Content uploaded by Pelin Dursun
Author content
All content in this area was uploaded by Pelin Dursun on Oct 14, 2016
Content may be subject to copyright.
inking and designing with the
idea of network in architecture
Abstract
A spatial setup is designed considering the network of interrelations between
its constituent units. is is a network signicant for architectural discourse as it
maps the interactions and social relations between users, denes the functional
and latent routes, and indicates spatial proximities. Although design is subjective,
design tools and methods provide objective criteria to interpret and iterate. Com-
mon tools of network thinking allow us to invoke scenarios that will lead us to
visualize and exchange ideas about architecture, extrapolate up to date functional
ratios, dene ranges of proximities to bring forth spatial and potentialities of ar-
chitectural program and test them within criteria.
is study focuses on the idea of networks in architectural design and discusses
the use of graph theory based tools in the design process. It presents the possi-
bilities of systematic mapping of relations among spatial elements through their
neighboring and attracting qualities in the initial phase whereby the relational
network is still dynamic and non-hierarchical. e topic will be expressed by
presenting two examples, one from an academic setting, the other elicited from
practice. e rst describes a workshop on systems thinking demonstrated with a
game called “İkidebir”. e second is an iterative hospital campus design scheme
in which functional and site specic relationships are modeled and animated with
network modeling and assessment tools. Network-based thinking, graphs mea-
surements, and the diagrammatic assessment of relationships between spatial or-
ganizations as a design exercise are valuable both for those who are in practice and
in the education of architectural design.
Keywords
Networks, Architectural design, Relational thinking, Space syntax.
Nilüfer KOZİKOĞLU1, Pelin DURSUN ÇEBİ2
1 nilufer.kozikoglu@izmirekonomi.edu.tr • Department of Architecture, Faculty
of of Fine Arts and Design, Izmir University of Economics, Izmir, Turkey
2 dursunpe@itu.edu.tr • Department of Architecture, Faculty of Architecture,
Istanbul Technical University, Istanbul, Turkey
Received: May 2015 Final Acceptance: October 2015
ITU A|Z • Vol 12 No 3 • November 2015 • 71-87
ITU A|Z • Vol 12 No 3 • November 2015 • N. Kozikoğlu, P. Dursun Çebi
72
1. Introduction: Space as a networked
artifact
Today, networks, which can be de-
scribed as structural and organiza-
tional models, are pervasive in every
aspect of our lives and range from
genes to power systems and from so-
cial communities to transport routes.
ese networks are concerned with the
structure of relations between things
and are informative as they allow us to
uncover those inherent principles and
behaviors that regulate a variety of nat-
ural and articial systems (Lima, 2011;
Wigley, 2007).
In the eld of architecture, the study
of networks has emerged as an inspir-
ing concept in the description of built
environments. Aer all, the design of
a spatial setting inherently implies the
network of interrelated spatial units,
and so we can view the practice of ar-
chitecture as mainly involved in the
creation of the specic conguration of
this network. In other words, the out-
come of an architectural design process
is essentially a conguration (Nourian
et al., 2013). Network relationships are
thus tools that the architect utilizes to
propose his/her perceptions. ese
relationships once regarded as a mu-
table also constitute the potentials of
encounters for the users through con-
nections and borders, including even
new ranges and thresholds. us they
make up the base for the interactions
and social relations between users,
dening both functional and latent
routes, and indicating spatial prox-
imities and neighbors. According to
Dovey and Dickson (2002), the spatial
dispositions of buildings constitute so-
cial organizations. ey are not formal
types or archetypes, but, rather, clus-
ters of spatial segments structured in
certain formations with syntactic rules
of sequence and adjacency. Lawson
develops this view by dening archi-
tectural and urban spaces as containers
that accommodate, separate, structure
and organize, facilitate, heighten, and
even celebrate spatial behavior. He says
that space creates settings that orga-
nize our lives, activities and relation-
ships (Lawson, 2005). Hillier suggests
that buildings carry social ideas within
their spatial forms (Hillier, 1996) and
spatial formations can be seen as visual
symbols of societies. We read the space
and anticipate a life-style (Hillier and
Hanson, 1984).
To date, most of the research studies
that set out to reveal the potentials of
network systems have utilized graph
theory, a theory that relies on the con-
version of information into a network
diagram that can be mathematical-
ly analyzed to determine the relative
depth or signicance of the nodes or
edges that make up the network
(Ostwald and Dawes, 2013). Archi-
tectural applications of this method
have also been developed by several
researchers (Alexander, 1964; March
and Steadman, 1971; March, 1976;
Steadman, 1983; Hillier and Hanson,
1984; Hillier, 1996). Generally speak-
ing, these works discuss some of the
concepts of mathematics and diagram-
ming or graph theory based tools that
have potential value in understanding
architectural forms and spatial or-
ganizations. ey primarily present
the architectural designer with some
mathematical methods of conceiving
and manipulating the spatial congu-
rations.
An analysis of utilizations of graph
theory based tools in architecture sug-
gest there are in three dierent modes:
(1) to analyze existing spatial forma-
tion (Hillier et al., 1987; March and
Steadman, 1971), (2) to generate spa-
tial form, (Mitchell et al., 1976; Stead-
man, 1983), and (3) to evaluate archi-
tectural design (March, 1976; Hillier,
1998; Space Syntax, 2002). e rst
of these types of utilizations begins by
exploring the intrinsic nature of the
existing built environment and then
decoding the underlying principles
and meanings. e second group uses
a series of predened rules in a com-
puterized, automated process to search
for a desired spatial product. e last
group provides tools that architects
may use to evaluate their design pro-
posals and also gives them opportuni-
ties to argue for the best performing
proposals. e criticism leveled against
these approaches mostly stems from
the following questions: To what de-
gree does an architect become involved
in this cognitive process and how does
he/she evaluate their designs consid-
ering desirable social implications
inking and designing with the idea of network in architecture
73
rather than focusing on an automated
evolutionary process? (Nourian et al.,
2013). In the last decade an analytical
approach, space syntax theory and its
applications, has made great strides in
showing architects the possible eects
of their design solutions and have en-
abled them to learn from their design
solutions (Dursun, 2007, 2012). In this
way such utilizations constitute evi-
dence-based design processes (Han-
son, 2001).
Space Syntax theory is constituted
on two hypotheses (Dursun, 2012):
1. e built environment functions as
a spatial / social network. In this net-
work the main interest is about rela-
tional characteristics of spaces rather
than individual ones. Space is experi-
enced through this spatial networks
or relations. 2. Spatial networks create
potentials of movement and describe
a living pattern. Movement is the key
element to decode man-space / man-
man relationship. Based on this net-
work structure spatial congurations
embody social or cultural meanings
and generate or inhibit social interac-
tions, movement patterns in built en-
vironments.
Such analysis tools guide in the
comprehension and depiction of the
relational structures however there has
been complicacies in the transfer of
this research-based knowledge directly
to design process. e design process
is conventionally perceptive, experi-
ential and subjective. e method to
reference research based knowledge to
the design process is a typically a mat-
ter of concern for most. Dovey tries to
explain this contradiction by focusing
on relation between phenomenological
philosophy and Cartesian world. He
describes these poles “lived space” (the
realm of personal feelings, emotions
and particulars) and “geometric space”
(the space of plans, forms and univer-
sals) (Dovey, 1993). According to Dov-
ey, geometric space is a representation
of lived space with the meanings and
values extracted. For him, geometric
space is a universal language of spatial
representation that has predictive val-
ue. How can one creatively externalize
the spatial knowledge in a measurable,
visible manner for evaluation for as-
sessment and improvement even from
the initial stages of the design process?
Design is a complex cognitive pro-
cess that continuously engenders both
problems and solutions (Lawson,
2003). It is a kind of experimental pro-
cess that is largely learned and prac-
ticed through “making” (Schön, 1987;
Al-Sayed, 2012). Rather than searching
for optimal solutions (Simon, 1996),
design is about experimenting and
probing. Experiments lead architects
to discover something, and then these
help them to redene their underlying
concepts (Dursun, 2007). In network
thinking the investigation focuses
on systematically mapping relations
among spatial elements through their
shared and relative characteristics, in
other words, neighboring and attract-
ing qualities in rule-based dynamic
network models. e “relations of the
relations” and “the protocol between
the rules,” which refer to the order and
the scale that the rules will be enacted
during the design process, are of prime
importance in these models. By ob-
serving the eects, the creative process
can be interpreted as a kind of chore-
ography, one in which “pace” is also
interrogated for the elements of the
parametric model.
It is possible to deduct that relation-
al qualities that suggests life inside a
spatial construct i.e. social interactions
and the movement (form prone to ow
patterns) and proximities are built up
by formal qualities dened by rule sets
i.e. distance close or far, vertical posi-
tions, below or over, and whether clus-
tering or disparate. Dynamic network
models suppose that spatial entities are
in constant motion during the design
process. eir exact positions are yet
ambiguous, they hang in air, and sway,
or jump from one location to another;
they start to presume specic locations
and concretize as their relationships
among each other become more and
more dened.
is study aims to explore following
question: How we can use the idea of
network in architectural design? By
focusing on the experimental and in-
tellectual characteristics of the design
activity the study tries to examine how
this kind of thinking can be used as a
creative and informative tool in design
process. In the scope of the study rst,
ITU A|Z • Vol 12 No 3 • November 2015 • N. Kozikoğlu, P. Dursun Çebi
74
the main question is opened for dis-
cussion conceptually with architectural
students by the help of a game, İkidebir.
Secondly, the authors try to explore
how this kind of thinking can be uti-
lized in the design process by focusing
on an iterative hospital campus design
scheme from practice. Cross indicates
that design has its own distinct intel-
lectual culture and has its own ways
of knowing, thinking, acting (Cross,
2007). Based on this idea this experi-
mental study aims to open a discussion
about how, a scientic and graphic tool,
network thinking and modeling, could
feed the design thinking and making.
2. Playing the systems game –
“İkidebir”
Played by architecture students as a
component of the Architectural Mor-
phology class at the ITU Faculty of
Architecture in 2014, “Ikidebir,” is a
game in which simple rules make up
a network where the nodes are in mo-
tion until they asymptotically settle
into a conguration that satises the
rule for each individual. is game
engenders a dynamic system in a giv-
en space, and has the following rules:
(1) Players initially announce an av-
atar, a spatial entity in this case, they
selected for themselves and represent it
as a node; (2) each player then selects
two other announced nodes in order
to follow in discrete this time. (3) All
players randomly position themselves
in the conned space (game area). (4)
Hearing the start signal the players try
to stand at equal distance to the two
nodes whom they have picked to follow
(Figure 1). e students rst write their
selected spatial entities and later draw
the relations that form on the board,
(Figure 2). en the system is opened
for discussion with the students. Af-
ter introducing some analytical tools,
space syntax and other dynamic net-
work models such as cytoscape to de-
code this relational structure, the au-
thors re-evaluated the process by the
feedbacks of the students.
Network can be described as a struc-
ture that is constituted by the links be-
tween nodes. ese nodes can repre-
sent dierent entities such as individual
person, object, space or concept. Both
countable and non-countable entities
can be interrelated. For example in the
rst series of the workshops for this
game in the Architectural Morphol-
ogy class between 2008 and 2014, the
students selected ctitious avatars and
that had caused a more concentrated
discussion on the nature of networks.
However in the last workshop students
selected to represent spatial units. us
the composed networks lead the play-
ers to question the adopted relations
that provide typical congurations.
Recorded sessions are revealed at the
blog: http://ikide1.wordpress.com.
As soon as the game starts, play-
ers move in order to position them-
selves between their selected players.
However as those players are also in
movement, they continuously have to
recalculate their target positions. is
can be seen as a systemic ow, which
sometimes accelerates and sometimes
slows down. e simple rules create a
dynamic set of nodes until the game
settles into an arrangement that satis-
es the rule for each player.
Discussions with the students yield-
ed the following key aspects:
Figure 1. Ikidebir Game – A Demonstration of Game Evolution.
Figure 2. Choice of Spaces and Relational Characteristics
Visualized by Cytoscape Program.
inking and designing with the idea of network in architecture
75
• e networked structure is in-
formed by the choices made by the
players.
• e rule that the relation between
spaces must be equidistant to the
selected two spaces both triggers
and organizes the motion.
• e nodes, spatial entities in this
case are xed in terms of the links
created whereas their geometric
compositions are constantly chang-
ing.
• By default each player selects two
other spatial entities; therefore ev-
eryone is plotted into the network.
• Only two spaces may be selected.
is type of selection brings an im-
portant limitation for the interrela-
tions among the nodes. Such that
each space is connected by at least
two spaces and remains linked with
the whole. On the other hand some
spaces are selected more than the
others and this causes the system to
lose homogeneity and leads it to a
varied distribution.
• e nodes selected by more nodes
tend to be key elements in the sys-
tem, in the given case “kitchen, en-
tryway, and courtyard”. eir po-
sitions / or uctuations aect the
whole group causing both acceler-
ations and decelerations. erefore
these nodes are latent to change the
form of the system.
• Some nodes – such as porch and
sofa – are less signicant to the sys-
tem and thus they either may be se-
lected by only a few players or even
by none at all. eir actions do not
create major changes.
• However, even though they may be
less-selected, some nodes – such as
the winter garden – may prove to
be eective, especially so when they
are selected by a single player who is
selected by many.
By the introduction of graph the-
ory based tools such as space syntax
and cytoscape to analyze the network
structure the students tried to make
this network legible and accessible
to reading and assessing (Figure 2).
Based on mathematical and graphical
data, following questions are put into
considerations: How do the selected
nodes (avatars) behave in that particu-
lar system? How do they interact? How
many connections do they have? What
do they share? Are they interactive or
are they inactive? Is there any key con-
nection among them? Are there any
groups or divisions (clusters) between
them?
e relational whole in the graph-
ic and the calculated syntactic val-
ues, such as integration, connectivity,
depth, choice, etc., rationally support
the experience of the students’ percep-
tion of the choices (Figure 3). ese
explorations induce some valuable
insights associated with the network
structure:
• In order to play students made ran-
dom selections from spatial entities
as avatars. e choices are mostly
relevant in a residential setting, de-
ning a quality or a program inher-
ent to that space, like living room,
kitchen, bathroom, WC, entryway,
terrace, or nursery, or a few less
common spaces like a cellar. e
game also includes spaces more typ-
ical of traditional Turkish architec-
ture, like the inner courtyard, iwan
(vaulted hall) and sofa (connecting
hall or egress space).
• Hearing all the choices, students
then selected two other two spa-
tial avatars to be linked to from the
available set in the group. ese
choices result in conventional rela-
tions such as sofa-courtyard, living
room-kitchen, terrace-entryway,
cellar-kitchen, kitchen-WC and
some unusual relations such as liv-
Figure 3. e Relational Whole and Calculated Values Visualized by Space Syntax for Grasshopper - İkidebir.
ITU A|Z • Vol 12 No 3 • November 2015 • N. Kozikoğlu, P. Dursun Çebi
76
ing room-iwan, living room-bath-
room, bathroom-kitchen, nursery
room-WC. is allows the players
to experiment on uncommon or
secondary relationships.
• ese selections provide enough in-
formation to analyze and gure out
the key entities in the network are
“kitchen, entry way and courtyard”.
ese represent powerful nodes
that have strong relations with the
other nodes. e game also pre-
sented that these nodes initiated the
motion and acted on the pace of the
system. “Sofa and porch” tended to
be inactive nodes. As they do not
have strong relations with the oth-
er nodes, their eects on the spatial
system are limited. Syntactic anal-
yses clarify these characteristics.
Integration values for the spaces
reveal the following order: kitch-
en (2.636) > entryway = courtyard
(2.197) > storehouse = winter gar-
den = terrace = living room = WC
(1.883) > bath (1.757) > iwan =
stairs (1.647) > cellar (1.551) > foyer
(1.318) > nursery = porch (1.255) >
sofa (1.198).
• Networks do not need to link nodes
specically of the same genre. Stu-
dents’ selections included vague
spatial entities like “entryway” as
well as very dened ones like a “cel-
lar”.
• Networks by default defy physi-
cal dimension; however, discrete
groupings suggest varying snap-
shots of spatial possibilities. Iter-
ative playing out of the rule hints
form possibilities including pro-
portions, zones, interior and exte-
rior build-up, etc. Specic network
visualization layouts simulate part–
to-part and part-to-whole relation-
ships and spatialize the network in
2D (Figure 4). Visualizing the game
with cytoscape, it is possible to vi-
sualize adjacencies and clustering
possibilities, although the model is
exempt of physical dimensions.
In this workshop network thinking
in architecture have been opened to
discussion a. through students person-
al experience b. through graph theory
related tools that analyze the demon-
strated network. In other words ab-
stract spatial network that emulates a
spatial construct is experienced by the
students participation and then exam-
ined in a cognitive scientic platform.
e study imparts the following poten-
tials network thinking in architectural
design process:
1. e spatial whole can be described
as the relations among its constit-
uent parts rather than as a sum
of disparate units. e manner in
which these relations are constitut-
ed may infer diverse connotations
and there may be quantiable as-
pects of these relational patterns.
2. e rules that construct the net-
work (one space must be selected
by at least two other spaces) and the
rules that enact on the form or the
Figure 4. Network from the Game Modeled in Network Visualization Program Cytoscape.
inking and designing with the idea of network in architecture
77
conguration (the relation between
spaces must be equidistant to the
selected two spaces) conform the
ow and the proximities between
nodes, is therefore constrained. Ac-
tual design processes include more
complex and varied relationships
and rules. However in both cases
specied rules for distancing and
clustering are indicative for the
propositions of form. To under-
stand the implications of these rule
sets and their implementation is
signicant for the designer.
3. Certain relationships tend to be
prevalent and aect change to the
whole, whereas other clusters of re-
lationships are not at all eective to
the whole, yet are dynamic in their
groupings. It will be argued that this
phenomenon relies on the designer
and the brief. e game is a demon-
stration of the relational make-up
and the dynamic quality of these
relations when they need to attain
spatiality. is conceptual visualiza-
tion or modeling enables the archi-
tect to consciously model, through
play, the bonds and proximities of
spatial units and the site.
4. Concentrating on the idea of net-
work in architectural design, space
syntax helps designer to develop
spatial awareness by transform-
ing relational spatial structures
into graphical, mathematical and
scientic forms. It explains what
does these relations mean and how
does the system works. By mak-
ing non-discursive characteristics
of space discursive, it presents a
language for thinking and talking
about space (Dursun, 2007).
5. While space syntax provides a use-
ful tool for architects in decipher-
ing and assessing the relationship
among spatial entities in terms of
spatial accessibility and human ow,
other dynamic network models
such as cytoscape and customized
parametric modeling reveals possi-
bilities regarding on geometric-for-
mal characteristics of this relational
whole. In other words, these mod-
els visualize the possible formal end
products of applied rules.
6. Relevant graph theory concepts and
criteria, diagrams, and produced
data sets based on eective repre-
sentation of spatial systems lead to
powerful instigation, management,
and assessment of design phases. It
is thus that, in contrast to conven-
tion, these tools have potentials to
be tools with which we can think
(Hillier and Hanson, 1997) during
the morphological stages. ese
tools are creative and constitute
an educational component with-
in the research-based design. ey
also lead the designer to better un-
derstand the relationship between
form and its use (function), while
opening up new possibilities for de-
sign based on research results and
generative principles (Schneider et
al., 2013).
7. e experiment does not refer to
the use of graph theory based tools
including space syntax to extract
potentials aer the architectural
form is solid rather during the ini-
tial stages of design. In this context
it advances design thinking, enables
interactive exploration of the eects
of programmatic relations on form
and suggests a method to structure
correspondence of form and func-
tion.
3. Design Research: Method to design
a campus
e second example is taken from
practice and deals with a conceptual
design scheme for a campus on psy-
chiatry and neurology. Hospitals have
been the subject of a great deal of re-
search in the architectural literature,
especially in regard to their functional
and organizational structures. Human
ow and way nding issues appear
key concepts of these researches (Ünlü
et al., 2005, Setola, 2009, Khan, 2012,
Peponis et al., 1990). e aim of our
study is to impart potentials of net-
work thinking explored in developing
this master design scheme. In parallel
to existing research, this scheme also
focuses on the human ow in terms of
vehicular and pedestrian pace between
specic subunits. e programmat-
ic and site relationships and relations
to the varied qualities of the site are
modeled and animated by the use of
custom-made modeling tools based
on network thinking. Peculiar qualities
ITU A|Z • Vol 12 No 3 • November 2015 • N. Kozikoğlu, P. Dursun Çebi
78
of the site – such as emergency-prone
segments along a major artery, or the
more tranquil neighboring residential
areas, and/or the security latent zones
are represented as polar attractors.
is design exercise incorporates in-
terrelated positioning; programs are
attracted (tied up) to specic zones or
segments, and also to one another, as
are the nodes connected to each other
by the students’ choices in the case of
the game.
In the initial phase the separable
programmatic units, regardless of their
sizes, are scripted to move around an
abstract container, pulling and push-
ing one another and the poles of the
container in terms of their space/use
related attributes. ese disparate units
are determined according to the ad-
ministrative organization chart and the
patient ow described by the clinical
team. Attributed criteria to these units
are urgency, security and privacy. Each
program unit is specied with vary-
ing degrees of these attributes (Figure
5). “Urgency” pole attracts programs
with emergency zones such as the
emergency of the neurology hospital,
privacy node attracted the acute psy-
chiatric clinical program nodes, public
pole pulled the outpatient nodes, and,
nally, security node pulled forensic
clinical nodes. By regarding these con-
tained program units as a network, the
script allows similar attribute grades
to accumulate and the dened polari-
ties to pull each other, and to move the
groupings toward specied poles of the
abstract container. e script also al-
lows for negotiations among the vary-
ing degrees of these attributes.
In the second iteration, shown in
Figure 6, the group formations are
clustered in the layout to allow prop-
agation to the actual site. In this case
the rules for propagation are parame-
terized by diusion, overlap possibili-
ty, and size. e rule implementation
follows a hierarchical order. Certain
program units link to others like their
satellite, and certain units have priori-
Figure 5. Conceptual Polarities Mapped in Relational Modeling among Program Units and
Specific Attributes.
inking and designing with the idea of network in architecture
79
ty in maintaining proximity to the de-
ned abstract polar zones. For exam-
ple, the rehabilitation unit is a sub-unit
orbiting the psychiatric clinics, bound
by the everyday personnel and patient
ow, which has been quantied as
300m walking distance. e emergen-
cy department has rst priority to be in
proximity to the main artery which is
the main urgency pole (node).
ese interrelations and hierarchy
are plotted and evaluated on a table
matrix. Size is derived from the “List of
Requirements” as well as the height and
oor space limitations as a work area
serving both a group of patients and
a health team: 20-30 patients to one
oor, as in the case of the neurology
inpatient building – up 100 patients in
total. e oors of the building oors
are limited to eight in total, suggesting
a footprint area of 1500sqm depicted
with a circle with equal area (Figure 6).
In the site implementation, the value
sets and interrelated network are then
mapped to the site directly referencing
the preferred poles and axis to certain
nodes. is ‘machinic’ diagrammat-
ic exercise is modeled and run iter-
atively. e distances between units
are dened in ranges proportional to
time and the pace of pedestrian and
vehicle reach (Figure 7). For example,
the emergency pavilion for the three
departments (psychiatry, neurology
and neurosurgery) are located in the
same spot; however, once a patient is
to be transferred to an inpatient unit,
the neurological unit is accessed via a
ight of ramps and elevators, taking a
total of ten minutes, whereas psychiat-
ric patients are transferred by vehicle
to the psychiatric inpatient clinic. One
is vertical in positioning whereas the
other is horizontal.
Each unit “behaves” and situates
according to the specied rules, with
emergency related units tending to
prefer the artery neighboring zones,
the inpatient units moving towards the
residential borders, etc. e process is
further rationalized with the use of a
major axis for pedestrian and vehicu-
lar ow and its possible orientation on
one hand and the variations provided
by possible positions of a hypothetical
center of the system on the other, cer-
tain units only following other units as
satellites (Figure 8).
is exercise is repeated in iterations
for assessment of the resulting congu-
rations. Units that are directly linked to
site poles and units that have more links
to other units have greater potentials in
dening the working conguration.
e position of an emergency plateau
close to the major road is a straightfor-
ward design decision; however the role
of the diagnosis department and its lo-
cation to the other departments is one
example where probing is necessary.
e process enables ne-tuning and
easy reassessments of multiple possi-
bilities.
e space syntax analysis also
demonstrated that the diagnosis de-
partment is the key spatial unit in the
network as it represents a powerful
node that has strong relations with the
other nodes. e rehabilitation block
and inmate unit tend to be inactive
spaces. Based on the syntactic analyses
integration values for the spaces reveal
the following order: diagnosis (4.435)
Figure 6. Matrix of Relations of Program and Site.
ITU A|Z • Vol 12 No 3 • November 2015 • N. Kozikoğlu, P. Dursun Çebi
80
> psychiatric inpatient = umatem =
forensic (1.267) > neurology inpatient
= amatem (1.109) > psychiatric outpa-
tient (0.986) > rehabilitation block =
inmate units (0.634) (Figure 9).
e focus of these design research
sessions is to be able to abstract and
re-evaluate relationships regarding
the program, and the site, and recon-
struct corresponding layout options
with their interrelation degrees in ref-
erence to specic attraction criteria.
ese attractions and repulsions, in
other words the polarized units, hint at
building/structure-prone units by their
capacity to conjoin and to cluster as
Figure 7. Setup Order for the Abstract Programmatic Polarities Diagram.
Figure 8. Relational Site Model in Iterations.
inking and designing with the idea of network in architecture
81
buildings around a courtyard, for ex-
ample. rough a series of assessment,
multiple layout potentials are derived
and compared. e process gradually
narrows into a discursive scheme and
potentials for the master plan are ex-
trapolated.
With the parametric conguration it
was possible to convey the fact that the
project model is only a snapshot of the
possible set, and yet major decisions
are more dened than others, and that
there is room for development. It is
thus a map for action (Figure 10).
While the project was actualized
in 2009, it is still under discussion as
the stakeholders continue to bear a
great burden of existing patients and
economical strain; however, it is im-
portant to note that the project has
remained viable, despite the passage
of time and change of certain person-
nel. is is mainly due to the fact that
the project is in itself a tool that allows
evaluation of site and program condi-
tions, and has the potential to change
in accordance with modication of the
site and evolving needs.
e idea of network has been inu-
ential in the process o conceptual de-
sign scheme for this campus project.
e clinical team asked for the project
to correspond with the new under-
standings as well as the client required
to evaluate all possible scenarios at the
site. Both interests were met the proj-
ect. e process imparts the following
potentials of relational thinking in ar-
chitectural design process:
1. Same as case one, here it is demon-
strated that the spatial whole can be
described as the relations among its
constituent parts rather than as a
sum of disparate units. e manner
in which these relations are consti-
tuted may result is diverse conse-
quences as to form and these rela-
tional patterns can be mapped in a
quantiable manner although they
are based on concepts.
2. e rules that construct the net-
work (common conceptual/spatial
qualities that can refer to both the
site and the functional units) and
the rules that enact on the form
or the conguration (the distances
attributed between units as a func-
tion of pedestrian and vehicular
motion) conform the ow and the
proximities between nodes. Spec-
Figure 9. e Relational Whole and Calculated Values by Space Syntax for Grasshopper – Hospital Campus.
Figure 10. Propagated Site Model.
ITU A|Z • Vol 12 No 3 • November 2015 • N. Kozikoğlu, P. Dursun Çebi
82
ied distances as well as clustering
operations (orbiting, attraction to
axis and positioning with hierar-
chy) are indicative for the propo-
sitions of form. To understand the
implications of these rule sets and
the order and the pace in which
they are implemented are of value
for the designer.
3. Certain nodes and links (relation-
ships) are more eective to change
the whole, whereas other clusters in
the network are not at all eective to
the whole, yet can be active in their
local groupings. It is important
to note that the designer takes his
position rather than an automated
generation of form in orchestrat-
ing the relational model. Also in
relational models enable to demon-
strate peculiar qualities of networks
such as an overlooked unit linked to
a major node has equal eect on the
system, and therefore may act on
the design discourse equally.
4. e project made use of cus-
tom-made parametric models
that animated the network of both
subjective (based on functional
qualities) and objective (based on
functional size, distance and ori-
entation) relations. is enables to
link attractive qualities with their
corresponding spatial abstractions.
is is made possible by the ad-
vance in the now ubiquitous digital
tools that enable live change and
tracking of parametric relational
models. us it is possible to have
variations as well as breeds of solu-
tions to a brief.
5. is process requires a sustained as-
sessment strategy for the variations
arrived by the modeling. Space syn-
tax or network visualization and
assessment with soware like cyto-
scape enable the assessment of the
relationships among spatial entities
in terms of spatial accessibility and
human ow, as well as other net-
work measurements like closest
path, clustering, etc.
6. e process involves iteration: re-
structuring the initial relational
setup, remodeling, and reformulat-
ing the physical ties (distances and
ratios), reassessment of the order of
rule enactment. It is crucial to the
process that the model is remade
up aer the initial run which serves
more as a prototype to the ma-
chine-like dynamic model.
7. In this project case, the units were
thought of as clustering similar
attributes of spatial concepts like
public/private together. However
the pattern to distribute and prop-
agate the units at the site could have
been dierent then clustering the
likes. e model only allows the de-
signer to apply his design decisions
in a prototypical manner that he
can observe exceptions, derivatives,
and possible modications live on
the model.
4. Conclusion
Architectural design is ultimately
about the congurations, connections,
shape, and orientations of physical
forms (Do and Gross, 2001). It deals
with designing connections, borders,
new ranges and thresholds in the
space. Two case studies (one derived
from architectural education and the
other from architectural practice) are
valuable both in terms of their eort
to conceptualize the idea of network
in design and to use this idea to trigger
production of space in design process.
Networks are dynamic forms in
which relations are alive, in that they
are in states of constant change. By
exploiting this way of thinking in ear-
ly stages of architectural design, it be-
comes possible to keep the negotiation
alive, which is important for a creative
process. is approach also provides
informative tools for architects as it
permits designers to see dierent po-
tentials and possibilities in design and
constitutes mediums for experiment-
ing and probing.
is study mainly concentrates on
the idea that a critical understanding
of the network in spatial constructs can
inform, shape, and enhance the design.
To exemplify the discussion, the au-
thors rst engage architecture students
in a game designed to explore how a
space paradigm can be conceptualized
through a process of dynamic network
rules. Secondly, the authors also try to
explore how this kind of thinking can
be utilized in the design process by
focusing on hospital campus design
inking and designing with the idea of network in architecture
83
scheme from practice.
e rst example aims to trigger the
architecture students to develop the
idea of spatial network in design by the
help of a system thinking game in which
all the students are actively involved.
Network thinking in architecture is
opened for discussion conceptually in
order to decipher the potentials of the
space and make its un-discursive, in-
tangible characteristics discursive and
tangible. is experiment constitutes
a conceptual ground that permits the
designer to understand the dynamic
interaction among the design parame-
ters, and also permits evaluation of the
relationships and their meanings in the
design. Here network thinking appears
as a powerful tool in order to under-
line the notion that design activity is
neither a closed box nor an automat-
ed process, but is rather an intellectual
process in which the architect plays an
active role as a spatial choreographer.
e second sample concentrates on
a design practice, one in which pro-
grammatic and site relationships are
modeled and animated by customized
modeling and assessed by graph the-
ory based tools such as space syntax.
e main aim here is to explore how
relational thinking can be integrated
into the architectural and urban design
process. is example is important as it
regards a need for a dynamic design in-
strument that can satisfy the changing
needs in a long-term process. e ar-
chitect can use the resulting parametric
work and relational thinking to reveal
and/or meet the requirements.
In networks, nodes are not consti-
tuted from the same genre. ey can
be structured with dierent compo-
nents including not only spaces, but
also design criteria or concepts. is
is an opportunity to link tangible with
non-tangible qualities in a cognitive
process.
In terms of network thinking, the
two experiments in this study are struc-
tured through three main stages: (1)
Description of the relational structure,
(2) Analysis of this structure and (3)
Application of a rule-based design. e
rst process concentrates on achieving
an understanding of how the networks
are constituted and reveals the linking
lter that organize these complex sets
of relationships. e second process
deals with the analyzing or decoding
potentials of the constituted networks.
e third introduces a phase in which
denite metric design rules are applied
to the network of nodes. In this way,
relational structures are transformed
into spatial form from which the de-
sign proposals emerge (Figure 11).
In the fırst stage of the game main
determinant is the choicesof the stu-
dents. e constitutednetwork can
be referred toas a conceptual and
nonhierarchical one in principal. In
the example from practice however
the spatial relations are structured by
the clientspreferences and through
data arrived from userquestionnaires.
erefore in this case the network is
not only a mental construct but also
has physical impositions, yet they are
also nonhierarchical in terms of their
networking. In the following stages in
both cases, the spatial potentials of the
structured networks are expedited by
network assessment and graph theory
based tools that include space syntax.
In the process space syntax imparts
ow, transition, integration among
spatial units whereas other dynamic
network modeling whether analogue
or digital set forth clustering, neigh-
boring conditions and their meanings.
Such graph theory based tools includ-
ing space syntax appear as informative
and creative tools to think, talk about
and engage in space and spatial con-
stitutions. In the third stage we can
denote that form is designated by the
enactment of the geometric rules. e
operative rule is “to remain in the me-
dian axis of the other chosen two” in
the game described in the initial sam-
ple, and in the next sample it is the
distances designated for the units to
satisfy in reference to one another. It
is possible to say that design process is
the iteration between these stages, i.e.
the assessment of the “xed” form and
its consequences in the third stage are
examined and tested with tools men-
tioned in the second stage. erefore
the process continues with the feed-
backs of the second stage reconguring
rule sets of the third stage and rerun-
ning these relational metric rules.
Network thinking equips architects
with data regarding space and enhanc-
ITU A|Z • Vol 12 No 3 • November 2015 • N. Kozikoğlu, P. Dursun Çebi
84
es their spatial awareness. Mathemat-
ical and graphical tools render previ-
ously invisible characteristics of space
visible, measurable, and discursive. In
respect to other generative tools for de-
sign network modeling in architecture
can thus be transformed into a design
tool with which the designer can freely
think, play and model.
References
Alexander, C. (1964). Notes on the
Synthesis of Form. Harvard University
Press.
Al_Sayed, K. (2012). Can Knowledge
Inform Creativity? Paper presented
in Design Creativity Workshop 2012,
Texas A&M University, USA.
Cross, N. (2007). Designerly Ways
of Knowing. Birkhauser. Basel. Boston.
Berlin.
Dursun, P. (2007). Space Syntax in
Architectural Design. Paper presented
at the meeting of Sixth International
Space Syntax Symposium, ITU Faculty
of Architecture, 12-15 June 2007.
Dursun, P. (2012). Dialog on Space,
Spatial Codes and Language of Space.
AZ, Vol.9, 104-119.
Do, E. Y., Gross, M. D. (2001).
inking with Diagrams in Architec-
tural Design. Articial Intelligence Re-
view, Vol.15, 2001 Kluwer Academic
Publishers, 135-149.
Dovey, K. (1993). Putting Geometry
in its Place: Toward a Phenomenology
of the Design Process. Building, Seeing
and Designing: Toward a Phenomeno-
logical Ecology. Editor: David Seamon.
Suny Series in Environmental and Ar-
chitectural Phenomenology. Albany,
NY: State University of New York Press,
247-270.
Dovey, K., Dickson, S. (2002). Ar-
chitecture and Freedom? Programmat-
ic Innovation in the Work of Koolhaas/
OMA. Journal of Architectural Educa-
tion, Vol. 56 (1), September 2002, 4-13.
Hanson, J. (2001). Morphology and
Design. Paper presented at the meet-
ing of ird International Space Syntax
Symposium, Georgia Institute of Tech-
nology, Atlanta.
Hillier, B., Hanson, J. (1984). e So-
Figure 11. Talking about Two Experiments.
inking and designing with the idea of network in architecture
85
cial Logic of Space. Cambridge: Cam-
bridge University Press.
Hillier, B., Hanson, J. (1997). e
Reasoning Art: or, e Need for an An-
alytical eory of Architecture. Paper
presented at the meeting of First In-
ternational Space Syntax Symposium,
UCL, London.
Hillier, B., Hanson, J., Graham, H.
(1987). e Ideas are in ings: An
Application of Space Syntax Method to
Discover House Genotypes. Environ-
ment and Planning B, Planning & De-
sign, Vol.14, 363-385.
Hillier, B. (1996). Space is the Ma-
chine. Cambridge: Cambridge Univer-
sity Press.
Hillier, B. (1998). From Research
to Design. Urban Design Issue, Vol.68,
October 1998, 35-37.
Khan, N. (2012). Analyzing Patient
Flow: Reviewing Literature to Under-
stand the Contribution of Space Syntax
to Improve Operational Eciency in
Healthcare Settings. Presented at the
meeting of Eighth International Space
Syntax Symposium, Ponticia Univer-
sidad Católica de Chile, Santiago de
Chile, 2012.
Lawson, B. (2003). How Designers
ink. Architectural Press.
Lawson, B. (2005). e Language of
Space. Architectural Press, Oxford, UK.
Lima, M. (2011). Visual Complex-
ity: Mapping Patterns of Information.
Princeton Architectural Press.
March, L., Steadman, J. P. (1971). e
Geometry of Environment, An Introduc-
tion to Spatial Organization in Design.
London: Riba Publications Limited.
March, L. (1976). e Architecture of
the Form. Cambridge University Press.
Mitchell, W.J., Steadman, J.P., Lig-
gett, R.S. (1976). Synthesis an Opti-
mization of Small Rectangular Floor
Plans, Environment and Planning B,
Planning & Design, Vol.3, No.1, 37-70.
Nourian, P., Rezvani, S., Sariyildiz, S.
(2013). A Syntactic Architectural Design
Methodology. Paper presented at the
meeting of Ninth International Space
Syntax Symposium, Seoul, Korea.
Ostwald, M., J., Dawes, M. (2013).
Dierentiating between Line and
Point Maps Using Spatial Experiences:
Considering Richard Neutra’s Lovell
House. Nexus Network Journal, Vol.15
(1), 63-81.
Peponis, J., Zimring, C., Choi, Y.K.
(1990). Finding the Building in Way
Finding. Environment and Behaviour,
Vol.22, No.5, September 1990, 555-
590.
Schneider, S., Kuliga, S., Hölscher,
C., Conroy-Dalton, R., Kunert, A., Ku-
lik, A., Donath, D. (2013). Educating
Architecture Students to Design Build-
ings From the Inside Out: Experiences
from a Research-Based Design Studio.
Paper presented at the meeting of
Ninth International Space Syntax Sym-
posium, Seoul, Korea.
Schön, A.D. (1987). Educating the
Reective Practitioner. San Francisco:
Jossey-Bass, A Wiley Imprint.
Setola, N. (2009). A New Approach
to the Flows System Analysis in the
Teaching Hospitals. Paper presented at
the meeting of Seventh International
Space Syntax Symposium, School of
Architecture and the Built Environ-
ment, KTH, Stockholm.
Simon, H. A. (1996). e Sciences of
the Articial. MIT Press, Cambridge,
Massachusetts, London, England.
Space Syntax (2002). Tate Britain,
Report on the Spatial Accessibility Study
of the Proposed Layouts, Space Syntax
Limited, July 2002.
Steadman, J. P. (1983). Architectural
Morphology. London: Pion.
Ünlü, A., Ülken, G., Edgü, E. (2005).
A Space Syntax Based Model in Evacua-
tion of Hospitals. Paper presented at the
meeting of Fih International Space
Syntax Symposium, Del University of
Technology, e Faculty of Architec-
ture.
Wigley, M. (2007). Architectural
Brain. Network Practices, New Strate-
gies in Architecture and Design, Editors:
Burke, A. and Tierney, T., Princeton
Architectural Press, 30-53.
ITU A|Z • Vol 12 No 3 • November 2015 • N. Kozikoğlu, P. Dursun Çebi
86
Mimarlıkta ağ düşüncesi ile düşün-
mek ve tasarlamak
Bir mekan kurgusunun tasarımı
onun parçaları arasındaki ilişkiler ağı-
nın düzenlenmesi ile ilgilidir. Bu ağ ya-
pısı mimarlık söyleminde kullanıcılar
arasındaki sosyal ilişkileri, etkileşimle-
ri resmettiği, mekanda fonksiyonel ve
potansiyel rotaları deşifre ettiği, me-
kansal yakınlıkları gözler önüne serdi-
ği için önemlidir. Mimari tasarım öznel
bir süreç ise de kimi tasarım araçları ve
metotları tasarımcıya tasarlananı de-
ğerlendirmek, öğrendikleriyle yeniden
üretmek için nesnel kriterler sunar.
Ağ düşüncesi içeren mekan kurma ve
ölçme araçları mimarlığa ilişkin dü-
şünceleri görselleştirme ve tartışmaya
açmaya, verilerden mekansal ilişkilere
dair yeni oranlara ulaşmaya, mimari
programa yönelik potansiyelleri ortaya
çıkaracak çeşitlilikleri tanımlamaya ve
kriterlerle test edebilmek için senaryo-
lar geliştirmeye olanak verir.
Bu çalışma mimari tasarımda ağ
düşüncesinin kullanılmasına odakla-
nır ve tasarım aktivitesinin deneysel
ve zihinsel özelliklerine vurgu yaparak
bu tür bir düşünme biçiminin tasarım
sürecinde yaratıcı ve bilgilendirici bir
araç olarak nasıl işlevselleşebileceğini
araştırır. Çalışma, kural temelli dina-
mik ağ modelleri içindeki komşuluk
ve çekim özellikleri yardımıyla mekanı
oluşturan elemanların ilişkilerinin sis-
tematik haritalanmasına yönelik araş-
tırmalar sunar.
Mimarlıkta ağ düşüncesine odakla-
nan çalışmalar incelendiğinde temel-
de üç amaçla kullanıldığı söylenebilir.
(1) Var olan mimari biçimi anlama
(Hillier ve diğerleri, 1987; March ve
Steadman, 1971), (2) Mimari biçimi
üretme (Mitchell ve diğerleri, 1976;
Steadman, 1983), (3) Mimari biçimi
değerlendirme (March, 1976; Hillier,
1998; Space Syntax, 2002). İlkinde va-
rolan mekansal biçimlenmelerin ken-
dilerini oluşturan dinamiklerin keşfi
için analiz edilmesi hedeenir. Tanım-
layıcı ve açıklayıcı yönleri ön planda
olan bu çalışmalar mimarın mekana
ilişkin bilinç düzeyini arttırarak tasa-
rım sürecini besleyecek bilgi biriki-
mini çoğaltır. İkinci grup çoğunlukla
bilgisayar odaklı, mekanik bir süreç
içinde ve önceden belirlenmiş kurallar
bütününde istenen mekansal biçimi
aramaya niyetlidir. Burada çoğunlukla
üretilen biçimin nasıl bir yaşam biçimi
kurguladığı sorgulanmadan tüm ola-
sılıklar tasarımcının gözü önüne seri-
lir. Son grup çalışmada ise tasarımcı
üretilmiş mekansal kurgular arasında
istenen kurallar, sınır şartlarına uy-
gun en iyiyi seçme görevini üstlenir.
Burada kritik olan ve çokça eleştirilen
konu tasarımcının bu bilişsel sürece ne
denli dahil olabildiği, mekanın belirli
bir kural setini aramak ötesinde üret-
tiği olası yaşam senaryoları ile ne denli
değerlendirilebildiğidir (Nourian ve
diğerleri, 2013). Nitekim son donemde
mekan dizimi çalışmaları tasarımcıya
tasarladıkları mekansal kurguların na-
sıl yaşandığını göstererek, kendi tasa-
rımından öğrenmesine, önerisini yeni
düşüncelerle geliştirmesine olanak sağ-
lamaya, bilgi temelli tasarım sürecinin
de özünü biçimlemeye niyet etmiştir
(Hanson, 2001; Dursun, 2007, 2012).
Bu noktadan hareketle bu çalışma mi-
marlıkta ağ düşüncesinin tasarımcının
birebir dahil olduğu bir interaktif araş-
tırma süreci içinde tasarımın ilk evre-
lerinde, yaratıcı bir araç olarak nasıl
kullanılabileceğine odaklanmaktadır.
Yazıda bu olgu biri mimarlık eğitimi
diğeri mimarlık pratiğinden seçilmiş
iki deneyim üzerinden tartışılmıştır.
Bunlardan ilki mimarlıkta ağ düşün-
cesinin kavramsal olarak sorgulandığı
“ikidebir oyunu”dur. Bu atölye çalışma-
sında amaç, öğrencilerde ağ düşünce-
sine yönelik bir kavrayış ve farkındalık
geliştirmektir. Mekana ilişkin oluştu-
rulan karmaşık ağ yapısının ne tür po-
tansiyeller ürettiğinin, ağın karakteris-
tik özelliklerinin, bu ağ yapısının nasıl
görünür, tartışılabilir ve de değerlendi-
rilebilir kılındığının öğrencilerle bir-
likte irdelenmesi hedeenmiştir. Sen-
taktik ve grafik-teorik araçlar oyunda
kurgulanan ilişkiler ağını analiz etmek
için kullanılır. Bu deneysel çalışmanın
amacı mekan tasarımının belirli ku-
rallar çerçevesinde parçalarının, para-
metrelerinin karşılıklı ilişkide olduğu
bir sistem kurmak olduğunun soyut bir
model üzerinden altını çizmektir.
Yazıda tartışılan ikinci örnek ise
yerleşim ve programa ilişkin kararları-
nın dinamik ağ modelleme araçları ile
değerlendirildiği bir hastane kampüsü
tasarımıdır. Bu deneyim söz konusu
kavrayışın yani mimarlıkta ağ düşün-
inking and designing with the idea of network in architecture
87
cesinin, mekanı kurarken, tasarlarken
nasıl kullanılabileceği ile ilgilidir. Bu-
rada temsil edilen yerine deşifre edi-
lerek aranan, potansiyelleri sınanarak
geliştirilen bir mekansal kurgudan söz
edilebilir. Benzer şekilde sentaktik ve
grafik-teorik araçlar da mekanın po-
tansiyellerini çözümlemek için kulla-
nılır. Bu deneysel çalışma doğrudan,
üretilen bilgi ile sürecin beslendiği bir
mekan yapma pratiği ile ilgilidir.
Mimarlıkta ağ odaklı düşüncenin
mekansal organizasyonların ölçülme-
sine ve bir tasarım araştırması olarak
kullanılmasına yönelik olarak ortaya
konan deneysel çalışmalar mimarlığı
öğrenen ve gerçekleştirenler için de-
ğerli olacaktır.
