exterior at a building scale.The permeability
hypothesis helped the design team to establish novel relationships among
the elements of the building program. Holl [16] recalls: “Our project began
by rejecting an urban plan that called for a wall of brick buildings of a
particular ‘Boston type’. Instead, we argued for urban porosity”. At the early
stages, the design team developed a series of design alternatives. Contrary
to what is suggested in Ullman [2] – multiple concept generation and
evaluation – each of the early case-study-designs developed by Holl’s team
was a demonstration of implementing variations of the same conceptual
scheme of “porosity”.The schematic variations included “horizontal”,
“vertical”, “diagonal” and “overall” porosity alternatives, characterized by
various types and degrees of “permeability”.
Some early schematic arrangements appear in Figure 6.

Figure 6. Building types of
permeability. Steven Holl Architects,
New York, NY

After many of the schematic proposals were rejected by the building

authority, the design team shifted to the “sponge” example to implement
what was labeled as “overall urban porosity”. Overall porosity was based on
two general schemes: the abstract, recursive scheme of the Menger sponge
and the organic scheme of the natural sponge (Figure 7, up-left). A tectonic
version of porosity was realized by bringing in contact as much of the
building’s interior with the exterior as possible, by creating openings,
internal channels and cavities.This was accomplished in four ways: first, by
creating large-scale recesses of building mass; second, by creating
protrusions of building mass; third, by distributing multiple windows of
various size on the elevations; and fourth, by distributing a number of free-
form vertical cavities, penetrating the building from top to bottom.The four

From Design Concepts to Design Descriptions 345

 operations guided the production of design descriptions.The results were
depicted in sketches and models, a representative fraction of which appears
in Figure 7.

 Figure 7. Overall porosity. Early

design schemes. Steven Holl
Architects, New York, NY

The four design operations correspond to four parametric rule schemata A,

B, Γ, ∆.The rule schemata A, B operate in the algebra U33, which includes
solids manipulated in 3-dimensional space.The rule schemata Γ, ∆ operate
in the product U13×U33, which includes lines and solids manipulated in 3-
dimensional space.The four design operations, described by the rule
schemata, are: A) large-scale prismatic voids of building mass are produced
through subtraction, B) large-scale protrusions are produced through
division and translation of half of the building along its longitudinal axis, Γ)
sieve-like windows are applied on the façade panels, through subtraction, ∆)
vertical free, sponge like forms are embedded in the orthogonal building
grid.
Illustrations of the four parametric rule schemata appear in Table 3.

 Table 3:The parametric rule

porosity
schemata A, B, Γ, ∆
Rule schema A Rule schema B

Rule schema Γ Rule schema ∆

Rule schema A
The first operation allows the creation of prismatic recesses of building
mass.These are described by the design team as “large scale openings,

346 Sotirios D. Kotsopoulos

 corresponding to main entrances, view corridors, and the main outdoor
activity terraces of the dormitory”.The operation exposes more building
surfaces towards the exterior and forms terraces. A rule schema depicts the
operation: solids are subtracted from a larger solid, corresponding to the
overall building mass.The solids are parametric oblongs and prisms.The
application of rule schema A affects the form and the available square-
footage of the building.The parametric rule schema A is repetitive: it applies
several times under translation, rotation, reflection and scaling.

Rule Schema B
The second operation divides the building mass in two halves, and translates
one along its longitudinal axis.The specific transformation was labeled by
Holl’s team as “diagonal porosity”.The corresponding rule schema divides a
parametric solid into two equal parametric solids and translates one half
along its long axis, for some distance x.The result from the application of
this operation is that more of the building’s interior is exposed towards the
exterior.The application of rule schema B affects the building’s form without
altering its square-footage.The parametric rule schema B is non repetitive: it
applies only once under a single transformation, at the early stages of the
design process.

Rule Schema Γ
A third operation is used for the treatment of the elevations, to distribute
windows of various size.The operation has its conceptual basis in
mathematics and the concept of the Sierpiński carpet, or its 3-dimentional
extension the Menger sponge (Figure 7, up).The Sierpiński carpet is a 2-
dimentional fractal constructed as follows: a) a plane, square in shape, is
divided in 3 × 3 = 9 congruent squares, b) the center square is removed.
This treatment applies recursively to the remaining 8 squares and it may
continue indefinitely. A 3-dimentional version of the Sierpiński carpet,
involving cubes instead of squares, forms the Menger sponge. Illustrations of
Level 1, 2 and 3 Sierpiński carpets, appear in Figure 8.

Figure 8.The second, third and

fourth shapes from the left, depict a
Level 1, 2 and 3 Sierpiński carpet

In the context of designing Simmons Hall, rule schema Γ applies recursively

on a square façade panel to produce multiple openings, through subtraction.
Rule schema Γ introduces a mathematical expression of a “sponge”, which is
based on the recursive construction of the Sierpiński carpet.This
construction also responds to a need for better air and light circulation,

From Design Concepts to Design Descriptions 347

 which was a core design concern in designing Simmons Hall.The application
of rule schema Γ, affects the facades without altering the square-footage of
the building.The parametric rule schema Γ, applies several times under
translation and scaling.

Rule Schema ∆
A fourth operation, introduced by Holl’s team, was named “vertical
porosity”.Vertical, sponge-like openings penetrate the building from top to
bottom, allowing circulation among the different levels.These are
metaphorically described by the design team as “large dynamic
openings…the lungs of the building, bringing natural light down and moving
air up through the section”.Vertical porosity is depicted here by a
parametric rule schema that embeds sponge-like forms within the grid of
any two consecutive slabs.The rule schema ∆ introduces free organic forms,
which serve as a reminiscence of the natural sponge, as opposed to the
more abstract, recursive formulations of the Menger sponge (Figure 7, up).
The application of the rule schema ∆ affects the building’s available area and
volume, and the interior space.The parametric rule schema ∆ applies
several times under translation.

Derivation
A derivation involving the applications of the parametric rule schemata A
and B appears in Table 4.The derivation is presented in three columns, each
including six steps, performed in parallel.The main derivation appears on the
left column involving a series of subtractions among solids.The subtractions
are performed in the algebra U33, which contains solids manipulated in 3-
dimensional space. At the top of the left column, the initial shape is a
parametric solid representing the overall building mass. For brevity, the rule
schema A applies twice at the first three steps of the derivation. At each
step, the left column shows the produced shape: C' = [C – t(x)] + t(y).The
center column, presents the subtracted solids t(x) in the product algebra
U13×U33, which contains lines and solids manipulated in 3-dimensional space
The right column, presents the sum of the subtracted solids at each step
Σ[t(x)].The outline of the building is also presented with lines (for visual
reference to the overall building mass).
A derivation involving the application of rule schema ∆ appears in Table 5.
The derivation is presented in two columns, top to bottom, starting from
the left. It shows how the vertical, sponge-like cavities are embedded on the
3-dimensional orthogonal building grid. For brevity, the rule schema ∆
applies more than once at all the twelve steps.The product algebra U13×U33,
which contains lines and solids manipulated in 3-dimensional space, is used
in this derivation.

348 Sotirios D. Kotsopoulos


Table 4. Derivation of porosity
[C – t (x)] + t(y) t(x) ∋ [t(x)]
after rule schemata A and B.

A, A
⇓

A, A
⇓

A, A
⇓

⇓A

B
⇓

From Design Concepts to Design Descriptions 349

  Table 5. Derivation of porosity
after rule schema ∆.

350 Sotirios D. Kotsopoulos

 A derivation involving the application of rule schema Γ appears in Table
6. It presents the generation of perforated façade panels.The rule schema Γ
applies recursively on each individual panel to produce openings that are
arranged in a 3n × 3n (with n = 1, 2, 3) orthogonal grid. For brevity, the rule
schema Γ applies at once to all the panels of an elevation.The product
algebra U13×U33 is used in this derivation.

Table 6. Derivation of porosity

after rule schema Γ.

Γ
⇓

Γ
⇓

In the generation of perforated panels for the facades of Simmons Hall,

the rule schema Γ applies recursively on a concrete prefabricated panel to
produce openings as presented in Figure 9.

From Design Concepts to Design Descriptions 351

  Figure 9. Generation of façade
panels, after recursive application of
rule schema Γ

⇒ ⇒

Façade configurations involving panels like the above are depicted in early
sketches of Simmons Hall. However, quickly the abstract concept of
recursion gave its place to more tectonic considerations. In Simmons Hall,
the exterior concrete wall serves as the main load bearing grid of the
building.This dictates the standardization of the openings.The emergent
component was the “perfcon”, a structure described by Holl’s team as “a
design allowing maximum flexibility and interaction”.The building facades
are designed to have a total of 5538 windows nested in a uniform concrete
prefabricated wall 18'' thick that fuses windows and structure. Early studies
depicting the two design approaches appear in Figure10.

 Figure 10. Early studies of Simmons

Hall, by Steven Holl Architects, New
York, NY

The generation of a typical “perfcon” panel is depicted in Figure 11 by a rule

of the form x→y.

 Figure 11. Formation of a “perfcon”

panel in the product algebra U13×U33.

A typical “perfcon” panel has three 2' × 2' windows in height and six in
width (3 × 6). Each individual room has nine windows in total (3 × 3).
Therefore, typically, a perforated panel covers two adjacent rooms.
However, there are 3 × 5 panels and a small number of 3 × 4, 3 × 3, 3 × 2,
3 × 1 and 3 × 7 panels. A schematic presentation of the full panel vocabulary
appears in Figure 12.

352 Sotirios D. Kotsopoulos


Figure 12. Basic variations of
“perfcon” building panels

The Figure 13 presents the parametric rule schema Γ1, which captures the
generation of openings on panels.The rule schema has the general form
x → x – t (y).Three vertically aligned openings, are arranged in a 3 × n grid
(with n = 1,2,…6), while a standard panel-ending is applied to all panels on
their left and right side.

Figure 13. Parametric rule schema

Γ1 that describes the formation of
“perfcon” variations

Larger openings were applied in areas corresponding to lounges, while

select windows were filled in, in areas which were critically overstressed.
For the treatment of these two cases, two rule schemata are presented in
Figure 14. Rule schema Γ2 (left) unites a number of window openings into a
single window opening. Rule schema Γ3 (right) substitutes a window
opening with a concrete block.

Figure 14. Rule schema Γ2 (left) and

rule schema Γ3 (right)

In the derivation of Table 7, the panels are generated at the first step using
rule schema Γ1. Large openings are applied at the second step using rule
schema Γ2, and the blocking of window openings, after rule schema Γ3, ends
the derivation.This derivation, in the algebra U13×U33, does not capture the
order of panel placement.
In the example of Figure 15, the rule schemata Γ1, Γ2 and Γ3 apply on a
pair of concrete panels to create and to block opening window openings.

Figure 15. Sample derivation of

Γ1 Γ1 Γ2 Γ3
perforated panels, after rule schemata
⇒ ⇒ ⇒ ⇒
Γ1, Γ2 and Γ3

From Design Concepts to Design Descriptions 353

  Table 7. Derivation of porosity
after rule schema Γ.

⇓ Γ1

⇓ Γ2

⇓ Γ3

354 Sotirios D. Kotsopoulos

 The “perfcon” panels were used as building blocks in the erection of the
facades: 291 panels were used in total, with approximately 11 panels erected
per day. Due to the structural importance of the façade walls, the actual
placement of the different panel types, and the number and size of their
openings, were determined by structural and functional criteria. In a
computational simulation of the principles that guided the panel placement,
the discrimination of the various panel types can be expressed with the aid of
labels, while the act of panel placement can be controlled with substitution
rules. For example, at each floor: the typical 3 × 6 panels may get substituted
first, the 3 × 5 panels next, the 3 × 4 next, etc., and the corner panels last. Of
course, the determination of the position of each panel on a façade was the
result of calculations that occurred in a much later stage of the design
process.Accordingly, the formation of the appropriate computational rules,
while taking into account these calculations, requires a level of detail that goes
beyond the objectives of this discussion.An exposition of the necessary
construction rules may follow in a future presentation.

Figure 16. Placing the final piece

4.2. Discussion
Evidence of the application of the four parametric rule schemata A, B, Γ, ∆
can be traced at the early design representations, such as sketches, physical
models and schematic illustrations, of Simmons Hall. A possible
retrospective illustration, emerging from the application of the porosity
operations appears in Figure 17.

Figure 17. A possible retrospective

illustration of Simmons Hall, after the
porosity operations

From Design Concepts to Design Descriptions 355

 A comparison among the early design descriptions and the actual
Simmons Hall dormitory shows that some of the early conceptual schemes
were partially or totally abandoned in the process of implementation.The
building deviates from the descriptions of the design intent. For example,
the building recesses generated by the rule schema A were partially
reversed. And the results of the application of rule schema B (diagonal
porosity), were entirely eliminated during implementation. Several windows
generated by the rule schema Γ, were ultimately blocked by concrete
blocks, due to structural or other requirements. For the same reason, the
abstract concept of recursion was dropped rather early, and the intended
window variety on the facades, was restricted to a single standard window,
with the exception of a limited number of larger openings. Further, the
indented creation of multiple cavities, via rule schema ∆ (vertical porosity)
was hindered: only three vertical cavities were distributed to the overall
building mass. Last, due to the fire-safety regulations, the vertical cavities
were not allowed to penetrate the building from top to bottom, thus failing
to fulfill their original functional purpose of vertical circulation. A
summarizing visual comparison among the early descriptions of intents and
the finally implemented building configurations is presented in Table 8.

 Table 8. Intended (up) and actual

(down) implementation of Simmons
Hall

However, it is still accurate to say that the implementation of Simmons Hall

falls within the design space that was determined by the concept of
porosity.The final design can still be produced by instances of the
parametric rule schemata A, Γ, ∆.The rule schema A introduces “pores” at a
larger scale: that of the building mass.The rule schema Γ introduces “pores”
at a smaller scale: that of the façade panel. And, the rule schema ∆
introduces organic cavities affecting both scales: the building’s mass and the
interior space.The productive contribution of the “permeability” hypothesis
and of the “porosity” concept was to focus the designers’ attention on

356 Sotirios D. Kotsopoulos

particular features of the problem, and to provide a basis for the building of
a problem space containing these features. Making novel associations and
building a non trivial problem representation is essential in the production
of novel solutions. It signifies what Holl describes as “remaining
experimental and open”. Rather than simply solving a given program, within
predetermined confines, what the architect contributes to is the exploration
of new hypotheses.Then, the actual process of “making” supports or
invalidates the hypothetical constructions. Holl [17] concludes: “By making,
we realize that the idea is only a seed for extension in phenomena.
Experience becomes a kind of reasoning distinct to the making of
architecture.Whether reflecting on the unity of concept and experience or
the intertwining of idea and phenomena, the hope is to unite intellect and
feeling, precision with soul”.

5. CONCLUSION
For many architects and designers, the ability to diagnose problems and to
formulate productive hypotheses plays a key role in the development of
innovative design solutions in the studio. Productive hypotheses allow
designers to interpret the available design information in new ways.They
lead to the reframing of the problems and to the development of new
methods of production. Productive hypotheses are associated with the
introduction of concepts.The role of concepts in design is both descriptive
and productive.Verbal descriptions, keywords and conceptual schemes set
forth at the early stages of the design process are economical and inclusive
means to frame a general approach.
Background assumption and motivation of this study is that a design
concept is at its root a course of action meant to be performed by the
designers in the studio. Design concepts are introduced contextually and
jointly with a course of productive action that is organized and explained in
terms of them. Interpreting the output of the action confers meaning on the
concepts.This allows design concepts to evolve in parallel to designs. Design
concepts become explicit as designers adapt their general schemes to specific
contexts.This progression is analogous to moving from implicit principles to
explicit modes of action and their parameters. From a computational
standpoint this is analogous to moving from rule schemata to rules.
Novel aspect of this paper was to show how computational rules can be
useful in describing the early productive design processes and their
suppositions. It has presented a paradigm of how a design concept can be
treated by formal-generative means in designing from scratch. And it has
shown that a design concept can be converted into a system of visual rule
schemata to generate design descriptions.The rule schemata were
expressed by the means of shape grammar formalism.The paradigm was a
retrospective analysis of how the concept of “porosity” was used by
architect Holl and his team in designing the 350-unit student residence

From Design Concepts to Design Descriptions 357

 Simmons Hall, at MIT.
It was also pointed out that design concepts do not necessarily emerge
out of the existing design representations, alone. But, they can be imports
from domains foreign to design like biology, physics, mathematics, music,
etc., which are transposed to design contexts, with a view to focus the
designers’ attention on particular features of a problem and to propagate a
certain course of productive action. Of course, the use of a word like
“porosity” within a design context, does not guarantee the existence of a
straightforward, formal way to identify the actions or the things it denotes,
or to determine their features. However, the presented paradigm
demonstrated that when there is a recurring structure, or a “characteristic
pattern”, a domain can be specified for which rules and meanings can be
resolved.
Computational rule schemata were found ideal for the framing of design
concepts. It was proposed that there is no more to framing a design
concept than there is to setting a number of grammatical rules. Setting rules
is equal to identifying particular kinds of activity, usually of repetitive
character. It requires dealing with relations and features. Being unable to
make relational judgments prevents one from understanding how something
can be “part of” something else and makes the grasping of a rule, or a
concept, impossible. In applying a rule of the form x → y, “awareness” of
what stands as a complete x, and “judgment” to distinguish the appropriate
place and time to make x “part of” a description, are required. A
characteristic pattern x, singled out by name, has the privilege to expand the
domain of the matching places, as it prompts the viewer to discover
incomplete xs. It provides the best basis for distinguishing places having the
feature to be parts of x. Accordingly, a rule schema provides better means
for treating an x-like-family of forms, than a rule that treats only a specific x-
instance.
In conclusion, the early exploratory stages of design can benefit by a
formal-generative methodology in multiple ways.The description of design
concepts by rule schemata allows communicating abstract ideas with
precision without sacrificing generality. And the mapping of the actions
implied by a concept to computational rules, allows the implementation of
design tools with strong generative capacity.The organization of sequences
of rules leads to the construction of grammars. And, the rule sequences can
be transferred to a programming language, as scripts, to become available
for execution by digital machines.The encoding of design concepts into
rules and custom computational tools makes the design process transparent
and permits the future access and reuse of the design ideas.
In extension to the above, one may say that showing attention to the
early conceptual design schemes may provide a link between intuitive and
computational design methodologies.This is important in architectural
design, where the distinction between general problem solving methods and

358 Sotirios D. Kotsopoulos

case-specific techniques is not firmly established.The development of custom
digital design tools without an outline of formal or spatial expectations often
absorbs a lot of designers’ energy and attention and drives them away from
their intended spatial objectives.The ability to retain associations with early
conceptual schemes assists in framing ideas of explicit patterns of activity and
assures that one is not dealing with irrelevant issues.

Acknowledgements
I am indebted to the architects Steven Holl and Timothy Bade for allowing
access to the design material of Simmons Hall. I am also indebted to Prof.
Terry Knight and Prof. George Stiny, at MIT, for making useful suggestions.

References
1. Holl, S., Idea and Phenomena, Lars Muller Publishers, 2002, 73
2. Ullman, D. G., The mechanical design process, McGraw Hill, 1992, 120-172
3. Schön, D. A., Displacement of Concepts,Tavistock Publications, London, 1963
4. Schön, D.A., Educating the Reflective Practitioner, Jossey-Bass, San Francisco,1990, 3-7
5. Gero, J. S., Concept formation in design, Knowledge-Based Systems, 1998, 10 (7-8),
429-435
6. Suwa, M., Gero, J. S. and Purcell,T.,The role of sketches in early conceptual design
processes, Proceedings of Twentieth Annual Meeting of the Cognitive Science Society,
Lawrence Erlbaum, Hillsdale, New Jersey, 1998, 1043-1048.
7. Stiny, G. and Gips J., Shape Grammars and the generative specification in painting
and sculpture, Information Processing 71, ed. Freiman C.V., North Holland, 1972
8. Stiny, G., Shape:Talking about seeing and doing,The MIT Press, Massachusetts, 2006
9. Knight,T., Creativity. Rules, Proceedings of HI '05 Sixth International Roundtable
Conference on Computational and Cognitive Models of Creative Design, Heron Island,
Australia, 2005.
10. Sass, L., A Physical Design Grammar: A production system for layered
manufacturing Automation in Construction, (forthcoming)
11. Celani, M. G. C., Beyond Analysis and Representation in CAD: A new computational
approach to design education, PhD Thesis, Massachusetts Institute of Technology,
2002.
12. Duarte J. P.,Towards the mass customization of housing: the grammar of Siza’s
houses at Malagueira, Environment and Planning B: Planning and Design 32, 2005,
347-380
13. Kotsopoulos, S. D., Constructing Design Concepts: A computational approach to the
synthesis of architectural form, PhD Thesis, Massachusetts Institute of Technology,
2005.
14. Knight,T., and Sass L., Looks Count: Computing and Constructing Visually Expressive
Mass Customized Housing, Massachusetts Institute of Technology, (forthcoming)
15. March, L.,The logic of design and the question of value, in The Architecture of
Form, (March L., Ed) Cambridge University Press, 1976, 18
16. Holl, S., Parallax, Princeton Arch. Press, 2000, 174, 305, 308
17. Holl, S., Public lecture,Wood auditorium, Columbia University School of
Architecture and Urban Planning, New York,Wednesday, February 12, 2003, 6:30
PM

From Design Concepts to Design Descriptions 359

 Sotirios D. Kotsopoulos
Massachusetts Institute of Technology
Department of Architecture
77 Massachusetts Avenue, 10-482M, Cambridge, MA 02139
[email protected]

360 Sotirios D. Kotsopoulos

View publication stats