ETHNOCOMPUTATION

On Weaving Grammars for Architectural Design ASiS

MASSA s MnT TE
OF TECHNOLOGY
By
Rizal Muslimin SEP 22 2014
LIBRARIES
Bachelor of Architecture, Universitas Katolik Parahyangan, 1999
Master of Architecture, Institut Teknologi Bandung, 2002

Submitted to the Department of Architecture

in Partial Fulfillment of the Requirements for the Degree of

DOCTOR OF PHILOSOPHY IN ARCHITECTURE: DESIGN AND COMPUTATION

at the
Massachusetts Institute of Technology

September 2014

2014 Rizal Muslimin. All rights reserved.

The author hereby grants to MIT permission to reproduce and to distribute publicly
paper and electronic copies of this thesis document in whole or in part in any medium
now kn wn or hereafter created.
Signature redacted
Signature of Author:
Department of Architecture
Auui* I1
't

Certified by:
Signature redacted
m' - Knight
/

o ,eriyW. Knight
1Pr ssor of Design and Computation
Thesis Supervisor
Approved by: i
Takehiko Nagakura
Associate Professor of Design and Computation
Chairman, Department Committee on Graduate Students
2

DISSERTATION COMMITTEE

Terry W. Knight
Professor of Design and Computation
Massachusetts Institute of Technology

George N. Stiny
Professor of Design and Computation
Massachusetts Institute of Technology

Klaus-JUrgen Bathe
Professor of Mechanical Engineering
Massachusetts Institute of Technology
 3

ETHNOCOMPUTATION
On Weaving Grammars for Architectural Design

By
Rizal Muslimin

Submitted to the Department of Architecture on August 8, 2014 in Partial

Fulfillment of the Requirements for the Degree of Doctor of Philosophy in
Architecture: Design and Computation, at the Massachusetts Institute of
Technology

Abstract

The aesthetic, structural, and functional aspects of weaving are parallel to

the ideal Vitruvian model of an architecture that expresses an inclusive aesthetic,
resonating from the logic of the design, structure and construction process.
However, despite the strong historical connection between weaving and
architecture at their earliest inception, the ideal model of weaving does not fully
translate into most of today's architecture.

In a quest to reinvent architectural weaving, the conception of weaving is

challenged, reconfigured, and reformulated through in situ observation and
computational design formulation. This dissertation proposes methodologies to
interpret tacit knowledge in traditional weaving through an EthnoComputation
lens and to reinvent architectural weaving with two Weaving Grammars.

Thesis Supervisor: Terry W. Knight

Title: Professor of Design and Computation
4

ACKNOWLEDGEMENT

First and foremost, I would like to express my sincere gratitude to my

advisor, Terry Knight, who through her patience, motivation, and friendship has
guided and supported me throughout the course of my doctoral studies. I am very
grateful to have had the rare opportunity to assist her in teaching computational
design courses and conducting shape grammar research in Rio de Janeiro,
experiences from which many ideas for this thesis sprang. Witnessing her
meticulous class preparation and thorough research was inspirational; Terry is a
role model for academic excellence. Her careful critique of the draft of this
dissertation and her illuminating insights were hallmarks of ideal academic
mentorship-I could not have asked for a more supportive advisor. Along with
offering academic support, she cared for my well-being as an international
student, for which I cannot thank her enough. My family and I are blessed to have
her as an advisor, mentor, and friend.

I am also very grateful to George Stiny for his inspiring and enlightening
ideas. Many concepts in this thesis were fostered through my informal discussions
with him. I am also thankful for the chance to assist him in class. Seeing him from
behind the stage, I learned a great deal about how to deliver ideas consistently and
uses gentle pace to help students digest complex concepts. The more I understand
his concept, the more clearly I see my role as a scholar and as an architect. It will
take me years to become a shape-grammarist of his stature, but our conversations
have helped me build a strong foundation from which to explore other aspects of
shape grammar.

My warm thanks to Klaus-Jurgen Bathe for serving on my dissertation

committee. Prof. Bathe's warm personality and welcoming demeanor helped me,
as a non-engineering student, to feel less hesitant; I was just as comfortable asking
basic questions on finite element method as I was discussing high-level aspects of
my thesis. I truly appreciate the time he spent discussing my architecture thesis,
carefully analyzing my ideas and providing many valuable comments. His support
and insight revealed an area of design that is critical for engineers but often
overlooked by architects like me-a discovery for which I am most grateful.

I would also like to thank the other Computation faculty: Takehiko

Nagakura, for introducing me to scripting and sharing with me ways to make "Big
Data" meaningful; Larry Sass, for familiarizing me with digital fabrication; and the
late Bill Mitchell, for being so responsive and generous in sharing his views.
Discussions with other MIT professors also helped significantly in assembling the
framework for my study. I am grateful to Erik Demaine, Pavan Sinha, Marvin
Minsky, Patrick Winston, Michael Fischer, and to former professors Leah
Buechley and Lucy Suchmann.

Craftsmen in Kete Kesu, Toraja, have been welcoming and enthusiastic

during my on site research. I am thankful to Ting Ting for introducing me to the
local craftsmen; to Karaeng Toding for showing me the other dimension of
weaving; and to Acho, Ba, Nandong, Kadir, and Taufan for sharing with me their
 5

excellence at their craft. Their unspoken knowledge and demonstrated skill gave
me enough material for a lifetime of future ethnocomputation studies.

I am particularly indebted to Nathalie Miebach for showing me her

magnificent fiber arts; as an architect, I'm awed by and jealous of her collection
and her talent. Nathalie has been very generous in allowing me to join her 3D
weaving workshop, inviting me to share my works in her course at MassART, and
introducing me to other weavers in Boston such as Lois Russell and Jeanne
Flannagan. My conversations with Nathalie inspired me to think of what
architectural weaving can do to support handicraft weaving, instead of what
handicraft weaving can do to further the field of architecture. I also thank Olga
Popovic Larsen for our interesting discussion on reciprocal frames in Copenhagen
and Roxana Waterson of Singapore for her anthropological insight on
ethnocomputation.

MIT staff members also supported me during my studies, providing

invaluable services. I truly appreciate Renee Caso's tremendous assistance and
her immense patience in responding to my questions and confusion during my
time at MIT. I also thank Rebecca Chamberlain for warmly and patiently assisting
me in dealing with my financial paperwork. I am very grateful to Daniella
Ashbroke-Guichard and everyone at MIT-ISO for helping me navigate
immigration issues, and I appreciate the Cron crews for keeping all the machines
humming along. I also thank the MIT writing center and my editors, Amanda and
Patricia, for helping me improve this thesis and make it more readable.

I greatly appreciate the financial support offered by my sponsors: the

Fulbright Foundation, the MIT teaching assistantship, the Indonesian ministry of
education (Dikti), and MIT-SUTD. I also thank the institutions that gave me
financial awards for activities related to my doctoral studies: Harold and Arlene
Schnitzer award, the Avalon travel grant, the GSC travel grant, Brickstainable
(PVC bricks), and the South-East Asia Grant.

I am grateful to the Design and Computation Group for their constructive

criticism and friendly advice. I am thankful to Asli, Derek, Daniel Rosenberg,
Felecia, Onur, Katia, Juhong, Theodora, Rachel, Dina, Ki, Daniel Smitchwick,
Duks, and many others. Their camaraderie and daily conversation is inseparable
from the development of this thesis. I also thank various alumni, including Daniel
Cardoso, Mine, Gaby, Lira, Kenfield, Jose, Andrew, Steffan, Skylar, Murat, Adela,
Joseph, German, Stylianos, and Vernelle for sharing ideas and discussion, and
simply for keeping me in the loop.

I am glad to be part of the Indonesian family in the United States. At MIT, I

thank Poetro, Nova, Jonathan, and the other AIS-MIT fellows. In greater Boston, I
thank Panji, Aris, Djunaedi, Najib, Daniar, Naim, Aryo, Sukidi, and many more.
Their friendship and solidarity has helped my family and me to feel at home here.
Rezy Pradipta has been immensely helpful in assisting me with adapting some
engineering courses. Graha, Desta, Ipung, and Hadi's hospitality has helped my
family and me settle in and feel comfortable here.
6

I feel very fortunate to have parents who have always been supportive and
encouraging, helping me reach my goals. Despite my absence during some critical
moments and their own difficulties, they have consistently helped and supported
me. This degree serves as a form of thanks for their effort and support, although
nothing can truly repay their sacrifices or thank them enough for raising me well.
I also appreciate my in-laws for giving their love and support to my family and me
from a distance.

Finally, I am very grateful to my wife, Ayu, for her love, friendship,

appreciation, and, most of all, patience; I am blessed to share my life with her. A
walk from MIT campus to home is always a wonderful journey, knowing that she
will be there, and my short trip to campus from home is always pleasant, knowing
that she fully supports me and expects my studies to go well. Her independence
and ability allow me to stay focused on my studies, knowing that my children are
in good hands.
 7

To Anevay andRyuzen
8

TABLE OF CONTENTS

1 INTRODUCTION 11
1.1 Background 11
Rational Weaving 12
"Educultural" Weaving 13
1.2 Issues in Architectural Weaving 14
Dormant Beauty in Architectural Weaving 15
Design Fixation 18
1.3 Research Statement 21
1.4 Methodologies 22
Shape Grammar 22
Finite Element Method 24
In Situ Research 24
Perception System Theory 25
1.5 Scope of this Work 25
The Method, not the Subject 25
The Process, not the Product 26
1.6 Thesis Outline 26

2 ETHNOCOMPUTATION 31
2.1 Outline 31
2.2 Introduction 33
The-Visual in Visual Ethnography 34
The Computation in Ethnocomputing 39
The EthnoComputation Approach 42
Situating the Computational Framework 44
2.3 Seeing-in-There 50
Site as a Schema 50
Reconfigured Sites 52
2.4 Mediated Bamboo 54
Generative Bamboo 56
Reciprocal Causality in Indigenous Materiality 64
2.5 The Weaver and the Weave - Sensorial Weaving 65
Sensorial Computation 66
Permanency and Temporality in Mental Shapes 70
Haptic Shape and Haptic Ambiguity 75
Oriented weaving 80
Symmetrical Stimulus and Asymmetrical Reaction 85
 9

2.6 Discussion 86
Sensing and Calculating in EthnoComputation 87
Hylomorphism Revisited 87
Ethno-Computation as Cultural Mimesis 89

3 COMPUTATIONAL WEAVING 91
3.1 Outline 91

3.2 Basic Weaving Grammar: On the Language of Weaving Design 93

Introduction 93
Phase 1: Defining Weaving Intention 94
Phase 2: Developing Weaving Action 97
Phase 3: Evaluating Weaving Condition 100
Discussion 103
Case Studies 111
3.3 Abstract Weaving Grammar: On Extending the Language of
Weaving Design 126
Introduction 126
FEM and Shape Grammar 126
Knot in FEM-Kawabata's Schema Revisited 128
Knot in Shape Grammar-Boundary Rules 131
Knot in FEM-Shape Grammar-Perceptual Optimization 133
The Children of Weaving 140
Case Studies 143
3.4 Discussion 155
Traditional Computation in the Computational Tradition 155
The Expanding Weaving Language 156
Grammar Assimilation across Cultures 156

4 CONCLUSIONS 159
4.1 Summary 159
EthnoComputation as a Fieldwork Companion 159
The Roundtable in Mind 160
4.2 The Limits and the Prospects 161
Augmented-Me: The Future of EthnoComputation 161
Digital Shape: On Weaving and Automation 162
4.3 Borderless Weaving 164
4.4 Epilogue 165

BIBLIOGRAPHY 169

LIST OF FIGURES 175

10
 11

1 INTRODUCTION

The aesthetic, structural, and computational aspects of weaving exemplify

the ideal Vitruvian model of an architecture that expresses an inclusive aesthetic,
resonating from the logic of the design and the construction process.

Weaving and architecture were strongly connected at their earliest

inception. However, despite this strong historical connection, the ideal model of
weaving has not been fully translated into today's architecture. The application of
weaving in architecture will be referred to here as architecturalweaving. In
today's architecture, the structure-aesthetic relation in architectural weaving
remains fragmented. Moreover, the progress of architectural design inspired by
weaving is far behind the evolution of weaving in traditional settings.

This dissertation proposes methodologies with which to interpret tacit

knowledge in traditional weaving through a computational lens and to reinvent
architectural weaving with hybridized schemas from both a traditional
environment and computational principles. In a quest to reinvent architectural
weaving, the conception of weaving is challenged, reconfigured, and repositioned
through in situ observation and computational design formulation.

1.1 Background

Weaving has been coevolving with human civilization, from the production of
early survival tools, such as shelter and clothing, to the automation of weaving in
the Jacquard loom. As the terms 'technology' and 'textile' are both derived from
the Latin word texere, which means 'to weave', 'to connect' and 'to construct', the
use of weaving can be traced back to the origins of textiles and architecture." 2
Gottfried Semper, in The FourElements in Architecture, states that the origin of
architecture overlapped with the creation of textiles at the time when humans

'John Gillow and Bryan Sentance, World Texdles:A Visual Guide to TradidonalTechniques(Thames
&

Hudson, 2005).
' Henry Barker, Egypdan Gods & Goddesses (Grosset & Dunlap, 1999).
12

invented the interwoven fence as the earliest partition for an enclosure. 3 This then
led to the invention of other woven objects for domestic household purposes. Frei
Otto makes a corresponding argument that early humankind's first dwelling was
woven with flexible and vegetal materials (young conifer, bamboo, or branches of
broad-leaved trees) because these materials are easily harvested and manipulated
with bare hands.4 Today, the legacy of weaving with flexible vegetal material can
still be found in many traditional crafts and traditional house constructions. In
some cultures, weaving is frequently used for wall-partitions, sometimes with
567
ornament that is applied in the textile's woven pattern, to decorate the house.

Rational Weaving
In addition to the role of weaving in architecture and craft, the logical aspect of
weaving has allowed it to travel across disciplines. The rationality of weaving is
construed from its discrete and continuous property. On one hand, the checkered
pattern from the over-and-under configuration appears as an assemblage of pixels
that is countable in the interwoven surface. On the other hand, the continuous
material (i.e., thread) allows for the depiction of a continuous line or plane made
by one thread as it passes over or under other threads.

The association of continuous material and discrete pixilation makes the

technique of weaving popular and effective in forming images or patterns in
textiles. It is this pixilation pattern that allow weaving to be converted into a
computable element. The Jacquard Loom machine, demonstrated by Joseph-
Marie Jacquard in 1801, became one of the most celebrated milestones in
computer science. Jacquard substituted for the weaver's role in handling the warp
and weft by using holes in a punched card as the pattern-keeper that instructed
the weft to go over or under the warp when looming the patterns.8 9 The Jacquard

I Gottfried Semper, The FourElements ofArchitecture andOther Wridngs, Reissue (Cambridge University
Press, 1851).
4
Frei Otto, IL-37AncientArchitects(UniversitftStuttgart, 1994).
sMiriama Evans, The Art ofMiori Weaving: The EternalThread(Wellington N.Z.: Huia Publishers, 2005).
'Martin Prechtel and Robert S. Carlsen, "Weaving and Cosmos Amongst the Tzutujil Maya of Guatemala,"
RES: Anthropolog andAesthedcs, no. 15 (April 1, 1988): 122-32.
7Kate P. Kent, "West African Decorative Weaving," African Arts6, no. 1 (October 1, 1972): 22-88,
doi:10.2307/3334637.
I T. F. Bell, JacquardWeaving andDesigning(Longmans,Green, 1895).
 13

loom started the era of automation, and many computational pioneers began
working with punched cards. Charles Babbage adapted the punched card
mechanism for his Difference Engine and Analytical Engine designs in 1820, and
in 1845 Herman Hollerith embedded census data into the pattern of holes used for
his tabulator machine. Using Hollerith's mechanism, IBM tabulation machinery
was developed, and IBM's card reader and cardpunch machine assisted the birth
of ENIAC, setting off the digital era in computer science."

Today, decades after punched cards have been replaced by circuit boards,
the computational aspects of weaving remain ubiquitous in multidisciplinary
fields." Experts in electronics and biotechnology make broad use of weaving
principles, including identifying proteins using knot-theory, knitting articulate
cartilage in tissue engineering and knotting conductive thread into circuit
boards.2 1314 15 In mathematics, the formal expression of numerical weaving has
been formulated in algebra, arithmetic and trigonometry equations - namely, in
knot theory, braid theory and topology. 16 17

"Educultural" Weaving
As part of everyday craft activity, weaving has been passed down through
generations with various purposes in different cultural contexts. In Zinacantan,
Mexico, for instance, cultural customs exist that lead to improved weaving ability.
As observed by Patricia Greenfield, Zinacantec infants are trained to be natural

'James Essinger,Jacquard'sWeb: How a Hand-Loom Led to the Birth ofthe InformationAge (Oxford:
Oxford University Press, 2004).
10
Norman Macrae, John Von Neumann: The Scientifc Genius Who Pioneeredthe Modem Computer, Game
Theory, NuclearDeterrence, andMuch More (AMS Bookstore, 2000); Essinger, Jacquard'sWeb.
" Liu Suyi, "Fabric Weave Design Based on Cellular Automata Theory," May 15, 2009, 145-47,
doi:10.1109/IUCE.2009.88.
" Franklin T. Moutos, Lisa E. Freed, and Farshid Guilak, "A Biomimetic Three-Dimensional Woven
Composite Scaffold for Functional Tissue Engineering of Cartilage," Nat Mater6, no. 2 (February 2007): 162-
67, doi:10.1038/nmatl822.
Nat Mater6, no. 2 (February
' Gerard A. Ateshian, "Artificial Cartilage: Weaving in Three Dimensions,"
2007): 89-90, doi:10.1038/nmatl830.
' 4 K. Murasugi, "Knot Theory in Molecular Biology," Knot Theory&Its Appicadons, 2008, 267-83.
' Mahiar Hamedi, Robert Forchheimer, and Olle Inganas, "Towards Woven Logic from Organic Electronic

Fibres," Nat Mater6, no. 5 (May 2007): 357-62, doi:10.1038/nmatl84.

' Alexei Sossinsky, Knots: Mathematicswith a wist (Harvard University Press, 2004); P. Gerdes, "Fivefold
Symmetry and (basket) Weaving in Various Cultures," FivefoldSymmetry, 1992, 245; Ada K Dietz, Algebraic
Erpressionsin Handwoven Textiles (Louisville, Kentucky: The Little Loomhouse, 1949).
" Rizal Muslimin, "One-Piece Weaving: Reconfiguring Folding and Knotting Algorithm in Computational
Design" (presented at the the 16th International Conference on Computer Aided Architectural Design
Research in Asia (CAADRIA), Newcastle, Australia, 2011), 9-18.
14

weavers from the time they are newborns. The infants are swaddled and wrapped
with cloth so that their upper arms are always close to their bodies. This
conditioning helps build a posture that enables them to weave easily with the
Zinacantec backstrap looms they use in their adult life. Additionally, Zinacantec
children are introduced to toy looms as educational playthings at age three to
seven. In their adult life, carrying a heavy load (of firewood) on their backs helps
Zinacantec girls and women improve their balance for weaving. Nurturing such
physiological capabilities from childhood provides a deep foundation for
Zinacantec weavers to be more stable in backstrap loom weaving, while their
childhood experience of sustaining visual attention (watching their mothers and
sisters weave) prepares them to endure large weaving tasks, which usually take
several weeks.I

While the Zinacantec developed their childhood education to support the

practice of weaving, in another context it is the practice of weaving that helps
support the children's education. According to Friedrich Froebel, the inventor of
kindergarten, weaving cultivates the muscles in children's fingers, strengthening
them and giving them greater flexibility, while training the children's vision to
measure angles and distances. As they exercise their aesthetic tastes in selecting
harmonious colors and patterns, their minds gradually gain the ability to work
with objects with different relationships and properties. Froebel uses the playful
logic of paper interlacing and mat weaving for his fifth and sixth occupation
methods. 0 In addition to perception and motoric skills, Froebel's mat-weaving
also stimulates children's intuition of numbers and arithmetical logic, as the child
is obliged to count verbally while weaving.

1.2 Issues in Architectural Weaving

The brief background above shows the versatility of weaving in early architecture,
computer science, and educultural domains. Given this versatility, the continuity
of architectural weaving, from its inception to today's architecture, might have

1 Patricia Marks Greenfield, Weaving GeneradonsTogether:Evolving Creadvityinthe Maya of Chiapas,

illustrated edition (School of American Research Press, 2004).
" Maria Kraus-Boelte and John Kraus, The KindergartenGuide: The Gifts.-V.& the Occupadons(Nabu Press,
2010).
 15

been seamless. Most architects today are consistently in pursuit of an inclusive

design, where function, structure and aesthetics are logically interrelated within a
building. This inclusive integration corresponds well with the values in weaving.

The inclusive model found in weaving, however, does not translate in

most examples of recent architectural weaving. The relation between structure
and the aesthetic features in some architectural weaving today is fragmented as
the beauty of the building does not come from the structural expression. For
instance, faqade components may expresses weaving pattern or texture, but these
components are not structurally woven. Moreover, some of the aesthetic
expressions in this fragmented practice are trendless, bluntly mimicking the style
from handicraft weaving. Existing weaving styles are magnified several times, a
practical adaptation, but one that does not necessarily consider how these designs
will perform when adapted to architectural scale.

This fragmented weaving and the trendless styles expose only the surface
of the deeper problems. When we look beneath the surface issues of architectural
weaving, we find issues that may be more conceptual than technical in nature. In
particular, problems may arise from the conceptions under which weaving has
been institutionalized, in both scholarly and industrial contexts.

In order to understand what has been left out of architectural weaving, we

need to understand the difference between the features of handicraft weaving and
the features of weaving that have been applied in architecture. To understand
these issues more comprehensively, in the following segment I discuss the
fragmented practice in architectural weaving, followed by the discussion on the
fixation on weaving conceptualization.

Dormant Beauty in Architectural Weaving

In handicraft, weaving can be experienced through different modalities: visually,
by admiring the pattern; physically, by feeling the texture and the stiffness;
functionally, by utilizing the weave. Moreover, through the integration of visual,
physical and functional qualities, the Vitruvian principles in weaving can be
experienced intellectually.
16

However, in most architectural weaving, aesthetic expression is singled

out from the rest of weaving qualities. This exclusion keeps weaving's Vitruvian
principles dormant. In the following, I illustrate some examples on the aesthetic
domination over structure in cases where building components are pseudo-
woven. Subsequently, I contrast the trendless styles in architectural weaving with
the progressive styles in handicraft weaving.

Pseudo-Woven
Despite the wide application of weaving's aesthetic in architectural projects, the
structural features of weaving have not followed the aesthetic features at the same
pace. Projects such as the Aragon Pavilion for World Expo 2008 in Zaragoza, the
Spain Pavilion for World Expo 2010 in Shanghai, and the Center Pompidou Metz in
Paris exhibit interwoven expressions aesthetically. Yet, on closer inspection, the
components that express interwoven style in these projects are not structurally
interwoven (Figure 2). In the Spanish pavilion, the weaving expression in the
facade is assembled of curved woven-mat modules attached to the steel structure
behind. In the Zaragoza pavilion, the wavy glass faqade (constructed so as to
create a weave-like impression), is sitting on a layer of zigzagged mullions. The
Pompidou Metz probably bears the closest resemblance to the original interwoven
pattern, as the triaxial timber pattern, in addition to its aesthetic function, does
serve as a structural component. However, looked at in detail, the triaxial timber
beams are basically stacked and layered one over another. In a traditional
weaving structure, the component is not only stacked in an over-and-under
configuration with the neighboring component but is also self-supporting with the
others in a cyclical composition, in that some of the distributed load from one
component to another will return back to itself.

Trendless Repetton
In addition to the lack of aesthetic-structural integrity in architectural weaving,
the references to weaving styles in architecture are still limited compared to the
diversity of weaving styles in handicrafts such as basketry and textiles. Not only
does handicraft weaving show greater variety of style compared to architectural
weaving, it is also more progressive. The structural logic and aesthetic expression
in handicraft are well integrated and mutually resonant in the progression. For
 17

instance, the works of fiber artists such as Mary Jackson, Nathalie Miebach, and
John Garrett in basketry, as well as Anni Albers and Stella Hicks in tapestry,
exhibit new aesthetics of fiber design derived from ancient weaving techniques
(plaiting, coiling, knitting, twilling, twinning, stitching, ikat, tatting, braiding,
beading and many others). Permutations and combinations of various weaving
techniques, supported with new materials and technology, have spawned
numerous weaving styles in fiber art.

Indeed, the use of non-woven structural systems to support weaving

expression is legitimate in architecture. Also, nothing is wrong with mimicking
the exact style from the handicraft. My question from these practices is more
about why weaving has yet to accommodate the Vitruvian model in architecture,
given its corresponding principles. Moreover, with such progression in handicraft
weaving, why is architectural weaving development relatively stagnant?

Understandably, a technical issue, in particular the characteristics of the

material to be woven, is commonly regarded as one of the main reasons why the
weaving diversity we see at the handicraft scale is difficult to adapt to the
architectural scale. The vegetal materials used in handicraft weaving (e.g., reed,
bamboo), which can be acquired easily from the surrounding environment, are
flexible enough to be manipulated by hand (cut or bent), hence liberating the
weaver to create various weaving configurations. In contrast, in architecture, the
vegetal materials are considered inadequate to meet certain durability standards
at building scale (e.g., structure, climate and safety). Additionally, handling
industrial materials that are generally used in architecture is not as easy as
manipulating the vegetal material used in in handicraft (Figure 3). Thus,
architects might be encouraged to apply the material in architecture in the same
style as the material is commonly applied in handicraft.

The examples above indicate that the problem in architectural weaving

might be less about technical issues of how to apply weaving, and more about
conceptual issues of perceiving whatweaving is. Specifically, the problem is a
general conception that registers weaving exclusively to the realm of handicraft
and not to that of architecture. In the following, I explain how the conceptual
18

issue of weaving-architecture incompatibility is rooted even deeper in handicraft

categorization.

Design Fixation
The psychologists Jansson and Smith coined the term "design fixation" to argue
that showing designers a picture of a potential design solution to a problem, prior
to a design process, could result in fixation - that is, as a precedent, the picture
would block access to the other ways of solving the problem.' In supporting this
theory, Jansson and Smith posit that a design process is operating with two types
of mental representations of the design problem:

Conceptualspace: representation of abstract knowledge about principles,

concepts and rules that can be used to solve the problem.

Objectspace: representation of particular physical objects and elements

that could form the physical realization of a solution to the problem.

Within these two mental spaces, design fixation occurs when a designer
perceives a certain design precedent in the object space. The precedent solutions
provided in this space will prevent the designer from moving to the conceptual
space, where a new solution could produce an innovative design.

This theory might explain design fixations in architectural weaving, where

some precedents from handicraft weaving seem to prevent architects from
reinventing another way to weave on the architectural scale. Therefore, an
architect may, in effect, copy and paste the precedent from handicraft into
architectural scale as is. This study considers that fixation in architectural weaving
may also be affected by the enumerative format in weaving studies and the well-
established mechanization in weaving manufacturing.

" David G. Jansson and Steven M. Smith, "Design Fixation," Destgn Studies 12, no. 1 (January 1991): 3-11,
doi:10.1016/0142-694X(91)90003-F.
" A. Terry Purcell and John S. Gero, "Design and Other Types of Fixation," Design Studes, Special Issue:
Design Cognition and Computation, 17, no. 4 (October 1996): 363-83, doi:10.1016/S0142-694X(96)00023-3.
 19

Enumerative Catalogue:FragmentedStyles
A long list of weaving types has been documented and categorized by many
scholars. For instance, in categorizing lightweight structures, Frei Otto defines
various types of netting.13In exemplifying basic types of textiles, Annie Albers
defines several types of weaves (e.g., coiling, netting).2 4 In proposing weaving as
the first element in architecture, Gottfried Semper distinguishes several basic
types of weaving (e.g., knitting, plaiting, and knotting).2 In cataloguing basketry
across culture, Bryan Sentance geo-tags different styles of basket weave based on
location (e.g., African weaving, Navajo weaving, Japanese weaving). 26 This
enumerative weaving documentation is important to exhibit the diversity of
weaving as well as to promote the craft of related culture.

Nevertheless, beyond these meticulous catalogues, certain information

that is needed for design purposes is not addressed by the enumerative format.
The broad spectrum of stylistic weaving in the catalogues provides many
precedents for an architect to apply architectural weaving in the object space
mode (according to design fixation theory). However, the knowledge needed to
learn the basic principles and rules of weaving--to reinvent architectural weaving
in the conceptualspacemode-is not available in the enumerative format. Some
books give step-by-step instruction for how to build a specific weaving style, yet
this instruction does not describe the weaver's design method - how he or she
invented the style.

In addition, the absence of other types of crafts in the catalogues that use
weaving principles but with different names, materials, or scales, could also cause
design fixation. These exclusions could cause any potential ideas and principles
from the so-called 'not weaving'to be perceived as incompatible with architectural
weaving, and thus, would not be considered in the architect's conceptualspace
during the design process.

2
Berthold Burkhardt, Jugen Hennicke Frei Otto, IL 10 - GridShells(Institutefor Lightweight Structures,
University Stuttgart, n.d.).
24
Anni Albers, On Weaving(Wesleyan University Press, 1965).
* Semper, The FourElements ofArchitecture and Other Wridngs.
2 Bryan Sentance, Basketry:A World Guide to TradidonalTechniques (Thames & Hudson, 2007).
20

MechanicalFixation:Binary Weaving
To understand how far weaving design has evolved, it is necessary to distinguish
design innovation from manufacturing revolution.

Weaving automation, as a means to delegate the role of the weaver's body

to an external apparatus, may have begun in prehistoric times. Prehistoric
weavers at that time invented ways to fix several threads in one orientation all at
once (warp) with a cane or hanging branch so they could weave another thread
(weft) in perpendicular orientation to the warp (this process later became known
as looming).

In their time, the prehistoric weavers' precedents probably came from

random weaving in nature (e.g., birds' nests). In revolutionizing weaving design,
they were able to think beyond these precedents by simplifying the random
weaving in nature into a two-axis plain weave that remains pervasive today (i.e., a
two-strands rectilinear weave composition).

This prehistoric invention liberated the weaver's hand from handling

several vertical warps at once and let the weaver focus solely on sliding the
horizontal weft through the warps. Subsequently, for the next few thousands of
years, through a series of looming mechanization advances (e.g., frame-support,
shuttle, and yarns), the process of sliding the weft and lifting the warp remained
in the weaver's hand until the beginning of the 19th century. As the role of the
hand was delegated to the punched card in the Jacquard loom, weaving became
hands-free. Thus, in terms of manufacturing, weaving mechanization has been
advanced through many generations since prehistoric weaving.

However, in terms of design, most of today's industrial weaving is

essentially weaving the same design that was invented by the prehistoric weaver

" Mattie Phipps Todd, Hand-Loom Weaving:A ManualforSchoolandHome (Rand, McNally & company,
1902); Alfred Barlow, The Historyand Principlesof Weavfn, by Handandbypower(London:Low,
Marston, Searle, & Rivington, 1878).
The conception of automation used in this framework is based on Bruno Latour, "'"Where Are the Missing
Masses? The Sociology of a Few Mundane Artifacts,"' in ShapingTechnology/ BuildingSociety:Studies in
SociotechnicalChange, by wiebe E. Bijker and John Law (Cambridge, Mass.: The MIT Press, 1994); Bruno
Latour, "Gaston, A Little Known Successor of Daedalus," trans. Charis Cussins, n.d.
 21

thousands of years ago: plain weave and its variations (e.g., basket weave, satin,
twill, etc.). A contributing factor in the design fixation on this mechanization
might be the hands' lessening involvement in weaving over time. Consider a
weaver holding a thread in each hand. There are many ways for the weaver to
weave the threads in 3D space to create points, linear, planar, or volumetric
objects (e.g., a knot, a braid, a mat or a basket). However, the first weaving
automation into looming effectively reduced the number of possible ways to
weave in 3D space into 2D space (fabrics) with 1D linear motion (i.e., by sliding
the weft only horizontally). The weaving advancement in the Jacquard loom
completely lifted the hand off the thread and only left patterns of OD points on a
surface, ready to be flipped between '0' and'1'to portray a pixilated 2D image --
hands off; codes on.

Thus, if the enumerative catalogue effects design fixation visually, then

weaving automation effects design fixation mechanically.

1.3 Research Statement

I have explained some issues in today's weaving application in architectural

design, such as pseudo-weaving and trendless repetition. I have also discussed
some potential issues in reinventing architectural weaving, such as weaving
mechanization and weaving enumeration. On that basis, strategies to reinvent
architectural weaving are outlined as follows: [1] To prevent the mechanical
fixation, we need to shift attention back to the condition before the weaving
process was mechanized and digitized (i.e., hand-weaving), and focus more on
interpreting weavers' methods during the weaving process. [2] To prevent design
fixation, we need to pair the enumerative representation of weaving productswith
a generative representation of weaving processesthat could demonstrate the basic
principles and rules of weaving design. Using these strategies could help
architects to analyze existing weaving designs and synthesize new weaving
designs for the architectural scale.

To pursue these strategies, the main objectives of this research are the
following:
22

Interpretation. In order to analyze weaver methods before weaving

mechanization, the analysis would need to take place in a traditional
setting where non-mechanized weaving practice and natural weaving
materials are available. Traditional weavers often make use of tacit
knowledge. To anticipate and capture tacit knowledge in the traditional
setting, this study develops EthnoComputatdon,a research method to
interpret a weaver's actions and perceptions into explicit representations.

Reinvention. To create a generative representation that resonates with the

basic principles of weaving, the representation would need to be obtained
from a rule-based design that encapsulates characteristics of weaving
designs. This study formulates two Weaing Grammars:Basic weaving
grammar and Abstract weaving grammara set of design rules that is
capable of characterizing existing weaving design as well as generating
new weaving designs.

1.4 Methodologies

To support the interpretation and reinvention strategies, four research

methodologies are used -- Shape Grammar, FEM, In Situ Research and Perceptual
System Theories. These four methodologies function both as an analytical
platform to interpret the logic of weaving based on weaver-weave interaction, and
as a synthesis platform to demonstrate the reinvention of weaving into
architectural scale.

Shape Grammar
Shape grammar, originated by George Stiny and James Gips in 1972, is a
computational design method used to analyze and synthesize designs by
embedding and calculating shapes with a set of shape rules.28 A shape rule is
visually represented by a left-hand shape and a right-hand shape (the new shape).
A shape in the rule can be defined by a certain parameter or description (e.g.,

' George Stiny and James Gips, "Shape Grammars and the Generative Specification of Painting and
Sculpture," Informatdon Processing71,no. 1460-65 (1972): 125.
 23

variables, conditions and data set). A shape computadon represents the iteration
of shape rules in a step-by-step notation to show how the rules generate designs.

left-hwAn right-hand

shape rules initial shape

shape computation

Figure 1. Shape Grammar Computation

"

Most shape grammar studies for design analysis are represented by

defining the shape rule first, and then showing how the rules generate designs in a
shape computation. In addition to this procedure, the present study also uses each
shape rule and shape computation as a stand-alone function to represent
information from traditional settings and knowledge from different disciplines.

For representing knowledge from traditional settings, such as crafting

activity, shape computations can independently serve as a recording instrument
to visually and explicitly document the iterative process of craftspeople's
activities. With this explicit representation, the iterative process of the
craftsperson becomes more amenable for computational analysis.

In representing knowledge from different disciplines, such as descriptions

from applied sciences, shape rules can accommodate certain scientific
descriptions in the shape rules. In the form of shape rules, the descriptions from
the other disciplines can be more easily be incorporated into a design synthesis.

"From: Terry Knight, "Computing with Emergence," EnvironmentandPlanningB: PlanningandDesign30,

no. 1 (2003): 125-55, doi:10.1068/b12914.
24

(More details regarding shape grammar's principles will be presented in chapter

2.)

Finite Element Method

Finite Element Method (FEM) is a method for analyzing the structural response of
a physical object that has been idealized into a virtual model (i.e., an assemblage
of vertices, edges, and faces called finite elements, which define the shape of the
object). Given the volume of the object, boundary conditions, degree of material
elasticity (Young's Modulus), and the applied external load on the model, FEM
solves the structural response of the idealized object by satisfying the Equilibrium
Solution (based on Hooke's Law) of each element inside the body and on the
surface.

This research uses FEM at two levels: the application level and the
conceptual level. At the application level, this study uses FEM simulation
software and compression testing to analyze the structural behavior of rule-based
weaving design under different design rule parameters. At the conceptual level,
this study revisits the FEM idealization process in converting a physical object to a
virtual model. This conversion is represented with shape grammar notation, in
order to integrate computational design and computational mechanical features
early in the process of designing architectural weaving. This integration is used to
analyze and synthesize potential weaving derivations in the other type of crafts
(e.g. tensegrity, beading, folding, etc.). (More details on the Finite Element
Method idealization process will be presented in chapter 3.)

In SituResearch
This study uses three types of in situ research: (1) field observation, (2) a weaving
grammar workshop and (3) a building installation workshop, to observe
traditional weaving knowledge on the site and to inspect the adaptation of the
rule-based design on the site.

31
Klaus-Jurgen Bathe, FiniteElement Procedures(Englewood Cliffs, N.J: Prentice Hall, 1996).
* Miguel Luiz Bucalem and Klaus-Jurgen Bathe, The MechanicsofSolidsand Structures- Hierarchical
Modelingand the FiniteElement Soludon, 2011 edition (Berlin; New York: Springer, 2011).
 25

FieldObservation.The observation included a generalobservationof the

surrounding environment on the site to note activities and elements
related to weaving, and specific observationof the weaver's interaction
with the weave during the making process.

Weaving GrammarWorkshop. This workshop included teaching rule-

based weaving design to groups of participants to evaluate their degree of
comprehension in understanding a rule-based design approach.

BufdingInstalladon Workshop. In this installation, a group of local

carpenters built a mock-up of an architectural weaving installation in the
field to evaluate the degree of applicability of a rule-based design during
the construction process in a traditional context.

The primary site for my in situresearch was Kete Kesu, a traditional

village in Toraja, Indonesia, chosen for its rich diversity of crafts. A secondary site
of my field study was Cambridge, Massachusetts, where I conducted a 'Weaving
Grammar' workshop for a group of 5s to 9t graders to assess weaving grammar's
universality across age groups.

Perception System Theory

This study also incorporates a theory of perception to analyze the traditional
weaver sensorial experiences during the weaving process: James Gibson's
'Perceptual Systems'.Y

1.5 Scope of this Work

The Method, not the Subject

The scope of research in the field study is microscopic, concentrating on a specific
interaction between a traditional weaver with a specific weaving design, parsed in
a very definite moment, and geographically confined in a remote location. Thus,
although the weaving procedures in this definite moment are omnipresent

' James Jerome Gibson, The SensesConsideredasPerceptualSystems(Boston: Houghton Mifflin, 1966);

James Jerome Gibson, The EcologicalApproachto VisualPerception(Routledge, 1986); E. Bruce Goldstein,
"The Ecology of J. J. Gibson's Perception," Leonardo14, no. 3 (Summer 1981): 191-95.
26

beyond geographical boundaries and across cultures and periods, the ubiquity of
weaving should not position this study as a generalization of weaving.

Through the field research, this thesis should be seen as a proposal for a
design method on how to learn design from the weaver - here, in particular, from
the tacit environment of traditional weaving. The research should not be
evaluated with a statistical mindset so as to legitimate a certain weaving method
based on quantitative findings. The value of this study should be appraised
according to the capacity of the research method to extract information from the
designer's activity and transform it into an explicit knowledge for design
purposes.

The Process, not the Product

The scope of design generation is limited to generating a very specific
architectural component that demonstrates weaving principles, instead of a whole
building configuration. Yet the process of developing this architectural
component encompasses several phases: interpreting design knowledge from the
traditional weaving, formulating the design knowledge into a Weaving Grammar,
teaching the Weaving Grammar back to local people, and testing the buildability
of the architectural component (generated from the Weaving Grammar) using the
local skills and local materials at the site.

Therefore, the conception of architecture in this thesis is not confined

solely to the architectural component, but rather extended through the cycle of
the component's design derivation from a traditional site, and its application back
to the traditional site. Thus, the evaluation should be made on the
comprehensibility of the Weaving Grammar in a traditional context, and the
Grammar's capacity to analyze and synthesize weaving designs.

1.6 Thesis Outline

This thesis is divided into four parts, beginning with this introduction (part one).

The second part, "EthnoComputation," outlines my computational

approach (EthnoComputation) in assessing and demonstrating the analytical use
 27

of shape grammar to observe traditional weaving activities. The procedures of the

EthnoComputation study are computationally detailed and findings from the
observation are presented visually.

The third part, "Computational Weaving," formulates two weaving

grammars, Basic Weaving Grammar and Abstract Weaving Grammar, based on
activities of the traditional weaver and theories from Shape Grammar and Finite
Element Method. Case studies (including weaving grammar workshops, building
workshops), as well as demonstration of designs generation from the Weaving
Grammars are outlined to demonstrate the process of reinventing weaving for an
architectural scale.

The fourth part, "Conclusion," summarizes and discusses the role of

computational design in interpreting and reinventing weaving in architectural
design and envisioning the future work of EthnoComputation and architectural
weaving.
28

a.

b.

C.

Figure 2. Pseudo-Woven
Architectural weaving from the left: [a] Aragon Pavilion for World Expo 2008 in Zaragoza
by Olano and Mendo, ' [b] Spain Pavilion for World Expo 2010 in Shanghai by Benedetta
Tagliabue," [c] Center Pompidou Metz in Paris by Shigeru Ban.' Close up images on the
bottom of each building show faqade components expressing the aesthetic of weaving, yet
structurally, the components are not interwoven.

" Images available from: huep//en.wikipedia.org/wiki/Zaragoza and M

IImages available from: http//commons.wikimedia.or/wiki/File:Spain Pavilinn of Expo 2010 2,ivg and
https://www.flickr.com/photos/13286981(&NG4/5360958173/
' From: http://www.archdaily.com/490141/centre-pmpidou-metz-shigeru-ban-architects/ and
http://www.checkonsite.com/centre-gomidou-metz/comment-page-1/?pid=870
 29

ab
Figure 3. Woven Bentwood and Sculptural Weaving
[a] Frank Gehry's Knoll Chair (Power Play) is woven with Bent wood technique using hard
white maple veneer strips' and [b] Frank Gehry's Fish Sculpture at the top of Vila
Olympica in Barcelona is woven with Bronze Strips and supported by steel frame structure
from behind.' The Fish is woven using a similar type of weave as the Knoll chair (i.e.,
plain weave).

" Fromht://areaneo.comblog/2013/05/1/frankgehry-lounge-chair-power-play-for-koll-iteratoal-
192/ and http://hivemodern.com/pages/product537/knoll-frank-ehry-power-npla-lounge-chair
' From: http://www.europaenfotos.com/barcelona/eng pho bcn 175.html and
http://www.shutterstock.com/ic-30686695/stock-photo-barcelona-april-frank-gehy-s-peix-d-or-whale-
sculnture-on-the-beach-of-barcelona-aoril.html
30
 31

2 ETHNOCOMPUTATION

2.1 Outline

This chapter describes my field study, in which I used computational means to

interpret the craft of weaving in a traditional village in Toraja-Indonesia. In the
field study, the craft of weaving is analyzed beginning from the acquisition of
materials from nature (bamboo) to the weaving process itself.

Challenges of interpreting crafting activity in this context lie in

representing, explicitly, the tacit knowledge in traditional practice.3 The
indigenous craftspeople in this field study craft with skills passed down from
generation to generation. They work based on sensory input from engaging with
their tools and materials (e.g., sight, touch, sound, smell and taste). They learn
tacitly, by observing other craftspeople's activity and other crafts objects, rather
than by attending formal institutions with explicit instructions and design
theories. To represent this tacit knowledge, I observe a craftsperson's activity and
analyze his observable interactions with the craft object, i.e., eyesight direction,
hand contact and body movement. From this interaction, I interpret the
craftsperson's sensorial experience explicitly through computational
representation. To represent how he learned from other works, I observe the
variety of crafts objects that are available on the site, and interpret the crafting
process into computational rules.

The chapter is divided into five sections. The first section situates my
computational approaches through reflections on other disciplines' work in
traditional communities; in particular, I reflect on the ways in which scholars
analyze and represent traditional craft in their case studies. Based on this
reflection, I propose my in-situresearch methods, EthnoComputadon,which are

" Michael Polanyi, Knowing andBeing: Essays by Michael Polanyi, First Edition edition (Chicago: University
Of Chicago Press, 1969).
32

built upon Shape Grammar principles. A discussion of how Shape Grammar

principles describe traditional crafting processes on the site is provided.

The second, third and fourth sections describe my EthnoComputation

method in interpreting weaving activities, from three perspectives: [a] from a
general observation of the site, [b] from an observation of weaving's material
derivation from nature into crafts components and [b] from focused observations
on the weaver-weave interaction.

The second section demonstrates how EthnoComputation works on visual

observation by showing the application and benefits of seeing-in-therein
conjunction with beingthere.* I illustrate my procedures for representing a
traditional village as my site, and for delineating the scope of my research through
visual computation. From the visual computation, I generate several reconfigured
sites, each of which can be experienced and studied from different perspectives.
Two of the reconfigured sites highlight relationships between craftsperson, craft
objects and the nature. The relationships highlighted in these two frameworks
serve as the research framework for the third and fourth sections.

The third section interprets how nature is translated into material for the
craft of weaving. Drawing upon Aristotle's matter and form conception
(hylomorphism),4 1 42 through an interpretative computation, I associate the
abundance of bamboo as the main weaving material and the diversity of bamboo
crafts in the villages. I develop rules and schema to exemplify how craft material
is made from two directions: by reverse-engineering the material manufacturing
process back to nature and by generatively deriving the natural material into a
diverse type of craft component. This interpretation highlights a pivotal
assumption about how the weaver perceives a shape, an assumption I use as I
observe the weaving process, as detailed in the fourth section.

' Clifford Geertz, Works andLives: The AnthropologistasAuthor, 1 edition (Stanford, Calif.: Stanford
University Press, 1989).
41
Christopher Shields, "Aristotle," in The Stanford Encyclopedia of Philosophy, ed. Edward N. Zalta, Spring 2014,
2014, http://plato.stanford.edu/archives/spr20l4/entries/aristotle/.
' Frank A. Lewis and Robert Bolton, Form, Matter, andMixturein Aristotle (Oxford, UKE; Malden, MA:
Wiley-Blackwell, 1997).
 33

The fourth section describes the interaction between the weavers and
weaves during the weaving process. Based on my observations on the site, I
present the weaver's sensorial interaction with the weave through an iterative
computation. The sensorial interaction is represented within two timeframes:
slow motion and fast motion. In the slow-motion analysis, the analysis focuses on
the weaver's visual and haptic interactions with the weave, to interpret shapes that
the weaver embedded perceptually in the weaving process. In the fast-motion
analysis, I analyze the weaver's body orientation toward the weave, as he moves
from one position to another, to understand how the weaver views the shapes as
he adds a thread on the weave in different positions.

The discussion in the last section of this chapter reflects on how

traditional weaving activities, as a tacit knowledge, can be interpreted by an
observer using the EthnoComputation approach.

2.2 Introduction

In situ research of traditional crafts has focused on different subjects, including

ornamentation study, such as Franz Boas's documentation of the art of North
America's North Pacific Coast,43 socio-symbolic study, as seen in Roxana
Waterson's documentation of the cosmological aspects of Southeast Asia's
traditional houses,44 and socio-educational study, exemplified by Patricia
Greenfield's study of the social dimension of back-strap looming activities in
Zinacantin, Mexico. 45

Examined by researchers of various backgrounds and built upon a

multidimensional cultural foundation, traditional-craft studies inevitably lead to
several disciplinary convergences; each discipline focuses on a specific aspect of
the craft. Two disciplines that intersect with computation and visual material are

4 Franz Boas, PRimidve Art, 1st Dover Edition (Dover Publications, 1955).
4 Roxana Waterson, The Living House:An Anthropology ofArchitecture in South-EastAsia (Tuttle Publishing
2010).
' Greenfield, Weaving GeneradonsTogether.
34

47
ethnocomputing, which study the application of computing in cultural settings,
and visualethnography,which study the use of visual media and material, as well
as the incorporation of a visual lens into ethnography.48

The explicit representation and visual awareness proposed in these fields

coincide with my objective in studying weaving in the traditional context.
Nevertheless, considering that visual ethnography originated from social sciences
and that ethnocomputing was derived from computer science, it is necessary to

understand how the views of craft in these disciplines intersect with the view of
craft in this thesis. In this thesis, I view craft as a design activity. How are the
perspectives from which scholars in visual ethnography and ethnocomputing
describe the crafting process, and present it to their audiences, useful to design?
How might a design study diverge from these perspectives?

The Visual in Visual Ethnography

Visual anthropology is generally known as an anthropological practice using a
49
visual medium, as well as a study of visual phenomena in culture and society.
The notion of what visual means in a certain culture and how to capture visual
culture through a visual lens has held the spotlight in academic debates in visual
anthropology. As a method in visual anthropology, visual ethnography has been
re-defined, promoted and criticized among anthropologists. Anthropologist and
ethnographic filmmaker David MacDougall is a strong critic and also a leading
advocate of visual anthropology. MacDougall builds upon his concern for the
textual narration of visual material, exclusive of ethnographic film studies, and

the myopic view of the use of visual technology for anthropological practice rather
than the visual form of anthropology to re-conceptualize visual anthropology in
his responses to the earlier Principlesof VisualAnthropologyproposedby Paul
Hocking.-0 In The CorporealImage,MacDougall posits three principles:

' Tedre Matti, "Ethnocomputing: A Multicultural view on Computer Science" (M.Sc Thesis, University of
Joensuu, 2002).
* Karen O'Reilly, Key Conceptsin Ethnography(SAGE Publications Ltd, 2014).
* Encyclopxdia BritannicaOnline, s. v. "visual anthropology,"
PuHt://www.britannica.com/EBchecked/tolic/e1072842/visual-anthrog
.

' Paul Hockings, Principlesof Visual Anthropology, 3 edition (Berlin: De Gruyter, 2003).
 35

(1)To udlize the disdncdve expressivestructureofthe visualmediarather

than those derivedfrom expositoryprose;(2) To develop forms of
anthropologicalknowledge that do not dependupon the principlesof the
sciendficmethod for their validity;(3) To explore areasofsocial
experiencefor which visualmedia have demonstratedexpressiveaffinity
-in pardcular(a) topographic, (b) the temporal, (c) the corporeal, and (d)
51
thepersonal.

MacDougall's principles imply an invitation for visual ethnographers to

build more confidence in the validity of anthropological knowledge and
ethnographic method in studying the visual media and visual culture. His
principles of visual anthropology were later echoed by social anthropologist Sarah
Pink, who defined ethnography as a "methodology; as an approach to
experiencing, interpreting and representing culture and society that informs and
is informed by sets of different disciplinary agendas and theoretical principles".5

Pink further underlines the role of the ethnographer's personalexperience

in represendngknowledgethatdoes not claim to produce an objective or truthful
account of reality, but rather an 'individualexperience ofreality.0 In other words,
the same reality will be experienced differently by different individuals with
different reasoning and perceptions. These differences would lead to different
findings (Figure 6). In practice, the content of visual ethnographic studies ranges
widely. In terms of visual material, the studies range in coverage from the visual
material ownedbya subject, to the subject activity as the visual material. In terms
of visual media, the studies range in coverage from the technology used to
capture the visual material (Figure 4), to the media used to publish the visual
material (Figure 5).

51David MacDougall, The CorporealImage: Film, Ethnography, and the Senses (Princeton, N.J: Princeton University
Press, 2005).
" Sarah Pink Doing Visual Ethnography,Second Edition edition (SAGE Publications Ltd, 2012).
3 Ibid.
36

Figure 4. Street photographer in Dehra Dun, India, 1989

"

People and devices in this picture are a few of many subjects from MacDougall's visual
anthropology study, which covers multi-dimensional aspect of photographic practice in
India, ranging from the social life of the photographer to the context and history of the
practice to the mechanism of the camera.

54 From: MacDougall, The Corporeal Image. p.157

 37

1~ I'.iid knr J h hi~

~~ Ib, -1 ~. C
( inmrm.nt d kIla vd' v Illa M
l zdji
.

pt t'ard by I. At{muu, I rncuh ( mgt, a. I 0"

Figure 5. The Disappearing European man, 1905

"

Two postcards by French photographer, Jean Audema, at the French Colonial in

Mayumbe, Congo, serve as MacDougall visual material for his study on the colonial
photographer. Visual anthropological study from this image can go far beyond the
intriguing subject in the picture (i.e. the disappearance of the bearded European in the one
picture and ways of posing the natives); for instance, the postcard itself as a visual medium
along with its embedded properties: caption, written text, style, public tastes, etc.

* From: MacDougall, The CorporealImage. p.177

38

Now
M

Figure 6. Women in Bullfighting s

Pink illustrates three possible interpretations of the sequence of these images from three
groups of people: [1] fans who support women as matador, [2] fans who are skeptical about
the idea of a'women matador' and [3] people who are against bullfighting. The top image
may entertain the first group, while the other images may justify skepticism in the other
group, i.e., that women shouldn't perform in bullfighting (group 2) or that the bullfighting
itself should be banned (group 3).

" From: Sarah Pink, Doing VisualEthnography, Third Edition edition (SAGE Publications Ltd, 2013).p.77
 39

The Computation in Ethnocomputing

In a more formal discipline, a group of computer scientists approach the
traditional site through digital representation. Drawing from his concern on the
dominant'western' culture in computer science education in developing
countries, Tedre Matti, a computer scientist, offers the term ethnocomputing, as
"a field of research that studies the phenomena and applications of computing in
different cultural settings". Here the term "computing" refers to the use of
computer applications in an educational context. In his thesis, Matti proposes
ethnocomputing as a "tool for developing a multicultural approach in Computer
Science education, recognizing the influence of societal and cultural backgrounds
on learning Computer Science."*

Matti's illustration of indigenous crafts as exemplifying ethnocomputing

practice, such as Marcia Ascher's study of the Inca Quipu 6'1 and Ron Eglash's study
of the African fractal6 2, is mostly metaphorical. The computing aspect of the craft
is proposed by mapping properties of computer language onto features found in
the artifact (e.g. color coding and number of knots in Quipu). Matti looks at Quipu
as "a structure that represents organized information, and can be considered as a
data structure".6 Matti has continued to work in Information and Communication
Technology (ICT). His projects are conducted primarily in rural settings in
Ethiopia, Colombia, Tanzania, South Africa, and Sri Lanka, where he uses
computers to analyze how students interact with teaching materials. The original
concept generated in Matti's thesis, computing= computer, continues as the
model in ethnocomputing research (Figure 7). Recent ethnocomputing studies
seem to focus mainly on understanding how ICT affects and interacts with society
and culture."

60
Matti, "Ethnocomputing: A Multicultural View on Computer Science."
Marcia Ascher, Mathematicsofthe Incas: Code ofthe Quipu (Dover Publications, 1997).
* Ron Eglash, African Fractals:Modern Computing and IndigenousDesign, 3rd edition (New Brunswick N.J:
Rutgers University Press, 1999).
" Matti Tedre et al., "Ethnocomputing: ICT in Cultural and Social Context," Commun. ACM49, no. 1
(January 2006): 126-30, doi: 10. 1145/1107458.1107466.
" Matti Tedre et al., "Ethnocomputing: ICT in Cultural and Social Context," Commun. ACM49, no. 1
(January 2006): 126-30, doi: 10. 1145/1107458.1107466.
40

Ron Eglash, who supports Matti's definition of ethnocomputing, is

probably closer to the design discipline. Eglash's ethnocomputing studies aim to
help students learn math using indigenous design and digital simulations." This
scope is apparent in his interactive digital representation of traditional design,
such as the African Fractal software, to produce a computer-generated design
resembling traditional designs. On his website, "Culturally Situated Design Tools:
Teaching Math and Computing through Culture," Eglash provides Internet access
to various simulators of the traditional design. Under the title, there is an
encouraging description about the software:

Many culturaldesigns arebasedon math andcompudngprinciples. This

software will help students learn theseprinciplesas they simulate the
oiginalardfacts, anddevelop theirown creations.6 7

However, the simulator does not show the design process or how the
craftsperson interacts with the design. The software lets the user generate the
design by typing an input for the design parameter -- for example, the shape
position (point coordinate), symmetrical properties (rotation angle), and recursive
function (number of steps). Thus, what the user will learn from the software, as
promised, is the math and computer principles. To be precise: users learn the
Principle of Fractal, which may or may not relate to the original design principles.
After I tried the software, the website description looked different to me:

Many-euite designs-rehosed can be explained with math-and

eemp fdtg fractalprinciples. Thissoftware will help studentslearnthese
principlesas they simulate the eriginlfractalshape thatis similarto
existng designs, anddevelop theirown ereadonsfractaldesign.6

Eglash himself did not claim that his fractal description is the main
concept underlying the African built environment. In African Fractals,he

"Ron Eglash et al., "Culturally Situated Design Tools: Ethnocomputing from Field Site to Classroom,"
Ameican Anthropologist108,no. 2 (June 1, 2006): 347-62, doi:10.1525/aa.2006.108.2.347.
"' Ron Eglash, "Culturally-Situated Design Tools," CulturallySituated Design Tools: Teaching Math andComputing
through Culture,accessed May 16, 2014, http://csdt.rpi.edu/.
" The software is still fun to use (if not addictive). However, after several design iterations, my initial
appreciation on the unique traditional design has shifted into the universality of the symmetry. From the
software, I appreciate the fractal more than the traditional design. What left from the cultural notion is
simply the ethnic-label indicating the place where the artifact found.
 41

acknowledges, "African fractal geometry is not a singular body of knowledge, but

rather a pattern of resemblance that can be seen when we describe a wide variety
of African mathematical ideas and practices, and it is not the only pattern
possible". 69 For Eglash, the Fractal representation of African artifacts is a thin
70 The question of
description,a study of surface particularities of an artifact.
whether the artifact may or may not symbolize deeper meaning than the fractal
appearance is out of the scope of his thin description.

Figure 7. Introduction to Java Programming Language in Tanzania 7 1

Local traditional houses used as a graphic metaphor for the online-programming course.
The metaphor used to make the course material become familiar is easily grasped by the
learners.

Figure 8. African Fractal simulation"

Using a set of fractal algorithms, the software (left) generates shapes that are similar to the
top view of the Ba-ila village, Zambia (right). The iteration button (e.g., ITER1, ITER2)
shows different steps of fractal iteration The same software simulates other traditional
artifacts in Africa that can be explained using Fractal principles."

6 Eglash, African Fractals.P.182

' Ibid. p.181, in response to Clifford Geertz "Thick Description", that criticize structuralism approach in
reducing the cultural symbols into a mechanistic syntax.
" From: Mikko Vesisenaho Erkki Sutinen, "Ethnocomputing in Tanzania: Design and Analysis of a
Contextualized ICT Course.," Research and Pracdcein Technology EnhancedLearning1(2006): 239-67,
doi:10.1142/S1793206806000238.
' From: Eglash, "Culturally-Situated Design Tools." htp://csdt.rp.edu/african/African Fractals/culture4.html
" Ibid. See ht://csdt.rpi.edu/ for more example of Eglash's Fractal program.
42

The EthnoComputation Approach

The above reflection provides a new path for me to situate my approach to
observing traditional weaving on a field study site. My aim in this field study
corresponds to a visual ethnography project in terms of drawing a distinctive
picture of a mix of knowledge, through personal experience on the site. My aim
also parallels the ethnocomputing practice of representing the local knowledge
explicitly, based on observations of the artifact, without verifying the cultural
symbol or meanings underlying the artifact. However, this field study does not
aim to promote digital technology or apply a particular logic to a traditional
object, as demonstrated in ethnocomputing. Nor does it advocate for a particular
culture, nor debate the use of the visual media as visual anthropology does. This
study aims to understand how craft is perceived during the making process, by
interpreting human-craft interaction explicitly in the traditional context.

In studies of traditional craft, the crafts objects themselves are often

viewed as artifacts; the culture in which they are produced is viewed as the
context for their creation. In this study, I view culture as the way humans
represent how they understand their environment. Different understandings of
the environment -- seeing it in different ways, with different goals -- can lead to
different representations (i.e., the creation of different kinds of objects). From
this perspective, the craft object itself is not the artifact; the artifact is in the
interaction between a person and the object.

In The Sciences ofthe ArtiBcial, Herbert Simon posits the artifact as an

interface between an inner environment and outer environment -- that is, "if the
inner environment is appropriate to the outer environment, or vice versa, the
artifact will serve its intended purpose".' For example, as an artifact, a bird's
wing serves as an interface. Observers of the bird's wing interpret how the inner
and outer environment interacts in the bird's flying mechanism. In aerospace
engineering, the interpretation has been focused on the shape of the bird's wing,
and has led to different types of airfoil for artificial flying. Different observers may
have different interpretations, or focus on a different aspect of the inner-outer

74
Herbert A Simon, The Sciences of the Artificial -3rd Edition, 3rd ed. (The MIT Press, 1996).
 43

environment of a bird's wing; such different interpretations could lead to different

innovations. A different focus on the bird's wing -- a focus on the flapping
mechanism, rather than on the shape -- helped the Festo Company develop a
bionic bird that can fly by flapping and twisting its wing (instead of using turbine)
75
in their SmartBirdproject.

Drawing from Simon's framework, I assign the term "artifact" in this study
to the interface between objects and the craftsperson, i.e. the craftsperson's tacit
knowledge and perceptual system as the inner environment, and the craft's
material and the craftsperson's habitat as the outer environment. This interface is
built upon the assemblage of pertinent disciplines (e.g., computational method,
perceptual system) and the evolving knowledge based on the researcher's
experience at the site; to reason clearly about the tacit knowledge on the site, I
rely upon the explicit representation of my own experience.

In this study, I tweaked the term ethnocompudng, described above, to

EthnoComputadonto stress the significance of 'ethnos'in positioning humans at
the center of the subject and'computation' to emphasize the need for explicit and
generative analysis. The scope of EthnoComputation in this study is
conceptualized as follows:

1. To contextualize a computational framework supported by pertinent

disciplines to observe artifacts and human activities on a field study.
2. To interpret physical and perceptual shapes through the interactions
between humans and their artifacts.
3. To develop forms of interpretative computational knowledge unique to a
particular site and individuals, at a particular point in time.

Thus, the hypothetical reconstruction of craft study in the

EthnoComputation method is not archeological or historical; the goal is not to
authenticate the craft object. The engagement with the participant on the site is
not anthropological or psychological; the goal is not to verify the craftsperson's
thinking process or his/her world-view. The way the subject is observed may be

7
s SmartBird," accessed May 16,2014, http://www.festo.com/cms/en-corp/11369.htm.
44

similar to the methods used in these disciplines. However, the research focus is
different. The research focuses on the interaction between human and craft
because in EthnoComputation, the explicit representation from this interaction is
the artifact.

Situating the Computational Framework

The computational method for observing traditional contexts must be intuitive
(i.e., easy to use in a traditional site) and adaptive so as to anticipate possible
emergent interpretations of the site. The method must accommodate both
physical and perceptual experience to enable the observers to observe how
humans perceive the object. Furthermore, the computational method should not
intrude upon the research subject.

To meet these requirements, the study uses Shape Grammar, a

computational method used to analyze and synthesize design with a set of shape
rules, as the foundation of the research method.76 Shape Grammar accommodates
the above requirements with its intuitive and explicit reasoning process. 7 The
simplicity of Shape Grammar rules and the fact that it uses shapes to analyze
objects makes it more practical for both interpretation of the subject on the site
and knowledge representation for the reader. In addition, Shape Grammar
computation does not depend on a particular technology or media, as shape rules
can be computed in both analog and digital media.78 All of these features adapt
well to the conditions in a traditional context. To obtain the optimal benefit from
Shape Grammar, I contextualize some principles in Shape Grammar for in situ
application.

First, I present the mathematical abstraction of the Algebra of Shape in

Shape Grammar as a perceptual-physical abstraction (Figure 9). In the algebra of
shape, the shape is represented as a point, line, plane or solid in n-dimensional

'See the introduction chapter for Shape Grammar definition (page 22)
" Terry Knight and George Stiny, "Classical and Non-Classical Computation," Arq: ArchitecturalResearch
Quarterly5, no. 04 (2001): 355-72, doi:10.1017/S1359135502001410.
' See the first shape grammar computation with non-digital representation in Stiny and Gips (1971) and
some examples of Shape Grammar digital interpreter by Andrew Li from http://andrew.li and Miranda
McGill from httn-/klesignmasala.com/nortfofio/ui-shaner2d.html.
 45

space (i.e., OD, 1D, 2D, and 3D)." However, the shape and space relationship on a
site is represented, not mathematically, but perceptually, relative to the eyes of
the beholder. Although object and space in an in situ observation are always
three-dimensional, that does not necessarily mean the shapes on the site cannot
be perceived as two- or one-dimensional representations. Therefore, in this
observation, a shape is represented by how its characteristics are perceived (e.g.,
atomic, linear, planar and volumetric), where the space is represented as a
medium that the shape inhabits.

0: atomic U00 U01 U02 U 03

1: lnear U 11 U 12 U13
2: planar U22 U 23
3: volumetrIc U 33

Figure 9. Algebra of Perceptual Shape

Based on Stiny's Algebra of Shape,"' a number in a column represents the dimension of the
perceivedspace which a shape inhabits, while number in a row represents the dimension
of the perceptualshapethat exists in the perceived space. For example, U12 indicates a
linear shape in the planar space. Uo1 is atomic shape in linear space, thus can only do linear
motions in an axis (Atomic shape is a type of singular shape that is perceived as not
decomposable into smaller parts, regardless of its dimension).

A set of objects -- for example, a pencil resting on a table -- can serve either
as a shape or a space based on how an individual perceives the relationship
between the pencil and the table. The pencil on the table can be perceived as a
one-dimensional line on a two-dimensional plane (U1 2), yet the eraser surface on
the end of the pencil can be regarded as a two-dimensional circle inhabiting a
three-dimensional cylinder (U 23). Later, if an individual uses the pencil to write or
draw, he/she may perceive the graphite tip of the pencil as an end-point of a line-
stroke on the paper (U02), or, as a peak point of the graphite cones of the pencil
(U03). The former may be more apparent visually (as perceived by the eyes), and

7 George Stiny, Shape: Talking about Seeing and Doing (The MIT Press, 2008).
8 Ibid.
46

the latter may be more apparent tactilely (as perceived by a sense of touch via
sensory nerves in the fingertips, fingers, hand). Thus, our perceptual mechanisms
(visual and tactile) can readily shift the shape-space relationship.

Second, I make an adjustment to situate Shape Grammar rules in order to

observe the site. In a rule notation, the shapes are placed on the left-hand side of
the arrow (LH-shape) and on the right-hand side of the arrow (RH-shape), where
the RH-shape substitutes for the LH-shape (Figure 10). Although the RH-shape
sometimes appears as the transformed LH-shape, this does not mean that the RH-
shape is always related to the LH-shape. The RH-shape can be anything,
including empty space or an empty element. Therefore, any object on the site can
be regarded as the LH-shape as well as a RH-shape.

LH-shape RH-shape
STF P (W0TE1 9
RULEe-

RULE ...

E2
2

3il

Initial Shape

Figure 10. Rules and Computation in Shape Grammar "I

With three rules on the upper left and the initial shape on the lower left, the computation
process on the right shows how rule iterations generate new shapes whenever there is a
shape that matches with the left-side shape of rules 1 and 2. Rule 3 will erase the label
(teardrop-shape) and stop the computation process because when the label disappears;
there will be no more shapes that match with the rules.

" From: George Stiny and James Gips, "Shape Grammars and the Generative Specification of Painting and
Sculpture," in InformationProcessing71 (presented at the International Federation for Information
Processing (IFIP), Ljubljana, Yugoslavia, 1971).
 47

A B

-- OR

1 2 3a 3b

82
Figure 11. Rules A -4 B and Computation

In a rule A -> B, the arrow does not refer to a particular transformation

method for changing the initial shape A into the new shape B. The process of
"making a shape" in Shape Grammar computation is quite distinctive, that is: to

see A, erase A and show B.1 For example, consider the shape rule that turns shape
3

A into shape B, and the three iterative steps in Figure 11. Between these steps,
more steps are required to make the shape. In Shape, Stiny states the procedure to
apply the rules:

Finda transformationt thatmakes the shapeA partof C. Thispicks out

somepartof C thatlooks like A

Subtractsthe transformadonofA from C, and then addthe same

transformationof the shapeB. This replacesthepartof C that'slike A with
anotherpartthatlooks like B. 84

Based on this instruction, the three iterative steps can be extended into six
steps in Figure 12, to ensure that the rule is properly applied.

[:] =:> [I] =g II OR

Initial shape match A erase A apply B match A erase A apply B apply 8
1 2 3 4 5 6 7A 78

8
Figure 12. Extended Rules Iteration s

The single-lined arrow (->) in the rule is, therefore, open for

interpretation as it is up to the rule user to employ the necessary transformation

" Find more example of this shape-rule in Stiny, Shape.

3 Stiny, Shape. p:232
8 Ibid.
' Find more example about seeing and doing in Terry Knight and George Stiny, "Making Grammar: From
Computing with Shapes to Computing with Things," in ComputationalMakfng(presentedat The 6th
International Conference on Design Computing and Cognition - DCC'14, London, 2014).
48

to subtract, addandreplacein Stiny's instruction. For example, adding a shape in

different orientation will result in different designs (see two possible orientations
at step 7A and 7B in Figure 12).

Double arrows (=>) in the rule iteration signify the process of applying a
rule (r)fora shape () in an iterative computation r(x). Thus, an iterative
computation r(x) implies that rule(r) and shape (A) exist. By reversing this
procedure, in a condition where craft objects exist on the site, we can apply an
inverse computation by assuming the existing crafts as the iterative computation
to retrieve back the rule (r) and the shape (x).

IF (r) AND (x) exist THEN IF r(x) exists THEN

compute r(x) return (x) AND return (r)

The return function to retrieve the rule (r) and the shape (x) is
interpretative because many rules and shapes can produce the same result(s). Just
as embedding the shapes from the shape rule into shape computation can be
ambiguous, likewise, embedding the shapes from shape computation into shape
6
rules can also be indefinite.

In a traditional culture, most of which consists of tacitly passed-down

traditions, access to explicit representation of the rule is not always direct. For
instance, suppose a craftsperson uses the rule in Figure 11 and uses shape B as the
initial shape to generate a set of craft objects (Figure 13-1). Later on, if an observer
happens to have access to these craft objects but not to the craftsperson's rule,
there is no guarantee that the observer will retrieve the rule in Figure 11 (Figure
13-2). Each of the possible embedded sub-shapes in the craft objects could be the
result of a different rule interpretation, because each embedded shape can be
used to deduce rules. For instance, based on different observations of the
embedded shapes in Figure 13-2, different rules (i.e. R2, R3, and R4) can be
defined to interpret how a craftsperson generates the craft object in Figure 13-3
and 13-4.

MTerry Knight, "Computing with Ambiguity," Environment andPlanningB: PlanningandDesign 30, no. 2
(2003): 165-80, doi:10.1068/b12915.
 49

Given the many possible defining rules from the available object-craft on

the site, it would be irrelevant to claim the original shape rule, even if a (set of)
87
rule(s) could generate an object that resembles the original craft.

1. Local craftsperson generated new crafts

2. The resulting crafts perceived by the observer

1 -+

R2 R3 R4

3. New rules interpreted by the observer

L R2 LR3 L11
R4

LP =L Iy
4. Crafts regenerated with interpreted rules

Figure 13. Inverse-computation from existing shapes into interpretive rules

7Other constraints, such as crafting skills, could greatly vary the craft objects. If a physical constraint is
increase.
included in the computation process then the possibility to retrieve the original rules will vastly
initial shape, the
Assuming craftsperson A and craftsperson B are working with the same rule and the same
in the
resulting designs can look somewhat alike but can vary based the physiological differences
in their
individuals. For example, craftspeople A and B may have different levels of tactile perception
that could lead to different designs based on the same rule.
fingertips and/or different visual perceptions
50

2.3 Seeing-in-There

Figure 14. Kete Kesu, Toraja, Indonesia

The region of Toraja, Indonesia is well known for its ornamental crafts,
woodcarvings, unique traditional houses (known as Tongkonan), funeral rituals
and other traditional ceremonies.8s9 90 The site location, Kete Kesu, is a village in
Toraja of about 350 residents. Culture in the village embraces a wide range of craft

diversity. There is an abundance of interwoven artifacts and weaving materials

(bamboo) around the village. Interwoven bamboo is commonly used to craft
household items such as chicken cages, carriers, sacks, room and building
partitions, and decorative hangings.

Site as a Schema

Figure 15. The initial setting for EthnoComputation

8 Kathleen M. Adams, Art as Politics:Re-CraftngIdendties, Tourism, andPowerinTana Toraja, Indonesia

(Honolulu: University of Hawaii Press, 2006).
" Jowa Imre Kis-Jovak, Banua Toraja:ChangingPatternsin ArchitectureandSymbolism Among the Sa'dan
Toraja, Sulawesi, Indonesia(Amsterdam: Royal Tropical Institute, 1988).
" Roxana Waterson, PathsandRivers:Sa'dan TorajaSocietyin Transformation(Leiden: KITLV Press, 2009).
 51

Just as there are different ways to see a shape, there are different ways to perceive
a site. I represent the setting in Figure 15 in a non-judgmental view where all
objects on the site are recognized as shapes that hold equal Weight. 91 For instance,
objects in the foreground, mid-ground and background have the same level of
hierarchy (the person is no more important than the other objects or vice-versa).
The non-judgmental view is analogous to an "image processing framework",
proposed by David Marr in computer vision. In Marr's framework, retinal image,
a representation of object in the real world that is projected into a retina, has no
meaning attached. A retinal image is processed through a bottom-up approach
from low-level vision (differentiating edges of one object with another), to mid-
level vision (recognizing shape attributes such as color, shading, texture, motion
flow), and then to shape recognition in a high-level vision, where shape attributes
from mid-level vision is being recognized with meaning.

Figure 16. The initial setting (left) is weighted with a term 'weave' to highlight
woven shapes on the site (right).

In this setting, I only recognize objects in their basic terms (e.g., a person,
threads, a panel), but their value and their relationship to another are neutralized.
This non-judgmental view serves as my initial setting to compute my research
focus on the site. In this setting, I begin to add Weight to the shapes that match the
description of 'weave' as a cyclical over-and-under shape relationship, such that:
IF weave is found THEN adds Weight to the weave. The Weight will contrast
weave apart from other objects on the site, and gradually, only parts considered as
woven are perceived in this scene, e.g., woven-partition, woven basket-base,
woven screen, etc. (see Figure 16).

" G. Stiny, "Weights," EnvironmentandPlanningB: Planningand Design 19, no. 4 (1992): 413-30,
doi:10.1068/b190413.
92 (David Marr, Shimon Ullman, and Tomaso A. Poggio, Vision: A Computational
Investigationinto the Human
RepresentationandProcessingof Visual Information (MIT Press, 2010).
52

tna~I wew ~ce~nbamboo tree

bjatboo thread weaer

arnae tal-pauttlun

Figure 17. Weighted objects on the Site

Weight highlights shape related to weaving in its particular form, e.g. ornamental partition
and screen, triaxial-weave, bamboo thread and bamboo tree as a weaving material, and the
weaver.

Reconfigured Sites
Objects in the setting are recognized separately in the weighted site as their
relationship is not yet defined (Figure 17). This setting serves as a schema that can
generate emergent research frameworks for perceiving the site. For example, in

the site schema, an observer might recognize ornamental shapes on the

interwoven partition in the background and further investigate differences

between weaving for ornamental and weaving for practical purposes. The
emergent relationship between the weaver and the ornament in Figure 18 shows

one of many possible frameworks that can be generated from the site schema.

For this study, I limit my scope to weaving in two frameworks (Figure 19).

The first framework highlights the transformation of weaving material from

nature into craft components, given the ubiquitous bamboo trees in the

background and the pervasive interwoven bamboo artifacts in the setting. The
second framework focuses on the weaver's interaction with the weave during the
production process, given the weaver's constant visual and physical focus on the
material.
 53

weaver: weave triaxial-weave: biaxial-weave nature: craft

weaver: nature material: design weaver: ornament

Figure 18. Object associations

Objects in the setting are paired to generate different frameworks in perceiving the site.

Nature: Craft Weaver : Weave

Figure 19. Reconfigured Sites from Object Association: Nature-Craft and Human-
Craft
Two frameworks from object association highlight my research focus on the site: [1] How
nature is transformed into material, and [21 how material is transformed into a design by
the weaver.
54

2.4 Mediated Bamboo

This section illustrates my observations on bamboo as the weaving material by

focusing on the way that bamboo's shape is transformed from matter (nature),
into form (craft).

In view of the dominance of form in the form and matter relationship, Tim
Ingold, anthropologist, proposes an alternative model to Aristotle's
Hylomorphism. Ingold criticizes means of reproduction that occur between the
conception of a material as a "formless lump of matter" and the material as "form-
bestowing agency of human beings".93 Building upon philosophers Gilles
Deleuze and F6lix Guattari's model, Matter-Flow, Ingold renders his Environment
Without Object (EWO) model, where matter flows through the physical
environment. Thus, even in its'brute' form and in the absence of human agency,
material is always in a state of flux, changing as it interacts with its environment.95
%97

Ingold's EWO model supports my aim in investigating the flux of shapes in

bamboo's transformation. Yet, in spite of Ingold's criticisms of Hylomorphism,
the theory of four causalities (a theory embedded in Aristotle's Hylomorphism
model) provides checkpoints that could help to navigate the flows of the shape.
The four causalities in this theory are the materialcause(that from which
something is generated and out of which it is made); formalcause (the structure in
which the matter comes to be something determinate); efficientcause (the agent
responsible for a quantity of a particular matter that is formed) and finalcause
(the purpose or goal of the object created from form + matter).98 In my case, the
efficient-cause is represented by the weaver; the material-cause is represented by

" Tim Ingold, Bringing Things Back to Life: CreativeEntanglements in a World of Materials,Working Paper
(Realities / Morgan Centre, University of Manchester, July 20, 2010),
http://www.manchester.ac.uk/realities/publications/workingpapers/.
" Hylomorphism (matter: huld and form: eidos or morphe) holds that no such object contains two distinct
metaphysical elements: one formal and one material.(Shields, "Aristotle.")
"Based on James Elkins, What PaintingIs, 1 edition (Routledge, 2004).
"Tim Ingold, Malng:Anthropology, Archaeology, Art andArchitecture (Routledge, 2013).
"Tim Ingold, "Of Blocks and Knots: Architecture as weaving," accessed May 17, 2014,
http://www.architectural-review.com/comment-and-opinion/of-blocks-and-knots-architecture-as-
weaving/8653693.article.
" Shields, "Aristotle."
 55

the bamboo; formal-cause is represented by the basket; and final-cause is

represented as the carrier."

Responding to Ingold's EWO and Aristotle's theory of causation, in

observing bamboo as a local weaving material, my observation is situated at the
flows in-between these causalities. Therefore, an initial mapping of these
causalities -- for instance, in basket weave, "a basket is made by linear thread and
woven into a cylindrical shape by a weaver to be used as a container to carry
things," -- is not perceived as the final explanation. Rather, it is an interface to
understand how the four causes affect one another.

Figure 20. Abundant Bamboo trees (top) and various bamboo crafts in Kete Kesu
(bottom)

" Anna Marmodoro, "Aristotle's Hylomorphism without Reconditioning," accessed May 17, 2014,
https://www.academia.edu/4195031/Aristotles-hylomorphism..withouLreconditioning.
56

I investigate the dynamic of these causal interactions in different

materialization states. I begin from a state where the bamboo tree is considered
matter and the pole of the bamboo is form, and then moving to the state where the
pole is considered matter and the thread, form, and later on in the process, to a
state where the thread is matter and the completed basket, a form.

In each of these recursive states, I focus at the moment when the

craftsperson perceptually gives form to the matter. In particular, the moment
when s/he visualizes the desired shape, right before his/her physical action is
applied to the matter. In cutting down the tree, for example, when a logger swings
his axe to cut the tree, he may visualize a shape on the tree as a target area for the
axe in order to properly hit the tree.

Generative Bamboo l0
The wide range of bamboo crafts in the village can be represented with a
parametric schema. In this schema, the part of the bamboo that is cut off
(negative shape) and the part that is uncut (positive shape) are equal. For instance,
a rectangular shape appears in the bamboo crafts, both as a positive shape (i.e., a
rectangle as a two-dimensional projection of the cylindrical bamboo) and as a
negative shape (i.e., rectangular holes in bamboo).

One of the pivotal moments in making these positive and negative shapes
occurs when the craftsperson's knife-edge comes in contact with the bamboo
surface to cut it into different shapes. This contact links the inner environment of
the craftsperson, and the outer environment of the bamboo. Therefore, the knife's
cutting line, mentally embedded on bamboo by the craftsperson in such a
moment, can be regarded as an interface with which to interpret the
craftsperson's knowledge and the bamboo-craft diversity (Figure 21).

Klaus Dunkelberg, IL 31, Bambus. IL 31, Bamboo. BambusAlsBaustoff Bauen Mit PflanziichenStiben.
(Kramer, Stuttgart, 2000). Provide numerous detail on bamboo utilization in Indonesia, including Toraja,
which has been used as a main reference in this study along with my direct observation on the site.
 57

Figure 21. Bamboo-Cutting Line

Weaver cutting-bamboo activity as the interface and the embedded mental-shape as the
artifact for ethnocomputation study. Contact area between the knife-edge and the bamboo
is represented with the cutting shape (red line), as an interface between the craftsperson's
inner-environment and his outer-environment.

Based on the weaver's cutting bamboo activity as the interface, I describe

my interpretation of the cutting process using two representations of bamboo: a
side-elevation and a cross-sectional view.

Side-ElevationRules
The cylindrical shape of the bamboo pole is represented through the bamboo's
side elevation with two cutting rules: longitudinal cut and lateral cut. The distance
between the end of the pole and the cutting line is parameterized with variables:
length (L) and width (w) (See rule Ri and R2 in Figure 22). The two cutting rules
can be applied recursively to generate more cutting lines, in order to subtract the
bamboo. After the cutting rule is applied, the desired shape can be signified with
Weight by applying rule R3. The three examples of rule R3 application indicate
different desired shapes from the same cutting lines.
58

R1K R3

- - w-

R2K

R3
initial
shape

R2 R2 R1

K R1 R3

R3 n

Figure 22. Cutting rules on the surface

Two parametric rules (R1: axial and R2: longitudinal) add cutting line to bamboo with a
distance variable (L and W). The rules can be applied recursively to generate several
cutting lines (e.g. R2-R2-R1-R1). Within these cutting lines, two-dimensional weights
(shading) from rule R3 will mark the desired shape for the subtraction. Note that the grain
direction of the bamboo, which runs parallel to its longitudinal axis, would make it easier
for the craftsperson to cut the bamboo using rule R2, as forces from the knife would
traverse easily along the grains direction."""

See more about grain property of Bamboo in Dunkelberg, IL 31, Bambus. IL 31, Bamboo. Bambus Als
02
1
Baustoff Bauen Mit PflanziichenStsben.
 59

The form of knowledge in interpreting the cutting process with these rules
can be accumulated into a lattice of rule iterations. The lattice diagram in Figure
23 illustrates a bamboo derivation from nature to several craft components. The
two ends of the diagram represent the original bamboo pole, one projected as a
two-dimensional shape at the top and the other as a three-dimensional object at
the bottom. In between, the rule iteration guides the process from one step to the
next. An observer who sees bamboo as part of nature on the site can use the lattice
to interpret how craftspeople transform the natural object (the bamboo pole) into
various crafts components and utilitarian objects based on the cutting rules in
Figure 22.

Or, the other way around, an observer viewing the finished craft
component can interpret how craftspeople made it from a bamboo pole. Given
existing bamboo crafts, an observer can trace the shape, beginning at the
subtracted shape at the bottom part of the diagram, back to the top of the diagram
(the original bamboo pole).

1
In addition, a desired shape can also be predefined and embedded in the surface at the beginning as the
initial shapes indicate the cutting lines.
60

Figure 23. Lattice of Cutting Rules on Bamboo surface

Using planetary pole analogy, two poles of the lattice represent the original bamboo from
nature in two projections, two-dimensional (top) and three-dimensional (bottom). Between
the two poles is the computation, using the rules in Figure 22. Moving to the left-bound,
cutting lines are generated using rule R2 (longitudinal) recursively. Moving to the right-
bound, the cutting lines are rule Ri (axial) iteration. The iteration in the middle indicates
the use of two rules simultaneously in particular sequences (e.g. R1-R2-R1 or Rl-R1-R2).
Shading in each shape indicates the desired area for subtracting the bamboo. At the end of
the iteration are the subtracted shapes in three-dimensional form. One who sees a similar
artifact to these shapes can interpret how the shape is made by navigating the lattice back
to the top. One who wonders how the bamboo is derived into various crafts can start
navigating from the top and go downwards.
 61

CrossSection View
In addition to the bamboo side elevation, the observer can embed the cutting lines
from the bamboo's cross section. After the pole is cut, an observer can see an
emergent circular shape. Here, the cutting rules are parameterized based on the
knife's position on the surface-edge of the bamboo (cross section). The examples
in Figure 24 show two schemas to place the knife's blade on the cross-sectional
profile (rule R5 and R6). Relative to the profile's center point, rule R5 translates
the knife parallel with the bamboo's profile diameter within a distance (D) while
rule R6 angles out the knife blade from the center by a certain number of degrees

().

R5 --

R6

R5 R5 R5

Figure 24. Cutting rules on cross-sectional profile

Two parametric rules place a knife in the cross-sectional view of the bamboo within
translation variation (D) in rule R5 and angular variation (a) in rule R6. The rules can be
applied repeatedly, as shown in the first computation (e.g., R5-R5-R5) or shifting back and
forth between the two rules in the second computation (e.g., R5-R6-R5). Note that the
infinite order of symmetry of the circle allows a user to apply the rule in infinite ways.
62

Rule R5 and R6 (Figure 24) are parametric. They can be applied

individually and also in combination with each another. The infinite orders of
symmetry of the bamboo profile (i.e., the circle) provide many ways to match the
initial shape in each step. The cutting lines outlined in rule R5 can be placed at
different angles, as any angle will always match the circle. The first iteration
shows a rule (RS) applied recursively in the computation. The second iteration
shows a combination between the two rules (R5 and R6) of cutting lines.

The newly-constructed profiles from the rule iteration generate many

types of bamboo threads. The form of knowledge in interpreting the making of
threads from bamboo is represented with a lattice in Figure 25. The lattice shows
different approaches to placing and angling a knife. At the top of the diagram is
the profile's two-dimensional projection and at the bottom is the three-
dimensional bamboo pole. Shapes in between the top and the bottom show
parametric rule iteration. In this iteration, the left-side iteration shows how the
bamboo profile would look if rule R5 were applied recursively and the right-side
iteration shows how the range of parametric variations of rule R6 would change
the profile. For example, if the bamboo profile is cut with lines shifted away from
the center (R5), the shape is generated in the left bound direction on the diagram.
Similarly, if the profile cut with an angle (R6), the shape is generated in the right
bound direction. If the iteration turns into the different direction, it means two
rules are being used simultaneously.
 63

-0

C,

Figure 25. Lattice of Cutting Rules on Bamboo Cross-Section

The two poles in the lattice represent the original bamboo from nature in two projections,
cross-sectional view (top) and three-dimensional view (bottom). Between the two poles is
the computation using the rules in Figure 24. Toward the left-bound, cutting lines are
generated using rule R5 (translational) recursively. Toward the right-bound, the lattice
shows rule R6 (angular) iteration. The iteration in the middle indicates the use of two rules
simultaneously in particular sequences. Shading in each shape indicates the desired area
for subtracting the bamboo. At the bottom sides of the lattice are the subtracted shapes in
three-dimensional form. Note the size of the bamboo profile becomes finer and thinner as
the rule is applied repeatedly toward each bound.
 64

Reciprocal Causality in Indigenous Materiality

In the previous section, I have described the role of bamboo as a weaving material
in the traditional village through computational design interpretation. Findings
from my EthnoComputation observation are reported with interpretive shape-
rules to describe, explicitly, how matter is translated into form. In these
interpretive rules, the dynamics of Aristotle's four causalities recursively shift by
their reciprocal relationship from one causality to another. The material-cause
(bamboo) provides substance to the efficiency-cause (craftsperson) by which the
craftsperson realizes his/her intentions by shaping the material into a new formal-
cause (new form). In turn, the new formal-cause provides new final-cause (new
purpose), and to derive new purposes, the new formal-cause becomes the new
material-cause.

nature component weave partition

matter#1
.atter#2 - - - - -> form#2 m *atter#3 - - - - -> - - - - - - ->
form#3 = matter#4 form#1 m

Figure 26. Recursive Causalities

The reciprocal causalities can change the agent or craftsperson's affordance in

perceiving matter, form and purpose. Shape polarization from the rule indicates
that the craftsperson (the agent) is not always guided visually, but also haptically
by the flexible behavior of the material. For instance, bamboo-thread flexibility
could expand the vocabulary of form from line to curve, which could eventually
extend the agent's function and material vocabulary (e.g. from rigid planar
partition into a flexible curvilinear basket).

In this reciprocity, craft variation is not limited to, in Deleuze's words,

"differences in type" which merely need dimensional manipulation, but also

"differences in kind, "which show few similarities between one and another, as
05 We can assume that the vast
exhibited in various forms of bamboo crafts.104

10 Gilles Deleuze, Bergsonism (New York: Zone Books, 1988).

'0 Gilles Deleuze, A ThousandPlateaus:Capitalism andSchizophrenia(Minneapolis: University of
Minnesota Press, 1987).
 65

diversity in bamboo crafts is a result of both different types and different kinds of
shapes that are mentally embedded by the craftsperson into the original material.

2.5 The Weaver and the Weave - Sensorial Weaving

In the previous section, I interpreted the way materials are perceived and
transformed, from their original form and function in nature to a craft
component. The different ways of perceiving the material's origins in nature lead
to multiple ways of shaping and crafting the material in Kete Kesu, Toraja. In this
section, I illustrate the ways in which material is transformed into weaving
through the craftsperson's senses, in particular the senses of vision and touch.

In analyzing craftsperson's senses, I position my EthnoComputation study

within the two leading approaches in the "Anthropology of the Senses": 107

(1) Communal sensing versus Personal sensing. Sensory studies can be

situated within the scale of community to represent the way a group's
senses symbolize their cultural model, or at the individual scale to analyze
human perception as a personal sensing endeavor. In approaching my
research subject, my observation leans toward the latter: the weaver's
personal sensing experience. While this research acknowledges that the
weaver's talent is shared by the other craftspeople in the village, my
observation does not attempt to characterize the 'weaving culture in Kete
Kesu'.

(2) Visual perception versus Nonvisual Perception. Visual dominance

has been challenged for being over-studied and overemphasized at the
expense of the other senses, such as hearing, taste, and touch.10 8 The form
of narration in this study relies upon the visualization of shapes as a media
to represent the weaver's sensorial experience. However, in this
visualization, shape is not exclusively registered to a weaver's visual

Based on Sarah Pink's DoingSensoryEthnography,who situates several issues in sensory ethnography,

by highlighting David Howes's and Tim Ingold's differences in approaching and framing senses of their
subject. Sarah Pink, DoingSensoryEthnography(SAGE Publications Ltd, 2012).
" In response to this, Howes, for instance in his book "Aromd", promotes smell as the dominant sense.'"
1
Ingold, despite his advocacy of the important of touch, " suggests positioning vision as part of intermodal
perception.
66

perception, but also to the weaver's haptic perception. In this registration,

shape is considered as a medium to represent both stimuli and reaction
from the weaver's visual and haptic perception.

Sensorial Computation
The weaver in my field study is a native Toraja man who practices various weaving
9 For this study,
activities to produce crafts such as hats, partitions, and baskets.
the man was weaving a basket. The basket consisted of 24 bamboo threads woven
in the triaxial weaving style (See basket in Figure 20). Figure 2726 represents a
process of weaving the basket's base. The process is divided into 24 steps where
the iteration is based on the number of threads in each step (from steps 1-13) and
the number of sides of a thread that defines the perimeter of the base (from steps
14-18).

This 24-step iteration is a typical representation of the step-by-step

instruction format that I mentioned in the introduction.110 It shows the weave's
evolution but not the weaver's interaction with the weave. In particular, not
shown here are the weaver's visual attention and the way he touches the threads
when making adjustments and correcting mistakes as he weaves. Based on my
observation of the weaver-weave interaction, the weaver's fingers, hands, arms,
legs, foot, and eyes were actively and simultaneously engaged each time he added
and adjusted a new thread on the weave; his entire body was engaged in the
process.

This study assumes that shapes the weaver sees and touches in his
sensorial interaction embody important aspects of his weaving. To interpret the
role of these shapes, in the following, I highlight the interaction that occurred in
between the 24 steps in more detail. Specifically, I focus on a weaver-weave
interaction between the third and fourth steps (marked by the red arrow in Figure
2726), and represent the interaction in a computational iteration. This
computational iteration of weaver's seeing and touching activity, hereafter

I Note that in this village, a person's profession is sometimes mixed with his/her other activities. To earn
his living, the weaver needs to simultaneously weave as well as work on other jobs, e.g. harvesting, house-
construction, etc.
no See a step-by-step example in OIta Prefecture- Weaving Technique
 67

referred to as sensorialcomputation, serves as an interface for me to interpret

how the weaver perceives the weave.

I 2 3 4

5 6 7 8

9 19 11 12

13 14 15 16

/l

Figure 27. 24 steps iteration in weaving the base of the basket

This step-by-step model is commonly used in weaving instructions, where the continuous
process is discretized based on the number of threads, e.g., if the numbers of thread
increased, the process translated into one step.

The sensorial computation between the 3rd and 4th steps is discretized
based on the weaver's vision and touch. Figure 2827 shows an example of the

shape that the weaver sees and touches.

68

In seeing: the blue color shows shapes from the weaver's visual attention
on the threads. This shape is represented in one step. As the weaver shifts
his gaze to see a different shape, the new shape would then be added into
sensorial computation as another step.

In touching: the red and yellow strips show the area that the weaver
touches (red for feet and yellow for hands). The haptic interaction is
discretized into one step based on his hands' positions on the thread, and
another step if the position of either of his hands changes.

Figure 28. Sensorial Weight

Weaver sensorial interface with the weave is represented with shapes based on his
eyesight on the weave (blue) and his hand contact (orange) on the threads.

Based on visual and haptic discretization, the weaver's experience is

represented in a sensorial computation in Figure 2928. In the weaver's sensorial
computation, the iteration shows eighteen steps between steps 3 and 4. In Figure
2726, eleven steps are shown for vision (first row), and nine steps are shown for
touch (second row).

In each step of the sensorial computation, I add labels to indicate a certain

weaving condition, such as position of the thread (pos), parallel condition
between two threads (prl), alignment of a thread with other shape (al), distance
into a specific location (dis) and crossing between two threads (crs). For each
label, the signs (+) and (-) indicate whether the weaver achieved the desired
condition (+) or not (-). For example, prl+ means the shapes are parallel and prl-
means the threads is unparalleled.
 69

I 2 3 4 5 6

\N

7 8 9 10 11 12
/

x\ N,'
/

\K
/ v N

\ \~,
V/

13 14 15 16 17 18

'\ \
N /\\
N
/

Figure 29. Sensorial Computation

The actual continuous process between step 3 and 4 is discretized based on weaver's
sensory attention, i.e., if the weaver changes his visual attention or his hand contact on the
thread, I discretize the process as one step. The first row represents the weaver's visual
interaction with his weave; the second row represent his tactile interaction (his hand-touch
on the threads). The two shapes at the end of the iteration represent the initial and final
condition between steps 3 and 4 (see Figure 2726).
70

1 2 3 4 5
crs- CS+ CrS+

New sitape is New shape as

two inetat sapes crossed UNDER crossed OVER
are not crossed the initial shape te initial shape

pos- pos+

f L/

weaver teet stands on alt hand pull initiat shape Adding new shape
existing shape right hend pick new shap edo the exist"i shape

Figure 30. Continuous Attention Span

The weaver's focus on the weave shifts back and forth between visual and haptic sense.
The long arrow in the top row indicates the weaver's eyesight continuously fixed at the
weave, while his hand adds a new thread to the weave (yellow color in the bottom row). In
the subsequent steps, the long arrow in the bottom-row indicates the weaver continuously
sensing, his hand on the thread, while his visual attention focuses on the knot condition
(blue) (see the label: crs- -* crs+)
.

The long-dashed arrows (====>) that subsume more than one step on the

above or below row indicate that the weaver is continuously sensing on that
particular row. For instance, in steps 2 to 4 in Figure 3029, while the weaver's hand
is adding a new shape in step three in the second row, his visual attention remains
focused on the area where he will add the new shape. Therefore, step three in the
first row is represented with a dashed arrow. In the second row, another dashed
arrow in steps 4 and 5 represents the continuity of the weaver's tactile attention.
In these steps, he continuously holds the thread at the same spots while he uses
his vision to evaluate whether or not the new and existing threads are crossed into
a knot.

Permanency and Temporality in Mental Shapes

Sensorial computation of this weaver highlights some shapes involved in the
adjustment of the weave. In representing these shapes, I incorporate information
 71

from my interview with the weaver about the basket he is making. As a weaving
goal, the basket has a cylinder-shaped dome topped by a circular lid. Thus, to
weave a basket properly, the end of each thread should coincide with the circle.

However, in step 7, there seems to be a misalignment between the thread

and the circle. To fix this, he pulls his arms back to align the shape and the circle
(Figure 3230). In step 8, the labels dist+ on the second row represent the desired
distance to align the thread is achieved, while labels al+ in the first row confirm

the alignment of the thread with the circle. This adjustment implies that as he
weaves, he is always considering the final shape.

In addition to the weaver's mental image of the basket-lid as the final

form, there is an indication that mental shape is not only embedded in the

thread's shape, but also in the space around the threads and in-between each
individual thread. In an interview following this observation, I showed the weaver
an image of unfinished two-dimensional woven lines and ask him to guess what
image it was (Figure 31). The weaver looked at the image carefully but had

difficulty identifying this object as a weaving pattern. He guessed that the image
was a partition, a road, but not an interwoven pattern.' When asked why he did
not perceive it as a weaving, he said because it doesn't have a contrast color in the
space between the threads. His answer raises an interesting point concerning the
2
attention he pays to the space between the threads while he weaves." In the

following, I incorporate this answer in the weaver's sensorial computation.

Figure 31. Ground shapes as mental shape in weaving

The an unfinished three-by-three weaving pattern (left) that was shown to the weaver. To
recognize the weaving pattern, he needs to see the shade in between the yarn (middle) to
perceive the complete weaving pattern (right).

" I also posed this question to other craftspeople on the site, and most of them were able to figure out the
weaving pattern in the image.
"2 I have described in more detail about how weaving is visually perceived according to Gestalt Law: see
Rizal Muslimin, "Recursive Embedding of Gestalt Laws and Shape Grammar in the Weaving Design
Process," vol. 1 (presented at the 30th eCAADe, Prague, Czech Republic, 2012), 443-49. And Rizal Muslimin,
"Gestalt Grammar in Weaving Design: Leveraging visual Affordance in Analysing and Synthesizing Weaving
Design Based on the Laws of Pragnanz and Shape Grammar (poster)" (presented at the 5th International
Conference On Design Computing And Cognition (DCC), Texas, USA, 2012).
72

6 7 8
al- al+

Y/

shae is net shape ts

ahigied with part allgrod with fte
of the goal-shap par of goal-sae

dist- dist+

Right hand slighty

Left hand release the initial shape pull the thread backwart
Right hand keep the w shape (lefthand is hoding a cgarette)

Figure 32. Visual-Haptic Coordination

In the first row, a circle is embedded in the weave as a mental shape representing the lid of
the woven basket. From this embedding, a misalignment position between the physical
thread and the mental shape is sensed. This visual sensing gives clues to the hands as to
which direction the thread needs to move to fix this condition. Here, the long arrow on the
bottom row represents the physical intervention to fix the misalignment, where the weaver
pulled the thread until both thread and the circle aligned (see the label: al - -+ al+) .

In step 9 in Figure 3331, an unequal knotting condition occurred. The area

of the new knotting pattern was larger than the area of the existing pattern (i.e.,
the triangular area between the thread). Here, based on my interview with the
weaver about the space between the thread, I presented this condition by adding
planes on the empty triangle area between the two threads (i.e., triangular
planes) (Figure 3331). These planes contrast the unequal knotting pattern. To fix
the asymmetrical knot, the weaver placed his hands on the newly- added thread
and then pulled it inward to firm up the knot. As a result, both triangles appeared
to be equal.
 73

9 10 11 12
ar- ar+

new gap is not now gets relatively at

snow
as smnall as the euxetmg gap the same size as the ealtig gap

dist- dist+

both hand touch both hands push the new shape

new shape on its side inward froM t side

Figure 33. Visual Assessment

The unequal knot in step 9 is represented by highlighting the negative space of the weave
(the triangular planes blue). This particular step interprets how the knot condition might
be evaluated more in a geometrical way rather than in a structural way using the knotting
pattern (see the equating triangle in shape labeled with ar - and ar+).

The sensorial computation from steps 1 to 18 interprets embedded mental

shapes involved during the weaving process as perceived by the weaver. The
shapes are projected visually as mental guidelines that can be permanent, as in
the case of basket-lid, or temporal, as with the triangular pattern of the knotting
pattern. In this case, the weaver uses the emerging temporal shapes during the
weaving process to help him weave his permanent shape. While embedding
emergent shapes in design exploration is usually to inspires new design,
recognizing temporary emergent shapes in this weaving activity is aimed more at
controlling or shaping the predefined design goal.'

..In contrast, in random-weaving or sculptural fiber arts, the final shape may not be defined at the
beginning, but rather gradually designed upon a series of emergent temporal shapes in an exploratory
process, which results in an art piece (source: based on my experience as a student in 3D weaving workshop
taught by Nathalie Miebach, MassART.)
74

14 15 16 17 18
pri- prl- prl+

new shape Is nrot paralel wirth unparattel suapes is new shape is now
part of in"ta shape being rnonillored parallel withn Initial shape

dist- dist+

both hands push tre new shape eft hand push the shape nward
inward krtn its sift right hand stays in fth samai part

Figure 34. Embedded and Embodied Shape

Notice how the weaver first gazes at the end of the thread to assess the unparalleled
condition and embed the 'V' shape, and then shifts his gaze to the knots in the center of the
weave and applies physical intervention to straighten the threads. Although he no longer
see the end of the thread, the weaver had embodied the 'V' shape in his mind, knowing that
parallelizing the middle part of the thread means the end of the threads will also be
parallel.

X
/

'/

//

Figure 35. Emergent perceptual shapes

Shapes from my interpretation of the weaver's sensorial experiences show that the weaver
may perceive the threads on the weave not as an atomic shape, but rather a continuous
shape that could contain many subshapes, either in the physical shape (the thread) or in
the space surrounding the threads (e.g., circle and triangle).
 75

Haptic Shape and Haptic Ambiguity

According to Gibson's haptic theory, the weaver senses the shape on the weave
using his/her haptic apparatus in three ways." 4 First, the weaver's hand has both a
passiverole, to feel the stress/strain of each knot, and an active role, as fabricator
of the weaving. Second, the thread, for weavers, is not merely a construction
material but also something that performs as the cutaneousappendagesand
receptive units, transmitting structural data of the woven object to the weaver's
hands as they pull and stretch the thread. Third, the weaver's body performs
dynamic touchingas skin, joint and muscle act together with different degrees of
sensitivity. The stretching muscles transmit stress/strain data, the joint's rotation
transmits the angle and position coordinates and the skin pressure provides
contact information.

To understand the role of these haptic apparatus in more detail, I interpret

the weaver's haptic and visual experience from step 7 to step 8 in the following
sequences (Figure 34):

In step 0 to 1, the weaver senses the shape in the area where the hand
touches the thread. The haptic apparatus involved in this step are mainly in the
palm and fingers as the skin makes contact with the thread's surface. Underneath
the skin, the weaver senses the thread's stiffness via the muscle's contraction and
compression so as to adjust the level of tension of the grip on the thread.

In step 2 to 4, the weaver senses a shape between his visual attentions at

the end of the thread and his hand grip as he starts to push the thread to close the
gap with the circle (dark-blue shape). In step 5, as his hand and arm muscles
adjust the force of the grip and pull the thread (through muscle tension and tactile
feedback), his visual attention transmits information about how far the arm
should push the thread and how much force should be applied to achieve a
desired distance.

" See further explanation of Gibson haptic theory in its relation to structural property of weaving in Rizal
Muslimin, "Learning from Weaving for Digital Fabrication in Architecture," Leonardo43, no. 4 (July 16,
2010): 340-49, doi:10.1162/LEON-a_00007.
76

In step 6, the weaver senses a shape from part of the thread between the
knots and his grip. Here, in addition to the hand and finger muscles and nerve
endings, the arm muscles, tendons, and joints are also involved in sensing the
friction of the knotted area. In step 7 and 8, the weaver moves his/her arm back
and forth to adjust the force so as to pull or push the thread through with the
proper force.

This synchronized haptic and visual coordination guides the shape

formation. The visual and haptic apparatus feeds information back to the brain
and instructs the arm/hand/finger muscles as to how much force should be
applied. In this case, the weaver's muscular force should be larger than the sum of
the knot's frictional forces to move the thread. All of this occurs while the visual
feedback informs the weaver when this force needs to be released, which is at the
exact time as when the shape formed by the thread and the part-shapes of the
final design (circle of the lids) are aligned in step 8.

The way the weaver positions his body to correspond to the thread
indicates that the haptic-visual coordination might have been mentally calculated,
even before the weaver takes action. The weaver grips his dominant hand on the
weave not too close to the last knot so as to easily push it, yet not too far from the
knot so as to maintain the arm in a stable position and to maneuver the thread.

The weaver's hand position on the thread is in some way analogous to the
way an artist holds a pencil close to the tip to draw fine details, and much farther
from the tip to make broad strokes in a rough sketch. As the weaver constantly
receives feedback from the weave's stiffness through his haptic senses, he can
shift the degree of freedom between the arm and the palm. In addition, the left
hand lifts the thread to helps reduce the friction between the thread and the floor.
The left hand also constrains the thread's degree of freedom to ensure that the
thread moves linearly rather than rotationally, to close the gap.
 77

5 6
4

8 9
7

Figure 36. The rise of the haptic shape

The images show a sequence of the weaver's haptic experience in sensing and adjusting
the misalignment in the weave (see also Figure 3331). In the first row, after the weaver
holds and sees the thread (orange and blue), he projects a circle, as the representation of
design goal, and notices the misalignment (dark blue). In the second row, he starts sensing
part of the shape that needs to be moved (dark grey area between his grip and the end of
the thread). As he starts to push the thread forward, the material responds, as a result of
the interwoven knots' friction forces (two red-squares), transmitting the stress information
from knot to the weaver's senses. Based on this information, he then focuses on the
thread's reaction between his grip and the knots (black), and gradually increases his
pressure force to the black area until the thread is placed in the correct position (indicated
by the disappearance of the dark-blue area).

HapdcAmbiguity
My interpretation of the weaver's movement and of the angle of his body and
limbs helps me to deduce how the weaver senses the shape through his haptic
apparatus. In this study, haptic sensing is represented not as a single event but as
an assemblage of multimodal sensing experiences, each of which works in
collaboration with the others to define the shapes. As there is more than one way
78

to see and embed shapes, 1 the assemblage of haptic actions suggests that there is
more than one way to touch or haptically sense shapes, as well. The term sensing
here is not the same as when a blind person senses or assesses a surface of the
ground with a walking stick.1 7 In this case, the weaver can see and touch the
thread but has yet to decide which part of the thread he should focus on, and how
that part will interact with his hand and the other threads. I call this haptic
ambiguity. In this ambiguity, the shape that he haptically sensed could involve the
whole thread, the part gripped by the hand, the part between the knot and the
grip, the part between his visual attention at the end of the thread and the knot, or
many other possible parts. Different perceived shapes are illustrated in Figure
3735 to show that in searching for an optimal haptic shape for subsequent physical
action, the weaver senses the thread as a continuous, not a discrete, shape.

Weaver Sensing Material Sensing

supported

fixed

pushd ? bo

fixed

Figure 37. Weaver-Weave Sensorial Exchange

This figure illustrates sensing from the thread point of view on the right-column images as
the thread's material properties corresponds to the weaver sensorial experience on the
left-column images (see Figure 34). As the weaver senses and perceives the thread
differently, the thread also "feels" the weaver in different way (e.g., being supported, being
fixed, being pushed, etc.) and responds according to the weaver's reactions.

's Stiny, Shape.

11 Gibson, The Senses Consideredas PerceptualSystems.
 79

In terms of structure, haptic ambiguity also affects the way the weaver
perceives the thread's stiffness. The way the weaver touches the thread changes
the thread's stiffness through the reaction applied by the weaver (In structural
terms: by applying a fixed support, simple support, pinned support or roller
support).118 In a way, the thread is also 'sensing' the weaver. The thread, in turn,
responds by allowing its shape to be altered by the weaver's action, which is vital
input for the weaver's sensorial experience. The thread's different structural
configurations in the right column of Figure 3735, show that the weaver's touch
affects structural sensing of the material. As the weaver applies different types of
forces to the thread, (i.e., pulling, pushing, pressing, rolling), he senses the weave
as a perceptualstructure.What is considered rigid in one step could be perceived
as flexible in another step, and vice versa.

The contact area of the weaver's haptic apparatus does not always involve
just the fingers and palm of the hand. In this observation, the foot also plays a
significant role, as the weaver always squats on the threads to firm the weave
position. Although his foot seems mainly static, the weaver's body-weight
pressure (through the foot) stabilizes the thread positions on one end while his
hands manipulate the other end. This position in turn may absorb the stress in the
body of the weaver while the hand manipulates the threads. Thus, through the
foot, the thread transmits more physical information to the weaver's intermodal
perception system.

In addition, the floor surface in this observation also serves as the

weaver's external haptic apparatus. This is particularly noticeable during the base-
making phase of the basket because the thread tends to bend and curve naturally,
thus making it difficult to control. Here, the floor acts as a counterpoint to keep
the thread straight and level with the planar surface of the basket-base. The
weaver keeps the weave on the floor until the base is finished. When the weaver
finishes the base, he stands up and lifts the weave off the floor to start weaving the
side of the basket. Now, the weave no longer needs the floor's counterpoint for

118Mario Salvadori, The Art of Construcdon:ProjectsandPrinciplesfor BeginningEngineersandArchitects,

3rd ed (Chicago: Chicago Review Press, 1990).
80

support. Instead, the weaver benefits from the thread's bending behavior, which
helps him form the vertical side of the basket.

Oriented weaving
The previous section discussed how shapes are perceived through the visual and
haptic apparatus during a short period (i.e., in-between steps 3 to 4 in Figure
2726). In the following section, I shift the time frame of my observation to a longer
period: the 15-minutes process of weaving the base of the basket (from step 1 to
step 13 in Figure 2726). During this longer period, the weaver moves his position
several times from one spot to another. His different positions appear to offer
different orientations to the weave. In this section, I discuss how the weaver
orients his body in relation to the weave, based on the Gibson theory of 'orienting
system.'

Gibson proposes an "orfendngsystem"to explain how we move, turn, and

orient ourselves to a new environments on three levels: (1) permanent orientation
to the earth (e.g., to gravity and the support surface such as a grassy hill, concrete
street), (2) temporary orientation to events and intriguing objects and (3)
locomotion orientation towards a certain goal (e.g., birds flying south, finding our
way home while walking in the dark, etc.)." 9 In this framework, the weaver's
orientation changes back and forth between these levels during the weaving
activity. In relation to the ground, the floor helps stabilize the weaver's body while
the squat pose serves to stabilize his position. In relation to the object, the
weaver's body and eyes are oriented toward the weave in each position (e.g.,
squatting, standing, moving). In relation to the goal of creating a basket, his
squatting position allows him to easily bend the thread to create the desired knots
and to create the shape of the basket. (i.e., the weaver squats when he makes the
basket's base, and stands when he makes the basket's side-surface.) Thus, his
permanent orientation with the surface of the ground, temporary orientation
toward the weave and locomotion orientation in making a basket, play significant
roles in his position.

"'Gibson, The Senses Consideredas PerceptualSystems. p.59

 81

Gibson's orienting system explains how the weaver's body is oriented in

space as he weaves. However, this explanation focuses more on how the weaver's
body orientation works and less on how the weave is being perceived in different
weaver orientations. To analyze how the weaver's and the weave's position are
reconfigured in the crafting process, it is necessary to widen the observation
parameter to include not only the weaver's body but also the weave itself. In the
following, as the weaver moves to different positions (standing, squatting), the
direction of his orientation toward the weave is analyzed.

Figure 3836 shows a diagram of the weaver's orientation, as he moves to

different positions toward the weave. As seen in the diagram, the weaver's
frequent orientation might be constrained by the location of thread-materials
stacked near this position. In addition, the camera and my position (which is
across from his dominant position) may have caused him, out of politeness, to
face me, the observer.
82

0 X.

Figure 38. Weaver's motion from the observer's eyes

The weaver's orientation as he moves to multiple positions. The pink ellipse indicates the
weaver's position at a point in time and space, and the green line indicates newly added
thread in each position. The images are then superimposed (bottom) to shows the weaver's
dominant position (the darker thread indicates earlier threads that were added to the
weave, e.g., darkest shape is the first thread).
 83

The superimposed images shown in the bottom of Figure 3836 describe

how the weaver orients his body toward his work, both functionally to remain
close to the stacks of weave material, and socially, so he can continuously face me.
However, this illustration does not show how he orients his body in relation to the
weave.

For a different perception of the weaver's orientation, I aligned my view

with the weaver's views of the weave. In particular, I look at how a new thread is
added at each of his different positions. To align our views, I rotated each image in
Figure 3836 so that my orientation axis coincides with his orientation axis in a line
(see Figure 3937).

I then superimpose these rotated images to identify how the weaver

perceives the weave every time he adds a new threading position. The
superimposed images in the bottom of Figure 4038 show two dominant thread
orientations. The weaver uses these two orientations more frequently, relative to
the other positions. The dominant positions indicate that, instead of making
random movements, the weaver might be consciously positioning himself with a
particular orientation to the weave.

0
0

0 *O

Figure 39. Weaver's - Observer's Orientation Alignutent

The weaver's position is justified by rotating the unparallel weaver's orientation (purple-
arrow), according to the observer's eyes, into one uniform orientation (vertical axis)
perceived by the observer. This way, the observer's view and the weaver's view toward the
weave are synchronized in each of the weaver's positions.
84

o
.r4

Figure 40. Weave position from the weaver's eyes

The weaver's orientation to the weave from Figure 3836 is justified using the rule in Figure
3937 so that each image would represent how the weaver sees the weave. Each justified
image is then superimposed over the others to create one multilayered image (bottom).
The green shapes represent the position of the new-added thread in each of the weaver's
positions.
 85

Symmetrical Stimulus and Asymmetrical Reaction

Gibson's principle of symmetrical stimulation states that we balance sensory
stimulation bilaterally, and we adjust our posture to make the input to the central
nervous system symmetrical.120 Gibson illustrates this by showing how an air-
stream influences the body to follow certain odors, how our ears respond to sound
by turning the head toward the sound, and how our eyes move to balance the
array of light so as to recognize objects. The principle of symmetrical stimulation
parallels the dominant thread's position in the weaving process. The first
dominant position of the threadis perpendicular to the axis symmetry of the
weaver's body while the second position is tilted toward his right hand (Figure
4139).

10000' -000' X. X

Figure 41. Asymmetrical Making

Two dominant orientations related to the weaver's body's bilateral symmetry (left) and to
his right handedness (right) from the superimposed image in figure 40. The orientation of
the new threads on the left (horizontal arrow) is perpendicular to the weaver's body, thus
allowing his right hand and left hand to manipulate the thread together. The image on the
right shows that the weaver always adds the new thread with his right hand, and provides
and explanation for the weaver's actions; many new threads are positioned on the right
side, instead of on the left or on both the left and right sides.

As seen in Figure 4139, the weaver's orientation seems to correspond to

the weave symmetry, the weaver's body's bilateral symmetry, and the asymmetry
of the weaver's right-handedness. In the left diagram of Figure 4139, the weaver

1" Ibid.
86

positions some new thread perpendicular to his body's bilateral symmetry. This
position helps to ensure proper thread position; if a new thread is perpendicular
to his body orientation, then the thread is symmetrical with the weave. The right
diagram in Figure 4139 shows that the asymmetrical position gives the weaver an
optimal distance for his right hand to maneuver the threads (e.g., to pull the new
threads inward toward his body).

The motions of the weaver, the thread, and the weave simultaneously
complement each other in this view. The bilateral symmetry of the weaver's body
and the radial symmetry of the weave allow him to make knots in any position if
his left and right hands are equally skilled (ambidextrous). For an ambidextrous
person, the location of each thread on the weave would be in reach of his two
hands, and therefore, the threads would be equally distributed around the weave
from one position only.

However, because he is right-handed, the weaver can apply a knot only in

a particular gesture according to his right-hand's distance; consequently, he
constrains his movements to a particular orientation to the weave. Thus, to apply
knots to the rest of the weave, he then needs to move around the weave. The
number of different ways to add a new thread, given the weave and the weaver's
symmetry, are reduced by the weaver's asymmetric right-handedness in the
making process. This asymmetrical-making process may contribute to his 'making
style'in weaving.' 2
'

2.6 Discussion

In this chapter, I have rendered my interpretation of traditional weaving on the

site (Kete Kesu, Toraja), from site reconfiguration to bamboo materialization to
weaver-weave sensorial experience. This EthnoComputation underlines the role
of sensing and calculating in the flux of form and matter.

" This symmetrical reduction of the weaver's body is analogous to the way symmetrical reduction in Shape
Grammar helps to define rules for a certain design language.' For instance, the order of symmetry of
square (eight) indicates eight ways to match the shape in shape-rules, through which many design
possibilities could be generated. Design languages using the square can be defined by constraining the
1
square's symmetry.
 87

Sensing and Calculating in EthnoComputation

I begin by portraying the site as a schema, which consists of a set of shapes that
have the same level of attention. I call this a non-judgemental view. This portrayal
does not diminish the context to become site-less. Rather, it shifts the context
back into a pre-sitecondition, where meanings of objects on the site are yet to be
identified or labeled. Next, I define the site's elements (e.g., weaver, weave,
ornament, material) and associate them to form certain relationships (e.g.,
weaver-weave; weave-material; ornament-weave; weaver-ornament). These
relationships provide me with different frameworks to perceive the context in this
EthnoComputational study.

Based on these frameworks, I present the role of human interaction in

crafting through indirect observation of the bamboo materialization and direct
observation of the weaving activity. Both crafts objects and craftsperson's
activities serve as my computational interface to analyze the crafting process.
With the absence of craftspeople in my indirect observation, I interpreted the
rules of how shapes in nature (bamboo trees) are transformed into diverse craft
components through the available woven crafts. With the presence of a
craftsperson in my direct observation, I studied the ways in which the weaver's
activity implies how he perceptually acquires the shapes through his senses.

Hylomorphism Revisited
This process of interpreting a traditional weaving craft was framed on two states:
continuous state, by using Ingold's EWO approach in navigating the matter-form,
and discrete states of Aristotle's four causal functions: material, formal, efficient
and purpose causalities.

The material cause in this study is examined as I perceive bamboo

transform from one form to another. In the case of weaving, my perception of the
material shifted recursively from a bamboo tree to a bamboo pole to bamboo
thread. Because a material-cause can recursively change, in turn, the cause can
extend the singular explanation of formal causes of weaving into numerous
explanations based on what the material can do (its uses, its flexibility) in different
states. For example, a long, rigid bamboo pole can be used to build a linear
88

structure for shelter; long, flexible bamboo threads can help weave a curved
surface for a basket; a planar, flattened bamboo pole can be employed to build a
two-dimensional partition in a house.

The efficient cause (agent) is represented not merely by the agent's

(weaver's) intention but also by how the spontaneous agent's intention emerged
from a reciprocal action-reaction in a weaver-and-weave interaction. This action-
reaction process is rendered through the weaver's sensorial experience (vision
and touch). In supporting the weaver as the main efficient cause, the bamboo
material property and the surface of the floor are also considered part of the
agency that affects the shape of the weave.

The final cause (purpose) is what use an object or creation offers. The
interwoven-craft in this village has various purposes (e.g., to support a sturdy
shelter, to carry things, and to decorate.) However, these apparent purposes do
not always explain why the weaving technique is chosen by the craftspeople,
because other non-woven crafts serve similar purposes and even have similar
forms. For instance, in Toraja's traditional architecture (the Tongkonan), both
wooden Tongkonan and interwoven-bamboo Tongkonan share similar forms;
thus weaving is not unique to the Tongkonan's form. However, in the other
causes, such as material and efficient causes, one can observe certain features,
such as the economic aspect of the construction process, that are unique to
weaving. Building a Tongkonan with bamboo can save construction cost and time
because of the abundance of low-cost materials (bamboo surrounds the village)
and the high supply of a labor-force (weaving can be performed by many people).
Material and labor affordability could suggest other final purposes, such as
funding issues in cultural sustainability to ensure that the cost of building
temporary Tongkonan to accommodate guests during ritual ceremonies remains
affordable.

Thus, by investigating the other causes - material, formal and efficient-

we can identify final causes, which in this case may not necessarily be drawn from
the artifact as a final product (e.g., a house is used to shelter people, a basket is
 89

used to help a person carry things), but rather in the process between matter and
form.

Ethno-Computation as Cultural Mimesis

The study employs a computational framework based on my personal experience
in interpreting traditional weaving. My interpretation, however, does not intend
to identify and verify the original rules of traditional weaving in order to establish
'a science of weaving.' It is meant, instead, to reconceptualize ways with which to
appreciate the craft of weaving. In this new appreciation, we can then reestablish
and enrich our knowledge to develop weaving, and use this new knowledge to
further interpret and reinvent similar artifacts. Through such reconceptualization
and enrichment, we can learn multiple ways to apply weaving techniques, even
beyond the traditional context (e.g. modern architecture, engineering, etc.).

Ethnos encompasses more than ethnic groups; it encompasses humans,

individually or collectively. In EthnoComputation, culture is defined as the way
humans represent how they understand their environment and craft is one of the
manifestations of such culture. This case study of traditional crafts-making should
not be viewed as an attempt to place a rigid boundary between traditional crafts
and modem technology, but rather as a means to enable people from different
spheres to exchange methods by which to compute shapes in crafting. Just as this
study represents the way the Toraja people manifest their understanding of the
environment through crafts, so EthnoComputation might be used to calculate the
interaction between humans and crafts in an advanced engineering context. For
instance, EthnoComputation can help to understand how engineers study and
represent the renowned nature-properties in biomimetic technology, or how
fabricators arrange their rapid-prototyping tools in digital manufacturing.
Recalling Herbert Simon's example, instead of seeing the bird as an object, an
EthnoComputation study would focus on the interaction between the aerospace
engineer and the bird as a research subject, in particular analyzing the way in
which the engineer perceives the object in nature and translates it into an
artificial object. Interpretive rules and schema from these examples could serve as
90

design currency in exchanging design logic and improving design affordances

across cultures and disciplines.
 91

3 COMPUTATIONAL WEAVING

3.1 Outline

This chapter aims to provide computational design methods to reinvent

traditional weaving in architectural design.

The ubiquity of weaving across scales in Toraja implies the presence of an

intelligible method shared among craftspeople, which allows them to use a
similar method in various applications. For instance, the method used to weave a
basket is similar to the method used to weave a wall-partition, floor mat, and
accessories. In this application, weaving principles embody different material
properties, e.g., more rigid for shelters, and more flexible for wearable items.
Through the tacit environment, the weaving method's comprehension and the
weaving design intention are mediated during the making process.

In this chapter, I provide computational design methods with the intent of

aiding designers to reinvent architectural weaving in an explicit manner; this is
comparable to my interpretation of the way the traditional knowledge assists
Toraja craftspeople in advancing weaving crafts in a tacit manner.

The chapter is divided in two sections:

In the first section, I describe the process of a basic Weaving Grammar

development. The Basic Weaving Grammar provides methods to reinvent weaving
design by reformulating the way thread is composed on the weave, using a set of
rules. The grammar development focuses on the moment in which the weaver
adds a new thread to develop a particular weaving geometry. From this focus, I
further discuss how the grammar evolves, from the weavers' mental shapes and
the weave's physical shape, to a set of rules using Euclidian transformations and
principles of symmetry. Several examples demonstrate the way rules in the
Weaving Grammar generate weaving design computationally. Using the
92

generative rules, traditional weaving designs as well as other designs can be

computationally analyzed and synthesized.

Following the description of Basic Weaving Grammar development, I

present four case studies in which I examined the grammar's contextualization for
architectural application. To evaluate how a builder learns how the weaving
grammar works, I conducted a weaving grammar workshop, where I taught the
grammar to people from different craft disciplines in Toraja, and to children from
different age groups in Cambridge. To investigate whether the architectural
weaving designs generated from the grammar are buildable, I inspected the
design's structural integrity with a Finite Element Analysis and Compressive Test.
With the rule readily understandable and the design structurally acceptable, I
further tested the design's constructability by building a mock-up installation in
Toraja. In the construction process, I inspected how the rules assisted the
construction process. The two workshops, in Cambridge and Toraja, showed that
individuals can readily learn shape rules, either visually, with paper and pen, or
physically, with wooden modeling pieces. The installation inspection on the site
showed that the rules helped the builder to assemble a design efficiently with a
minimal learning curve.

In the second section of this chapter, I explain the method of reinventing

weaving design using an abstract weaving grammar by perceiving weaving
material properties in different ways. This method was inspired by the way the
traditional weaver engaged with his materials (described in the previous chapter).
Different ways of perceiving materials allow traditional weavers to apply weaving
principles to different forms, such as partitions, roof structures, baskets, mats,
hats, etc. Thus, the aim of formulating material perception into an abstract
weaving grammar in this section is to accommodate different physical
requirements in architectural weaving, in particular, the requirement pertaining
to material properties (e.g. lightweight, flexible, rigid). Learning from playful
material application in traditional weaving, architectural weaving is hoped to be
adaptive to different types of material.
 93

Towards this goal of adaptability, I incorporate Finite Element Method

(FEM) approaches to the idealization of a shape, and its mechanical properties,
into discrete model. In this section, FEM idealization principles are represented
with shape grammar notation. With this notation, I develop a grammar to
reinvent weaving, benefiting from both shape grammar's visual calculation and
FEM's mechanical representation. The combined visual and mechanical
representation in the grammar allows a weaving design to respond to the external
forces on the weave based on its material composition and geometrical
configuration. Several examples of the grammar computation in this section
reveal the relationship between weaving and the other crafts, such as tensegrity
and folding, in a generative and explicit manner.

In the last section of this chapter, I reflect on the role of computational

design in reinventing architectural weaving.

3.2 Basic Weaving Grammar: On the Language of Weaving Design

Introduction
This section illustrates the process of developing a basic weaving grammar based
on the moment at which the weaver adds a thread to the weave in developing a
particular geometry. The aim of this grammar is to provide a comprehensive
architectural weaving design methodology, from formulating design intention,
defining weaving action and evaluating weaving condition.

Shape Grammar computation methods need certain basic requirements,

i.e. initial condition, shape rules and shape computation (see page 22 and page
44). To develop a comprehensive weaving grammar, the cascading processes from
weaving intention to weaving action to weaving evaluation in the grammar have to
be well integrated. To serves as an initial condition, the weaving intention needs
to be visually defined. To initiate a computation, the left-hand shape of the rule
needs to match the initial condition. To compute the shape recursively, the right-
hand shape needs to match the left-hand shape within the grammar.
Furthermore, to evaluate whether the generated shapes are woven, the grammar
94

should be able to evaluate the weaving condition and follow up the condition
properly.

In the following, the development of rules for Weaving Grammar is

rendered in three phases:

Phase 1: Defining Weaving Intention

Phase 2: Developing Weaving Action
Phase 3: Evaluating Weaving Condition

Following the development of the grammar, case studies of the Weaving

Grammar application in traditional context are discussed.

Phase 1: Defining Weaving Intention

Phase one provides a computational method to generate certain geometry for the
weaving grammar as the design intention -- specifically, in developing the left-
hand shape for the rules in Weaving Grammar.

The computational method discussed here is inspired by the mental and

physical shapes from the sensorial computation in the previous chapter. The
EthnoComputation study shows that the weaver utilizes mental shapes to make
sure the threads are woven into his design intention, a triaxial woven basket. One
of the mental shapes interpreted for this intention is the negative space between
the threads - the polygon. Considering the key role of the polygon in weaving
geometry, this phase begins the computational process by parameterizing the
polygon (Figure 4240).

The polygon in this grammar is construed as an assemblage of isosceles

triangles that are rotated using the polygon's center point. The rotation angle (a)
equals 3600 divided by the number of polygon sides (n).
 95

& P

mental polygon triangle rotation line

shape point

physical weave cross thread

shape

Figure 42. physical shape and mental shape

Mental shape is represented as an n-sided polygon. The parametric description of the
polygon can define different types of weave (e.g. bi-axial and pentagonal weaves).

The rotation of the triangles in constructing the polygon is parallel to the

rotation of the threads in weaving the weave. As the polygon's side coincides with
the threads, the number of the threads equals the number of polygon sides. As the
triangle baseline coincides with the partofthe thread that defines the polygon,
the triangle effectively focuses the user's visual attention on that particular part of

the thread. In addition to the triangle, the role of the point also corresponds to the
thread's addition. The point serves not only as an axis for the rotation, but also as
a visual reference in evaluating the polygon's symmetry as the weave develops.
With this mental-physical shape association, the shape relationship between the

thread and the triangles can now be represented. In the following, I use a square
as a polygon to demonstrate how a set of rules adds the triangles and the rotation
point to transform the thread parametrically (Figure 4341).

Rule R1 establishes the coexistence between the mental shape on the

thread (line) and off the thread (triangle), by adding the triangle to coincide with
the line (the line's length represents the thread's length). This rule is defined by
parametric descriptions pertaining to the polygon: (a) number of a polygon sides
96

(n) and the triangle's angle: a = 3600/ n.122 If the number of polygon sides changes,
for instance, from pentagon to hexagon, the shape of the triangle will change
accordingly.

Rule R2 adds a rotation point P within the triangle as a reference to weave

the thread into a polygon with rotational transformation. The role of the triangle
in this rule is to provide boundaries for placing the rotation point. Placing the

rotation point anywhere within the triangle, and within cx 3600 / n, will guarantee

the rotated threads intersect with the others. Otherwise, if the point is placed

outside of the triangle or a> 3600 / n, the rotated threads will not coincide with or
intersect one another; such threads cannot be woven. The point location's
parameter (xy) in this rule provides options to generate the thread into various
weaving styles based on different rotation axis (a discussion on how the point
location affects the weaving design is provided in page 103 to page 108).

P(xy) 0 0

R 1 R 2 R 3 R 4

Figure 43. Rules to define the left-hand shape Weaving Intention

Given a predefined n-polygon, a thread as a line and the mental shape as a Weight is
defined as the basis of the weaving geometry. R1 adds the mental shape as part of the rule.
R2 adds the rotation point P to generate the weave within the triangle. R3 and R4'erase' the
mental shape.

Rule R3 erases the triangle to signify that the process of defining the basic

geometry is finished. In other words, the question of 'what type of polygon should
I weave'is answered. The absence of the triangle fixes the point location because
the point can only be changed with the triangle present (as notated in rule R2).

Rule R4 deletes a point for two reasons: First, the deletion of the point

returns the line into its initial condition. Thus, makes the line is ready for yet
another new intention development. Second, without a point, the line can match

" The mental shape can be any type of triangle (beside isosceles) to define various mental shape. There
could be more than one rule defining the shape. However for clarity reason, here, I focus only on the
isosceles triangle and basic polygon.
 97

any lines in the computation. Thus, it makes the weaving intention becomes less
deterministic.

The development of the weaving intention in this phase is summarized in

Figure 42 through the following sequences: [1] given a design intention (polygon),
Ri defines the parameter for the weave's geometry (triangle). [2] Given the
weave's basic geometry, R2 defines the weaving style (point). [3] After the basic
geometry and the style have been defined, R3 fixes the weaving intention for the
next phase. Shape in step 4 and step 5 serves as the left-hand shape for the
weaving action in the next phase. The process of generating this shape is compiled
into rule R1-3 and R1-4.

R1-4
-

R1-3

R1-2

R1-1
-

R1 R2 R3 R4

1 2 3 4 5

Figure 44. the development of weaving Intention

Given a thread in the first step, the second and the third steps answer the question of 'what
polygon should I use' and 'how will the weaving be styled'. The fixed point in the fourth
step sets up the condition for the weaving action.

Phase 2: Developing Weaving Action

Phase two provide procedures to define rules for weaving action. In this phase,
weaving action refers to the addition of a new thread over the existing thread.
Rules to add the new threads, as the right-hand shape, over the initial thread, as
the left-hand shape, are discussed in the following:
98

Rule R 55 is parametric and corresponds with the design intention from

phase one. This rule adds a new thread over the initial thread with rotational
transformation (Figure 4543). The rotation point is point Pfrom rule R2, and the
rotation angle is alpha (a) from rule RI, i.e., a = 3600 / n. to modify the angle
between the initial thread and the new thread. Applying different values of a
would create different types of polygon, hence different weaving geometries (e.g.
hexagon, pentagon, octagon, etc.).

1-VT
R 5 <a=360 0 /n>
Figure 45. Additive Rule to define the right-hand shape for weaving Action

Because the right-hand shape matches the left-hand shape, rule R5 is

generative -- i.e., it can be repeated several times. With one point, rule R5
generates threads in cyclic weaving composition. To develop different weaving
compositions, such as linear or cluster composition, a rule to create new rotation
points is required, to propagate the weave in different directions.

Rule R6 transforms a new rotation point into different locations based on

the line's symmetrical properties (Figure 46). A line in 2D space has a symmetry
order of four, i.e. [1] by a 360* rotation [2] by 180* rotation [3] by reflecting on the
thread's lateral-axis and [4] by reflecting on the thread's longitudinal-axis. The
symmetry order of four indicates that new rotation points can be placed into four
locations. The parametric description in figure 46 encodes description for the
symmetrical transformation. S = symmetry number and [t(x)]= a set of number of
transformation based on the order of symmetry.

01 - 01 0j 01 01 010 oj 0:
R 6 <s,[t(x)]> R 6A <s,t(x),> R 6B <s,t(x)2 > R 6C <s,t(x)>
Figure 46. Shape's order of symmetry as the basis of transformation
R 6 generates the transformation axis based on the four orders of symmetry of the shape.
This axis then helps generate four locations for the new rotation points (blue points).
Variables indicates the symmetry number. Variable t(x)n indicates the transformation
used to generate the point based on the shape's symmetry.
 99

Lattice in figure 47 shows the computation of R4, R5 and R6 in adding and

erasing the point in the thread relationship. The rotation point position in the
thread relationship is critical in defining weaving action. In the first three steps,

after rule R5 adds a new line, rule R6 adds a new rotation point for the new line. In
the last three steps, Rule R4 erases the initial point, and again erases the new
point. Accordingly, the line relationship in step 6 has no points. Thus, R6
improves the generative aspect of rule R5 by adding a new rotation point, while R4

fixes the thread relationship in rule R5 operation by erasing the point. The
weaving style can be varied, but remains in the same basic geometry (e.g. hexagon
or pentagon).

The resulting shapes in the bottom of figure 47 express possible right-hand

shapes to generate weaving design for a certain basic geometry. Based on these
shapes, rules for weaving action are defined by pairing the thread from phase

one, as a left-hand shape, and thread relationships from figure 6 as the right-hand
shape (The rule computation is discussed later in page 106).

R5-5

R5-4 -- T
R5-3

R5-1 I-+

R1, R2, R3 * R5 1 R6 ' R4 R4

1 2 3 4 5 6

Figure 47. A new rotation point for the new shape

Based on the left-hand shape from phase one (in step 1 and 2), a series of shapes is
generated with additive rule R5, transformation rule R6 and erasing rule R4 (step 3 through
6). The new shapes serve as right-hand shapes. By pairing the left-hand shape in step 1 and
2 with right-hand shapes from step 3 through 6, series of additive rules to add new thread
are developed in the lattice. Rule R5-1 through R5-5 defines the weaving action for the
grammar, where R5-1 to R5-4 is deterministic and rule R5-5 is a non-deterministic.
100

Phase 3: Evaluating Weaving Condition

Phase three describes procedures to evaluate whether the designs generated from
the grammar are interwoven or not. It based on the way the weaver evaluates the
cross on the weave (see step 4 and 5 in figure 30). In particular, the interwoven
condition is evaluated through the over-and-under knot conditions on the weave.
Rules to evaluate the knot condition and rules to fix a knot in a non-woven
condition are provided in the following steps (see figure 48):

The first step provides a direction to evaluate the knot when the weave is
being developed or when the weave is finished. In developing a weave, rule R8E
visualizes an arrow onto an initial thread. In a developed weave, rule R8A
visualizes arrows on all threads to guide the knot evaluation within the weave in
order.

The second step evaluates each knot condition in the weave. By following
the direction of the arrows on the thread, one can trace and evaluate the over-and-
under condition of the weave in order. As the evaluation progresses, Rule R8B to
4 If the arrowed
rule R8G assigns values for a counting variable kto each knot.
thread is over another thread, then k is added by one: k -+ k+1 (as in rule R8F and
R8B). If another arrowed thread is under the other thread, then k value does not
change, k -> k, but the arrow direction is reversed (rule 8C).

The third step validates whether the weave is interwoven or non-woven

based on the counting from R8B-R8G. If the total k equals the number of polygon
sides (Xk=n), then the design is considered interwoven. If the total k is less than
the number of polygon sides ((Xk<n), then the design is non-woven.

If a non-woven condition exists, R8D-a and R8D-b could fix this condition
by switching the reversed-arrowed thread position to the over or under the other
thread. The thread's new position changes the k value, such that k -> k+1.

I Knot counting in this evaluation is derived from Alexander-Conway polynomial, see J. W. Alexander,
"ropological Invariants of Knots and Links," Tansactlons ofthe American Mathematical Society 30, no. 2
(1928): 275-306, doi:10.1090/S0002-9947-1928-1501429-1. and Sossinsky, Knots.
2 Mathematical Society American, Applications of Knot Theory: American Mathematical Society, Short
Course, January 4-5, 2008, San Diego, California (Providence, R.I: American Mathematical Society, 2009).
 101

Applying R8D recursively on the other reversed-arrow would increase the k value
and eventually equalize k and n. This equalization turns the non-woven condition
into an interwoven condition. However, R8D does not need to be applied to all
knots. Some arrowed threads may be kept under the other thread deliberately, to
create a weaving pattern.

The fourth step proposes what to do in a woven condition. The interwoven

condition signifies that the computation to develop the weave should stop
running. Otherwise, it will keep generating more shapes to overlap the existing
shapes. The computation of rule R5, for instance, will continue running as long as
there is a point inside the polygon. Unless the overlapping weave is part of the
design goal, previous rules need to be recalled to stop the computation, either by
erasing the point with R4, Xk=n: R4, or by moving the point into a different
position with R6, Yk=n: R6. The former stops the weaving action. The latter
generates new weaves in different directions (see the last iteration in figure 48).
102

k k+1
n-polygon k=9

-- I
R8A * - R8B -1

*
k k k

R8E R8C

k k+1 k k+I

R8F R8D-a --

-
Yk=n - 2 k+I k k+1

R8G --4 R8D-b -4

Knot counting in a Developed Weave

G +1 +2 +3 +4

R 8A R 8B R 8B R 88 R 8B R4
-i -i--- *J7=i7

4, f R4

+1 +1 +2 +3 +4

R 8A R 8B R 8C R 88 R 8B R 8D-b

4. nonwoven > woven

Knot counting in developing a weave

0 +1 +2 +3 +4

R 8E R 8F IR 8F R 8G R 8D-a R6

nonwoven > woven

Figure 48. Knot Counting

The series of rules at the top evaluates whether a weave is woven or not. Three iterations
show how the rules work in different contexts, i.e. in evaluating the developed weave
(middle), and in evaluating the weave as it developing (bottom). Example iterations in the
developed weave are provided with two cases: a case when the weave is woven and a case
where there is a cross that makes the weave non-woven. At the end of the iterations, two
rules follow up the woven condition differently. R4 stops the computation by erasing the
point, R6 continues the computation into the other direction by moving the point.
 103

Discussion
This section discusses the computation process of the set of rules defined in
previous phases. In particular, the discussion highlights the pivotal role of the
rotation point in determining and liberating the computational weaving process.
First, I discuss how the point position affects the weaving style. Second, I illustrate
how the point's presence or absence play key roles in the iteration process of the
grammar. 28

Point'sDeterminadon
The point's position directly affects the shape's relationship in the grammar and
therefore, may lead to a variety of styles in the language of weaving grammar.

To understand how the rotation point determines weaving styles, the point

location in rule R2 is constrained based on the underlying characteristic of the

shape's geometry. Three parametric rules in figure 49 represent the rotation point
interpolation within the triangle. R-2A places the rotation point along the triangle
reflection axis, R-2B places the rotation point along the hypotenuse lines, and R-
2C places the point in any location inside the triangle.

R2 R 2A R 2B R 2C

Figure 49. Rotation Point Interpolation

The rule-based design represents four 1D space options (blue lines) for the rotation point
interpolation within the triangle to understand the effect of each option in the rule
iteration process.

Different point interpolations in the set of rule R2 are used to define three action
rules in a set of rule R5: rule R5A, R5B, and R5C (Figure 5048). To illustrate the
point's location effect in weaving style, each of these rules is computed to

' The point, also known as a label, in Shape Grammar provides direction in rule iteration.
For more
example G. Stiny, "Ice-Ray: A Note on the Generation of Chinese Lattice Designs," Environmentand
PlanningB: PlanningandDesign4, no. 1 (1977): 89-98, doi:10.1068/b040089. And T W Knight,
"Transformations of the Meander Motif on Greek Geometric Pottery," Design Computing, no. 1 (1986): 29-
67.
104

generate new shapes. As shown in the iteration in figure 50, different designs
generated by rule R5s illustrate how the rotation point from rule R2s determines
the weaving style. R-5A produces a shape that resembles a Biaxial-weaving pattern

(plain weave). R-5B and R-5C produce a shape similar to a Reciprocal Frame,
where R-5C results in an extended intersection at the end of the thread. (I detail
this Reciprocal Frame later in this chapter.)

R 5A

IT 01
R 5B

0 0i 0

R 5C
Figure 50. Point's Determination
The different rotation point locations in each rule (R -5A to R-5C) determine the generated
design composition. The letters (A,B, and C) correspond to the rotation parameter in R2.

Point'sLiberation
The computation in Figure 5048 shows many possibilities for generating weaving
designs. However, the resulting design will always be in a cyclical composition

because it has only one rotation point that matches the initial shape and allows the
rule to run the computation. To expand the computation, rule R6 allows the
addition of a new point for such shapes.

Previously in Figure 4644, R 6 shows the thread represented as a line in 2D

space that has an order of symmetry of four. By using the shape symmetry as the

transformation basis, the rotations point is copied to its new location: [1] by 180'
rotation-R 6A, [2] by reflecting on the longitudinal-axis - R 6B, and [3] by reflecting
on the lateral-axis - R 6C.
 105

To see how different point impact the weaving design process, the set of
rule R6 is used to add a new point for shape relationship in rule 5B. Rules 7A, 7B
and 7C correspond to rule 5B in terms of shape-relationship, as well as to the set of
rule R6s in terms of new additional points. With the new point, the generative
performance of rule 5B is improved. The new set of rule R7 is able to generate
design in new directions, away from its original location (Figure 5149).

01 0f 0 +01 0 of 0o 01 0:
*0

*
R 6 R 6A R 6B R 6C

0o 0

R 7A

.-0.. R 7B
> 0 0 0
-

R 7C
Figure 51. Liberation of the Point
Four different weaving styles are generated from the same shape relationship in R 5A, yet
varied with different rotation point locations by R 6's.

The transformation of rule R6s seems to echo in rule R7's iteration. Figure
5149 reflects the characteristics of a certain symmetrical composition from R 6's.
When rule R 6A calls for a 1800 rotation, the new design generated using R 7A also
has a 1800 rotational-symmetry composition. When R 6B uses the thread's
longitudinal axis to reflect the point, the new design using R 7B also generates a
reflective-symmetry composition based on the same longitudinal axis. Similarly,
106

when R 6C calls for a lateral axis, the R 7C computation also generates a reflective-
symmetry design on the lateral axis. Thus, when a designer uses different
transformation in R 6, (s)he can anticipate these symmetrical characteristics in
the final designs.'2

Nevertheless, regardless of the different symmetrical transformations, all

designs in Figure 5149 express some consistency in terms of design language, as
they are generated by the same shape-relationship as in R 5B.

On picldng andaddingthe thread

The weaving intention, weaving action, and weaving evaluation have been defined
into a set of rules in this grammar. Within this series of rules, the decision
process, as it relates to how a particular shape and rules are chosen and
computed, needs to be based on certain understandings of the characteristics of
the rules.

In the following, I discuss some considerations for a particular moment

when a thread is about to be picked and added onto the weave, considerations
such as 'Which part of the thread should I begin with?' and'Where to add a new
thread? In discussing these considerations, I illustrate the way the left-hand shape
and/or right-hand shape of the rule characterize/s the weaving action in this
moment.

On which partofthe threadshouldlbegin?

The rule begins by matching the initial condition with the left-hand shape of the
rule. Thus, the left-hand shape determines which part of thread to begin with. In
this grammar, the left-hand shape is divided into two types: a line with a point and
without a point.

A line with a point specifies a particular part of the thread to be computed.

The ratio between the line length (1) and the point's distance to the mid-point of
the line (h) constrains the matching process (Figure 5250). As a mental shape,

m For more example on how shape's symmetry determine the rule iteration, see T. W. Knight,
"Constructive Symmetry," En vironment and PlanningB: Planningand Design 22, no. 4 (1995): 419-50,
doi:10.1068/b220419.
 107

points and lines can be embedded onto the thread in different ways. However, the
embedding must follow the ratio. Different embedding may vary the size of the
polygon, e.g., the shorter the point's distance to the thread, the smaller the
polygon, and vice versa. Yet, within this size variation, the line-point ratio ensures
that the style is consistent.

h I ]
Sh h

Figure 52. Ratio in point and line relationship in embedding shape on the thread

A line without point suggests that any part of the thread can be computed.
Without a point, there is no ratio to follow. The rules in Figure 5351 show a left-
hand shape without the rotation point as an initial shape. Any matching line found
during the computation process, whether it is a whole thread or part of a thread,
can run this rule. As a result, as shown in figure 12, the weave may have one
particular style, or another.'"

RI-5 R5-5 RU-5 R5-5

.
-
R5
R 7-5 7 - R5- R-5
r -5.V

Figure 53. What shape to compute?

By embedding the initial shape in a different way, the generated designs could be free of a
stylistic constraints.

"'For more examples on how a shape rule and a shape relationship are generated under non-deterministic
approach to create different designs, see T. Weissman Knight, "Languages of Designs: From Known to
New," Environment andPlanningB: PlanningandDesign 8, no. 2 (1981): 213-38, doi:10.1068/b080213.
108

Where to adda new thread?

After the left-hand shape finds its matches, a rule will substitute the matched
shape in the iteration with the right-hand shape of the rule. This substitution
indicates where the new shape should be located.

This Weaving Grammar contains rules where some of its right-hand

shapes have or do not have a point. Below, I list the interplay between the
presence and the absence of a point in this grammar based on the rule from
Figure 4745.

A rule where the left-hand shape and the right-

R5-1
hand shape have the same point location suggests
*

R5-2
'V+ locating the new thread in a cyclical fashion.

A rule where two points exist in the right-hand

shape and one point in left-hand-shape, in the
original and new location, provides options to add
the thread in two locations.

R5-3

I A rule where the original point is moved, instead

of copied, to the new location, signifies an action
*

to add the new thread only in the new location.

A rule where the left-hand shape has a point but the

R5-4
'I right-hand shape does not have a point anticipates
an action to end the computation process at the
next step.

A rule where neither the left-hand shape nor the

R5-5

I-+4
right-hand shape has a point liberates the user to
place the new thread in any place where the shape
matches the rule.

Finally, with sets of rules in the grammar, a specific weaving style can be
determined with rule sequences (Figure 5553). For instance, with two rules, A and
B, the rule iteration can be controlled, for instance, A-A-A-A or A-B-A-B, to develop
 109

a certain style. Figure 5654 shows how these rules could help analyze existing
interwoven structure from the past, in particular, Reciprocal Frame.13 1

I R5 -3 R5-3 R5-4

I I
-

R5 -3 R R5-4
-

Figure 54. Where to add a new shape?

New shapes are added in different ways with a deterministic rule.

R5- 3 RC R5-3
... R5-3=R6C=*R5-3=*R6C:R5-3=>R6C=>...

R5-4 RS-1 R5-4

.

TOF
...R5-4=>R - 1=>R5 -4=:>R5- 1=>R5 -4=>R5-1I=>...

Figure 55. Rule Rhythms

Different rule sequences create specific weaving styles

1 More detailed study of Reciprocal Frame, which inspired most of the examples of this basic weaving
grammar exercise, can be seen in: Olga Popovic Larsen, ReciprocalFrameArchitecture, 1st ed.
(Architectural Press, 2008).
110

960 1289 1519 1519

Rainbow Bridge De Honnecourt's Leonardo Da Vinci's leonardo da vinci's

floor frame rectilinear grid triangular grid

|IR7C *T* SR5AT 61 R7A f 28*

I-* *~ R7AT*9* * -- I

R557*

1545= 1670
-. 4 L

Sebastiano Serlio's John Wallis's Zollinger lamella Bunraku theater

floor frame ceiling frame frame roof frame

*j R5A q. * R7C].

26*
.RA> R5C
.R5CT
-+*go.
2005

Seiwa theater Balmond-Ito wooden Apple Store

ceiling frame frame Glass Frame

. R SR A
.

2
Figure 56. Analysis of Existing Design

2 Photos on top of each generated designs are from: Ibid. and Nexus Journal, Vol. 10, Issue 1, April 2008.

www.nexusjournal.com/volume-10/number-l-april-2008.html
 111

Case Studies
This section illustrates Basic Weaving Grammar contextualization for
architectural weaving in traditional context, through a series of case studies. In
contextualizing the grammar, the case studies seek to investigate the process of
building a tradition rather than simulating a traditional building as a product - in
particular, a tradition that emerged from the assimilation between computational
and traditional method.

Several considerations in contextualizing the grammar for this

assimilation are examined. First, the weaving grammar needs to be easily
comprehended. Second, the design must meet certain structural and fabrication
requirements. Third, the rules for design and construction must be affordable
with respect to the local skill and technology. To examine these considerations, I
conducted three types of case studies.

Weaving Grammar Workshop: This workshop examined Weaving

Grammar's degree of comprehension. The workshop was conducted in
two places: Toraja, Indonesia and Cambridge, USA. In Toraja, I taught
Weaving Grammar to a group of 5 craftspeople with visual media (pen and
paper). In Cambridge, I taught Weaving Grammar to a group of
elementary students with a physical model: a small wooden-beam (This
wooden beam serves as an example of architectural weaving component.)

Structural Simulation: This simulation analyses and optimizes the

structural behavior and soundness of the architectural weaving
component. The simulation uses a digital model with Finite Element
Analysis software and a physical model that undergoes a compressive test.

Mock-Up Installation: The installation process assesses how the weaving

grammar and the fabrication process interact during the construction
process. Particularly, I assess the way in which the grammar's rules assist
the builder in assembling the physical component into architectural
design. For this installation, I taught three local builders to build a mock-
up installation. The installation was built using timber from the local site.
112

Study.LA: Weaving GrammarWorkshopI - Measuringthe Weaving Grammar

Comprehension
The participants in the workshop included five local engravers in Kete Kesu,
Toraja. To these participants, I first explained the background of the workshop,
which was based on my observations of their colleague (the weaver whom I had
observed previously). I showed them an example of how to compute the rule in
the grammar using tracing paper and pen. The rule used in this workshop was R
7B, with an option to change the angle (a) of the shape relationship in rule R 5.

One of my concerns about this workshop at the beginning was that

embedding and transforming the shapes using the symmetry principle would be
unfamiliar for the participants. However, in the practice session, the participants
were less confused about this and more troubled by how to use the tracing paper.
Tracing paper is not a local commodity and thus they were not used to using it.
After I showed them the benefits of the paper's translucence --. that the see-
through quality was its key feature, and that they could explore embedding and
transforming shapes with it -- their troubles were resolved.

Once the participants understand how to use the media, they quickly
grasped how to compute the rule properly and could generate designs from rule
R5. After executing the rule for a while, the participants began to improve on
ways to use the media. For instance, while I usually match the shape during the
iteration process by moving the tracing paper with my hands (to align the overlaid
shape), the craftsmen embedded the imaginary rotation axis into the pen, as a
hinge to rotate the paper.

I found later that the Euclidian transformation and symmetry principles

are not new to them (although they may use it with a different term). The
engravers seemed to have been using the symmetry principles in their ornament-
engraving activities. Although their ornamentation rule is not visually
documented, the fact that most of their ornament can be generated with rotation
and symmetry infers that they may use the Euclidian transformation tacitly.'3

" For more detail example on the geometrical principle of the local ornament and the rule interpretation of
the ornamentation method, see Rizal Muslimin, "Decoding Passura' - Representing the Indigenous Visual
 113

Thus, for them, this workshop was essentially a continuation of their crafting
activities routine, but with different shape. At the end of this workshop, the result
showed that the local craftspeople can easily acquire an understanding of the

visual weaving rule and generate certain designs from the grammar.

Figure 57. 2D Computational Weaving Workshop

Local craftspeople generate design from the weaving grammar

Study 1B: Weaving Grammar Workshop 2 - Setting Up the PlaygroundRules

The previous study demonstrated that craftspeople could generate designs from
Weaving Grammar with tracing paper and pen. In this study, I conducted a
grammar workshop with groups of 90 middle-school students to examine how the
students could comprehend the weaving grammar with the 3D shape.'3 The
study uses R 7A, R 7B and R7C.

Messages Underlying Traditional Icons with Descriptive Grammar," in Open Systems (presented at the the
18th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA 2013),
Singapore: Southeast University Press Pub. DateW:2007-10-01, 2013), 781-90.
134 As part of the MIT Science, Technology, Engineering, and Mathematics (STEM) Summer program.
114

In the workshop, a wooden beam was used to represent the 3D physical

translation of a 2D line. The weaving grammar adjustment from paper to wooden
beam is described below in terms of point reference and matching processes.

Pointin 2D - Pointin 3D
While the rotation point helped the participant in study 1A compute the rule in the
previous study, in this 3D case, it would not be effective to ask the school children
to imagine the rotation point in the space between the wooden beams. For the
point to be effectively used, I projected the rotation point orthogonally onto the
side of the beam. As an oblong, the beam has an order of symmetry of eight. Yet,
Rule R6 copied the point onto the other location of the beam using the order of
symmetry 4 (Figure 5856). In other words, the beam's order of symmetry of eight
was reduced by four. By this reduction, the rule only allows four ways to
transform the beam in the 3D space.

*1

-4

Figure 58. 2D shape is translated into physical 3D shape.

Matchingin 2D - Matchingin 3D
In study 1A, superimposing 2D shapes with the tracing paper accomplished the
matching process. With a wooden beam, however, this process is impractical. So,
to bypass the 2D matching process, the position of the two beams was fixed by
 115

interlocking joinery. Each beam has four notches to allow it to be interlocked

with the other beam via one of the notches. By interlocking one beam with
another, the position of two beams is guaranteed to match the shape-relationship
in the rule.

With the rule's parameter embodied firmly into the shape through points
and notches, the rule for the wooden beam is no longer parametric. The user can
no longer modify the angle parameter (a) and the rotation point as shown in R5
(Figure 4543). Nevertheless, the beam has an order of symmetry of four, which
means, even without the parametric variable, two beams can still be assembled in
four spatial relationships. These options are important to explaining the
generative aspect of the rule in the workshop session.

Workshop Sessions
The workshop was divided into two 1-hour sessions for students in the 5th through
9th grades. For the first session, sets of wooden beams (1 cm x 6 cm x 0.3 cm) with
embedded points were distributed to the students. Each student had one set of
twenty-four wooden beams. The students were presented with visual instruction
(on an overhead projector screen) showing rules 7A and 7B. The students were
asked to assemble a shape based on the two rules shown in Figure 5957. No other
instructions, such as step-by-step images, video or demonstration, were given; the
only instructions were the images of the two rules. In this way I could assess how
the visual rule and the beam, with minimal instruction, helped the students create
the desired designs.

After about 20-30 minutes, most of the students were able to follow the
rules to generate the desired design (Figure 5957). The fixed embodiment of the
rule's parameter in the beam helped the students to compute the shape in two
ways: First, the notches constrained the shape-relationship, thus simplifying the
matching process. Second, the point position on the beam helped the students to
navigate where to place the beam according to rule R7 (Figure 5957).

For the second workshop session, another set of wooden beams, with no
points, was distributed to the students. Without points, the rules become non-
116

deterministic and the students could choose their own ways to assemble the
beams. The students were given three non-deterministic rules (on an overhead
projector screen) representing three different spatial-relationships, as shown in
Figure 6058.

Rule #1 Rule #2

Figure 59. 3D Weaving Workshop -Session 1: Weaving with rule

5th - 9th grade students were able to generate designs following the deterministic rules.

r'
K> -~

Figure 60. 3D Weaving Workshop - Session 2: The No-Rule Session

Students generated 'free' designs based on the non-deterministic rule.
 117

During the second session, the 5th - 9th graders seemed enthusiastic
about developing their own design with the shapes (Figure 6058). In their playful
designs, the students tended to use the generated design pattern as an internal
structure of certain objects, such as an airplane or robot. This tendency implies
that some students understood the inherent structural feature of this
configuration. Other students created rather abstract shapes that did not seem to
aim for a familiar object, but seemed to be an exploration of how the shape could
be assembled without rules. This exploration indicates that the students had
found new emergent shapes and new potential for the shapes.

Study2: StructuralSimuladon - Evaluatinga Design's Structurallntegrity

The two previous case studies illustrated the grammar's ease of comprehension.
Both adult craftspeople and middle-school students had no significant difficulty
understanding the Basic Weaving Grammar. These case studies indicate that
Weaving Grammar's methods of designing and developing architectural weaving
computationally are accessible to different members of society.

To ensure that the designs are not only comprehensive, but also applicable
on an architectural scale, this study tests designs from the weaving grammar for
their ability to withstand structural loads. Tests are carried out using two
optimization strategies. [1] Parametric Optimization for the beam as architectural
component. [2] Grammatical Optimization for an assemblage of beams as an
architectural design composition. Parametric optimization tests the design's
structural performance by changing the beam's dimension, specifically, the
positions of the notches in the beam. Grammatical optimization tests the design's
compositional performance by changing the rule within the same grammar.

ParametricOptimizationfor the Beam

The first test seeks to find the optimal distances between the notches, which
would be loaded by other beams, to meet both structural and fabrication
requirements. In terms of structure, the notches' distance should have a relatively
small displacement and small stress. In terms of fabrication, the notches'
distances should provide enough space to allow a builder to assemble and
maneuver the beam easily.
118

LIF

Figure 61. Idealized model for structural simulation

The Principle of Equilibrium signifies that the distance between

supporting-notches and supported-notches (L2) is linearly proportional to the
external force, but inversely proportional to the external forces (Figure 6159).
Thus, larger L2 means lower stiffness but more space for assembly, while smaller
L2 means more stiffness but less space for assembly.

To find the optimal distribution between these contradictive conditions, I

tested five idealized models under different load distributions, to examine the
displacement patterns using Finite Element Software (ADINA, Inc.). The left
column in Figure 6260 shows the result of the Finite Element Analysis simulating
the shape-displacement of five different notch distances given the external loads
(Figure 6260). Model 2 from the simulation shows a smaller displacement and the
lowest maximal stress.

Figure 6361 shows the how notch distance affects the design when the
beam is assembled into on a larger weaving pattern. Models 1, 2 and 5 were
generated using rule R7A into an interwoven pattern. The resulting pattern was
then compared to assess how the optimized load distribution and displacement
are related to the pattern aesthetic on a larger scale. In Figure 6361, the
comparison of three models shows pattern A, which has the widest notch
distance, and pattern C, which has the narrowest notch distance, are not
recommended for structural reasons. Pattern B, which has the lowest maximum
stress and displacement, meets both preferred structural stiffness and fabrication
requirements.
 119

AU Displacement a Stress Distribution

1 2.8 3.70E+10

2 1.7 4.60E+09

3 3.8 3.10E+10

4 4.7 1.26E+10

5 4.9 1.50E+10

Figure 62. Displacement and Stress Distribution from five notch locations
Maximum displacement and maximal stress tend to increase as the distance between the
notches narrows except model 2 (second from the top).

'L 0-~55

Figure 63. The diagram shows the effect of load distribution on the design pattern
generated with one rule: R7A.
120

GrammaticalOpdmization for a Design

Given the optimal notches for individual beams from the parametric optimization,
the next test aims to find the optimal beam configurations based on the design
generated by the grammar. Three models using the same beam (model 2) were
generated from the grammar. The first type, developed with R 7B, is named the
'da Vinci grid', the second type, developed with R 7C, is named the 'Wallis grid',
and the third type, developed with R 5B and R 5D, is the 'Nandong' grid. All three
were assembled with 24 pieces of Medium Density Fiberboard (MDF) and all
weigh 79 grams (Figure 6462). The assembled beams were compared using the
universal test machine to examine how much external force each design's
structural configuration could withstand. In particular, the design structure's
stiffness was measured by the deflection of the structure under a specific load.
These structures were tested in two orientations, vertical (a wall) and horizontal (a
floor). Each of these grids was then loaded with 40-newton pressure force,

- L L---------J

I - -R2 P34PA

"da vinci grid" "wallis grid" "nandong grid"

Figure 64. Three models of interwoven beams-of equal volume and weight are
generated from three rules in the grammar.
 121

Figure 65. Optimization Test and Weighting (vertical and horizontal)

From the simulation results, in the horizontal orientation, the da Vinci

grid has the lowest displacement at about 7 mm, followed by the Wallis grid (11
mm) and Nandong grid (12 mm). When placed in a vertical orientation, the da
Vinci grid again has the lowest displacement with a 5.6 mm displacement,
followed by the Nandong and Wallis grids, whose levels of displacement were
close (Nandong 8.7 mm, Wallis 9 mm)(Figure 6664).

In Figure 6664, the jagged line on the Wallis graph (representing the floor)
and the da Vinci graph (representing the wall) indicates that some of the beams
had cracked. As the compressive load rose, the number of cracks increased,
which means the number of solid beams structurally, supporting the increasing
load also decreased. Yet, because the grid interlocked using independent beams,
the crack of one beam did not spread immediately to the other beams; it remained
isolated for a certain period at least. At the end of the test, none of the models had

collapsed after the pressure force reached 40 Newton for both the wall and the
floor.
122

Load (Newton) da vinci wallis nandong

grid grid grid

45

40

35

30
--- Pattem i
25

20/

15 Paem

10

5
50 Deflection
Wwalli(
da Vinci

da vinci Wallis nandong

Load (Newton) grid grid grid

45

40

35

30
-Wan 1
25

20

15 Wan 3

10

5
Deflection
0

Figure 66. Results from Compressive Test

The top graph shows vertical orientation (wall) loading while bottom graph shows
horizontal orientation (floor). The da Vinci grid has the lowest deflection compared to the
other designs.
 123

Study 3: Woven-TimberInstailadon- Envisioningthe Grammaron the Vernacular

Context
The previous studies showed that physical models of designs from the grammar
were structurally sound and that the grammar was easily comprehended. In this
study, to inspect how the shape and the rules could be applied in an actual
construction project, I built a mock-up installation in Toraja. The installation
involved the work of three carpenters from the local village.1 37

The 2m x 3m x 2m installation took the form of a gabled house using the da

Vinci grid.'3 It was assembled near the entrance to the village using timber from
local material. The 10 cm x 60 cm x 3 cm beam were manually fabricated by the
local carpenter with conventional woodworking tools (saw, chisel, hammer and
wood-shaving tool) based on my design. In total, 135 timber beams were
fabricated for this installation.

In this shelter mockup, the beams' assembly did not require additional
binding components such as nails or screws and bolts, because, with four notches,
the wood serves both as an element and a joint (body-joint system). In a way, the
interlocking mechanism resembles the traditional weaving principles.

The positions of the notches on the wooden plank provided information to

the craftsperson as to how the piece of wood should be joined. At first, a problem
arose during the installation, involving adding the last piece of wood to make a
closed-loop knot (the fourth beam). The first three of the four beams could be
assembled as most wooden structures are traditionally built, by stacking one over
the other. This was achieved by placing all the notches in one direction (all up or
all down). However, the fourth beam in this case needed to be woven in so that
the shape-relationship between the beams would extend both overand under
(This is why the notch positions face both up and down.) In weaving, a new thread
is added by sliding the thread in between the knotted threads. To adopt this

" At the project's start, more craftspeople were ready to help, but we found that the installation did not
require that many people.
' In this case the installation was purely experimental; the aim of it was to enable my observation of the
assembly process and not for shelter purposes
1 Note that the beams' fabrication and assembly took place in different areas. This section focuses only on
the assembly part.
124

weaving gesture, the notches are chamfered and require a rotation movement as
the assembler slides one piece of wood over another. After I showed the
carpenters how to weave the beam together using the chamfered notches, the
carpenters readily understood the method and could easily assemble the timber
slats.

The 2D weaving rules helped the carpenter to weave the beam with
minimal instruction and no construction drawing. In addition to the rules, the
carpenters used a scaled model of the shelter that I made. The scaled model was
assembled with the same rules and methods (Figure 6765). The model directly
guided the carpenters not only in visualizing the final design, but also in
understanding the way the beams were to be assembled. In other words, the
fabrication method was embedded in the model just as the weaving method was
implicit in the interwoven crafts.

The three carpenters started the installation around 10:30 am, and
finished at 5:30 pm. I had not expected the installation to be assembled in just one
day, with only three people; it usually takes a few days to assemble 135 pieces of
wood (Figure 6866). Later, one of the local craftsmen explained to me that their
~500-year-old traditional houses were also assembled with wooden joints that do
not require nails. The wooden slats of these houses are notched, thus the body-
joint system was familiar to them. Yet the beams in the traditional houses are
stacked on top of the other, not woven. The difference between the traditional
houses and this installation is that, in the case of the installation, the timbers were
woven with principles inspired by the local weavers. Thus, somehow, this
installation married their ancestors' construction techniques with their
colleagues' crafting techniques.
 125

Figure 67. A carpenter holds the scaled model during the fabrication process to
verify where to place the timber.

Figure 68. The Mock Up Installation

126

3.3 Abstract Weaving Grammar:On Extending the Language of Weaving

Design

Introduction
This section explains the process of reinventing weaving with abstract weaving
grammar-a set of rules that can generate new designs by reconfiguring the
shape's mechanical description and the shape's boundary relationship. Inspired
by the traditional weaver's visual and haptic experience, my aim is to provide
computational design methods that can incorporate the aesthetic-structural
integrity of weaving into architectural design.

In developing abstract weaving grammar, I adapt some aspects from

traditional weaving by aligning the way in which traditional weavers sense the
weave with computational mechanics and computational design theory. In this
alignment, I include approaches from the finite element method (FEM) into shape
grammar's visual computation method to better integrate the shape's mechanical
consequences with the shape's visual significances. By relying more on the
abstraction of weaving mechanics properties and less on the common conception
of typical weaving material, the design that results from this grammar is hoped to
extend the language of weaving design.

FEM and Shape Grammar

FEM is an engineering method used to analyze the physical behavior of an object
that has been idealized into a mathematical model-an assemblage of vertices,
edges, and faces that define the object's shape. In this model, the object is defined
by the mechanical properties of its element, such as profile area (A), length
between vertices (L), boundary conditions (U), the material's degree of elasticity
(E), and the applied external load (R).

Given a shape's mechanical properties, FEM decodes the mechanical

behavior of the shape by satisfying the equilibrium equation (based on Hooke's
Law) to solve the unknown point of displacement (dU). The internal stiffness of a
bar, defined by the variables E, A, L, and dU, should be equal to the external force
on each element of the shape (R), as follows:
 127

E A
= R
-dU

To incorporate FEM principles into the computational design process,

shape calculation in FEM and shape grammar must be represented in the same
visual language.

In the FiniteElement Methodfor EngineeringAnalysis, Bathe describes

the process of idealizing physical objects into mathematical models in four
phases: geometry, material, loading, and boundary condition.'"

The geometry phase represents the kinematic shape of a physical object

(e.g., Is the shape a bar, beam, plane stress, or 3D object?).

The material phase assigns material properties to the kinematic shape

(e.g., Is the material isotropic, anisotropic, elastic, or rubber?).

The loading phase applies an external load onto the object (e.g., Is the load
concentrated in one point, distributed evenly, linearly, or centrifugally?)

The boundary condition defines the degree of freedom of the shape (e.g.,
Can the shape move or rotate around a certain axis)?

To better understand how this idealization process works in the context of

weaving, in the following, a knot of a weave is computationally represented and
scrutinized via three approaches. First, I revisit Kawabata's model in idealizing the
42 43
knot into a finite-element model through shape grammar rule notation.1 ,1
Second, I compute a knot's shape with boundary rules in shape grammar, without
any mechanical properties. Third, I alternately calculate the knot with the FEM
idealization rules and shape grammar boundary rules.

14 Bathe, FiniteElementMethod for EngineeringAnalysis (1996), 165

42 S. Kawabata, Masako Niwa, and H. Kawai, "3-the Finite-Deformation Theory of Plain-weave Fabrics
Part I: The Biaxial-Deformation Theory," Journalofthe Texdle Insdtute 64, no. 1 (1973): 21-46,
doi:10.1080/00405007308630416.
"' Kawabata's knot schema of a bi-axial weave laid the foundation of basic research of textile mechanics in
the clothing industry and has been subsequently studied and refined to include additional properties of the
knot such as the spring-moment at the friction between two yarns in the Kevlar model and shear-lock in
analyzing the knot's rigidity under shear stress in the King's model.
128

Knot in FEM-Kawabata's Schema Revisited

In this section, Kawabata's schema of knot of biaxial weave is revisited in the
order of the FEM idealization process described before using shape grammar rule
notation (Figure 6967).

Geometry
Rule G1 generates the parametric schema of a yarn. In idealizing the knot's
kinematic geometry, Kawabata perceives the physical yarn by embedding one-
dimensional weight in the knots-the structuralneutralline,(W13). Using this
neutral line, the 3-D yarn is represented as a 1-D bar in 3-D space (U 13). Variable
(L) represents the yarn segment's length and variable (A) represents the yarn's
profile area.

Material
Rule Ml assigns the material property to the bar to describe its material stiffness,
such as the material degree of elasticity (Young's Modulus-E). The assigned bar is
represented with a color that inhabits the ID bar (U,3 + W11). One can perceive
different material properties of the bar by assessing different colors; for example,
cyan means it is elastic, magenta means it is rigid. The color of the material can be
further parameterized to represent a wide range of material options.

Loading
Rule Li adds external force (F) to the endpoints of the bar, and L2 transforms the
point based on the force's magnitude. The force is represented by arrow (V,2), to
indicate direction in which the load is applied.'","I While the load in 3-D space
can be oriented in many directions, in the equilibrium equation, external force is
calculated independently as a 1-D vector in 1-D space (U 1 ). For instance, in case
there is an angular direction of the force direction, the force will be projected onto
the y-axis and the x-axis.

" This model ignores the bending moment (M.) from a yarn segment's deformation in plane xz, and
cross-over moment (My) from a shear deformation in plane xy. For more detail about this moment, see
Kings model.
1'" A label in shape grammar is commonly used to reduce the shape's symmetry (Knight), to show where to
add a new shape (Knight) and to attribute identity to a shape if some shapes look the same (Stiny).
 129

BoundaryCondidon
Rule Bi defines the boundary condition of the bar. The boundary condition refers
to the shape's degree of freedom (DOF) in responding to the external forces (in
this case, the external forces applied to the points at the end of each bar (Uoj)). The
arrow in the right-hand shape indicates the degree of freedom.

L L
geometry 61: -- + 0-o see the axis SG:( 7) -+(

A A

material Ml: -- E see the line SL: --

-
M2: - -+ -

o- F see the boundary SB: 0 -+ 0

loading LI: 0 -+

L2: o- F -+ -- oF see the boundary S81: - o"---o

boundary B1: o U. u inverse boundary SB2:o 0 -+

-
Figure 69. Rules for FEM idealization in shape grammar.

In addition to the mechanical rules, the grammar is also accompanied by

the set of seeing rulesto allow users to visually identify part of the idealized knots
that would be applied by the mechanical rules. Such rules are Rule G, to see the
neutral axis of a thread, Rule M to see the line, and Rule B to see the boundary of a
line (point)(marked by the orange shape in Figure 6967. Together, the seeing and
mechanical rules work recursively in the computation, where the seeing rules
highlight the user's visual attention on part of the knot, and mechanical rules
apply a mechanical description on the highlighted parts.

Figure 7068 shows the iteration of generating Kawabata's schema of a

plain weave's knot using the above rules. The iteration to the far right identifies
shapes that are visually attended, and the iteration in the middle shows the
embedded mechanical properties of the visually attended shapes. By applying the
rules recursively, four original Kawabata's schemas on the left images are
generated.
130

Kawabata Mechanical Seeing

Model Embedding

U.LII WVMV

1 SG

.................................................

I Y./2
x. Gi
P
P
L./2 see axis

.......... 0,

SL

The unit structure define truss-schema

Y./2.

L./2 M1
P Q h R see lines

X. SB

assign material

F,, F
p F Li
see boundaries
F

F, F,
SB
apply loading
Notation for the
forces and deformation

X,
F F. BI
IA: P F F see boundaries
F

FF

6
Fe,0 F,
The unit structure define boundary
in the deformed state conditions

Figure 70. Kawabata's Schema Reiteration 1

Rule derivation to revisit Kawabata's model of the knot, as a unit cell geometry of woven
fabrics (1973).

' Plain weave image and the original unit structure model are from: Kawabata, Niwa, and Kawai, "3-the
Finite-Deformation Theory of Plain-Weave Fabrics Part I."
 131

In this idealized mathematical model, the knot's optimization process can

be performed by modulating variables in the equilibrium equation to satisfy the
virtual equilibrium principles (Figure 7169). For instance, the material degree of
elasticity (E) and the yarn's cross-sectional areas (A) are proportional to the
external load applied to the knot (R). Thus, by increasing variable (E) (e.g., using a
stiffer material for the yarn) and/or increasing variable (A) (e.g., increasing the
yarn strand's dimension or density), we can increase the knot's rigidity.
Additionally, because variable (R) is inversely proportional to the yarn's length
(L), then decreasing variable (L) can also strengthen the knot's rigidity (e.g., by
49
adding smaller knots within the same knot module dimension (Y)).1 ,1

Given Change length Change thread Change thread

External Load between knots Material stiffness profile-dimension

EAdU =F EAdU =F EAdU =F dU F

L L L

Figure 71. Variables for optimization of the knot's rigidity in the Kawabata's
schema.

Knot in Shape Grammar-Boundary Rules

In revisiting the Kawabata's schema, I separated the shape of the knot from its
mechanical description with seeing rules. By this separation, the knot can be
calculated in shape grammar purely as a shape. In the following, I explain how
shape grammar visually calculates the shape of the knot with its boundary rules.

The boundary and shape in shape grammar is not tightly bound as in the
Kawabata model. Applying a boundary rule, x -+ b(x), to a line will produce the
line's endpoints, but applying the inverse-boundary rule, x -+ ba(x), to the points
will not only retrieve the previous line. Because the endpoints are independent
from the line, the results of applying inverse-boundary rules on the point are not

"' Kawabata solved this equation by offering two conditions: One in which the yarn is compressible and
another in which the yarn is uncompressible.
"" For instance, Kawabata uses this schema to analyze other weave types, such as the tri-axial weaves S.
Kawabata, Mari Inoue, and Masako Niwa, "Non-Linear Theory of the Biaxial Deformation of a Triaxial-
Weave Fabric,"Journalofthe Textle Insdtute 83, no. 1 (1992): 104-19, doi:10.1080/00405009208631180.
 132

limited to the previous line. It can be any shape that has the points as its
boundary. Similarly, in the knot, the result of the inverse-boundary computation
on the yarn's endpoints is not limited to the previous overlapping yarns. It can
also retrieve other lines where points coincide with the other line's boundaries.

Figure 7270 shows boundary-rules computation on a knot, where the knot

is viewed from the top. With the boundary rule, shape grammar computation
could generate various shapes on the knot using the yarn's original endpoints. The
three iterations in the figure show the boundary and inverse-boundary rules
generate new knots. Of the three, the first two steps from the rule, x -> x + b(x),
produce similar results-four points on the knots. For the third and fourth steps,
however, the inverse boundary rule, x -> b1 (x), begins to show different shape
compositions as a result of applying the rule only to specific points. For instance,
the boundary of the four points in step 2 could be four lines (a square) or two
triangles in step 3. Similarly, the inverse boundary of the lines need not be applied
to all four boundary lines in the square, as the original yarns can also be
considered as lines. As shown in the third row of the third to fourth step, the
inverse rule is applied to two lines and the original yarn (forming a triangle) to
retrieve the triangular plane (gray). The fifth step of the iterations shows a new
shape where the original knot typology has been fully altered, as some lines have
disappeared into endpoints as a result of applying the boundary rule to the lines.
1 2 3 4 5

x -+x + b (x) x -b(x) +x -+ b"(x) x -4 b (x)

+

0>

x -+ x + b(x) x- b"(x) x -+ b-(x) x -+ b(x)

+bx
x+b x
x -+x +b(x)
ON I
x -'(x) + x -+ b*'(x) +x -+ b(x)

Figure 72. Boundary rule iterations.

 133

Knot in FEM-Shape Grammar-Perceptual Optimization

The knot's calculation in FEM and in shape grammar shows a potential symbiosis
in the way in which a knot is seen as a shape with embedded mechanical
descriptions.

This symbiosis paves new corridors through which the way a traditional
weaver and engineer optimize a knot's structure can complement each other. In
particular, in the way an engineer and weaver consider the available options in
optimizing the knot (Figure 7371). In conventional building engineering practice,
the external force (R) is mainly considered as a given condition because the design
configuration is considered to be finished. To achieve a certain degree of the
knot's stiffness, the option is to modify the yarn's internal mechanical property
(i.e., variables E, A, and L). For the weaver, the bamboo's material property (E) is
considered as a given condition. To meet his design goal, the option is to modify
the thread's shape and the external force on the knot (i.e., variables A, L, and R).

Internal Optimization External Optimization

Figure 73. Different focus in in optimization between internal optimization and

external optimization.

Representing these two approaches in the equilibrium equation highlights

an intersection where some variables overlap while others are independent. The
overlapping variables, A and L, indicate that both the engineer and weaver are
keen to modify the shape of the knots. On the other hand, the independent
variables, E and R, underline the strong feature of each approach. The engineer
can modify material properties (E) but not the external force (R). The weaver does
not change threads' material properties (E) but can modify external force on the
weave (R).

By combining the features from the two grammars, there is an opportunity

to achieve the weave's structure and aesthetic integrity in architectural design by
134

modulating the thread's material properties (E) using the mechanical rules and by
modifying the external force 's shape (R) with boundary rules.

The following strategies: loading, boundary, and material ambiguities, set

mechanical rules and boundary rules by incorporating the traditional weaver's
haptic experience in the previous chapter.

Strategy#1 - LoadingAmbikuity
In the previous chapter, I described the haptic ambiguity that occurs at the
moment after the material "feeds" information to the weaver's perceptual system
and before the weaver decides what action to take with that information. At this
moment in time, the weaver has several options as to which direction and how
much muscular force he or she needs to apply to the thread (e.g., how much to
pull, push, or pinch).

The strategy proposed here is to formulate the weaver's haptic ambiguity

within the grammar. Represented with 6 nodes and 3 axes, the knot has 18 degrees
of freedom. This means that in each node, the external force can be applied in
many directions, for example, up (lifted), down (pressed), front (pushed), back
(pulled), left or right (tilted). The condition of having these different options for
which to apply external load is analogous to the moment when a weaver considers
which way to move the thread.

Figure 7573 shows four different ways rule Li applies external load to the
endpoint of the knot. Rule L2 shows the mechanical consequences of different
loading conditions by moving the knot as a rigid body (e.g., fall down, moving
apart, shifted horizontally). 5 ' As a consequence of applying rule L2, the shape-
relationship of the knot in the right column has now changed. In other words, the
knot design has been mechanically derived into four new spatial relationships.

1 As a rigid body, the thread moves to new position as a whole with a Euclidian transformation without
deformation.
 135

Strategy#2 - BoundaryAmbiguity
The traditional weaver responded to the material's reaction in different ways,
from various hand gestures to the use of external apparatuses such as the floor.
Thus, what mattered to the weaver was that certain forces (R) that he applied to
the material made the material "behave" or react in an expected way, regardless
of whether he applied these forces with his body or used other external
apparatuses.

This second strategy is in the way the weaver responds to the weave with
different gestures by incorporating the inverse-boundary computation. Here, the
new shapes from the inverse-boundary computation with rule SB1 and SB2 in
Figure 7270 serve as alternative shapes to mediate external force (R) in responding
to the knot's mechanical behavior. For instance, in a condition where a thread is
bending down, applying the inverse-boundary rule on the thread's endpoints
would produce new lines other than the thread itself. These new lines can then be
used as external forces on the thread to stop the bending.

The new shapes for the external forces are parallel to the weaver's
initiative to respond to the knot's mechanical behavior with different gestures or
apparatuses.

Strategy#3 - MateriaAmbiguity
One common feature the traditional weaver and FEM principle share is that the
knot's stiffness is relative. In FEM, this relativity is indicated by possibilities to
modulate Young's modulus of the elasticity (E) by the ratio of stress and strain.
Thus, to have a stiffer material, Young's modulus ratio needs to be higher and vice
versa. In traditional weaving, the relativity is signified by the way the weaver
constantly shifts his or her perception on the material stiffness, depending on his
or her design and structural needs. At one point, the weaver may perceive the
knot to be rigid, and at another time, flexible (see page 66).

This strategy is derived from this view on material relativity. In the

process of designing the knot, the material is not absolute. It can be considered as
rigid or flexible, tensile, or compressive. Rules Ml and M2 accommodate this
136

relativity. Figure 7775 shows the color is used to encode material properties to the
shape. Using color-coding rules, the user can generate different mechanical
behaviors of the knot by recursively changing material properties with this rule,

similarly to the way the weaver shifts his or her perception of the material's
rigidity.

Figure 74. Instantaneous and ever changing external force by the weaver by foot-
grips and mouth-press. 1s2

12 Photos from: Oita Prefecture Basket Weaving Manual

 137

LI L2

/00
Li L2

roooooil

Li L2

LI L2

Figure 75. Loading ambiguity-new shape-relationships.

A different external loading direction is applied to the boundary of the knot with Li. L2
transforms the shape as a consequence of each loading configuration.
138

/2 B1 B1

I
B1 81

/2 B1 B1

4,

B1 B1

Figure 76. Boundary ambiguity-new external forces.

The inverse-boundary rule is applied to the mechanically transformed shape from Figure
7573. In the 3-D space, the inverse-boundary of four points yields a linear shape once, and
applying the rule again to the linear shape yields a 3-D planar shape.
 139

M1 M1

M1 M1

Ml M1

4,

t t
M1 M1

Figure 77. Material embedding-new shape behavior.

Material stiffness is embedded into the shape with coloring rule Ml, recursively. In the left
columns, the shape does not have a material embedded (represented with black). In the
middle columns, each new shape is embedded with flexible material (magenta). In the
right column, the new shapes are embedded with rigid material (cyan), and some of the
original yarn is embedded with flexible material (magenta).
140

The Children of Weaving

Kenneth Snelson once said, "Weaving is the mother of Tensegrity."10 In his
comparative models, one can indeed see the resemblance between the type of
weaves and the type of Tensegrity modules. Yet there are limited references about
his design process on inventing the tensegrity modules from weaving. How did
Snelson sense weaving and come up with Tensegrity? If one understands his
inventing process, could Tensegrity have siblings other than weaving?

In this section, I demonstrate the way abstract weaving grammar derives

new knots designs mechanically and visually. In particular, I demonstrate the way
in which mechanical properties from FEM and boundary rules from shape
grammars work in tandem in generating new kinds of knot. This generative
process also allows us to interpret the way weaving develops into different
systems, including Tensegrity.

Figure 7876 and Figure 7977 show how a knot is generated into different
designs in terms of its mechanical properties. The computation provides corridors
where the user could alternate between seeing the knot as a shape and as a
mechanical configuration. In these figures, the middle column represents seeing
rules computation. The left column represents mechanical embedding. The right
column represents visual embedding.

The first three steps apply the basic mechanical properties on the knot. In
step 1 and 2, rule G1 assigns the thread as a truss and rule M1 applies a rigid
material description to the thread as rigid. In step 3, rule Li applies force
vertically to pull the threads away from each other. In step 4, rule Bi applies the
boundary condition that allows the thread to move only in vertical direction.

In step 5, Hooke's law is applied to the threads, and as a result, the threads
move away from each other. Here is where visual embedding comes into play. To
respond to the knot's disintegration, rule 5B highlights the endpoints of the

I "Kenneth Snelson," accessed November 23, 2009,

http://www.kennethsnelson.net/new-structure/structurel.htm; Eleanor Heartney, Kenneth Snelson: Forces
Made Visible (Lenox, MA: Hard Press Editions, 2009).
 141

threads. With inverse boundary rules, rule B-1 retrieves the boundary of these
points back into lines. Rule M1 is then applied to the new lines, embedding them
with tensile material property. These tensile lines hold up the knot from being
disintegrated, similar to the way Tensegrity structures hold the struts with cables.

Mechanical seeing
Embedding

GsG
./2 01G1

SL

define
2-nodes truss

M2
P see line

SB
apply rigid
material

LI
F, see points
F,
s e SB

apply external
force

B1
P R
F, see points
F,

s F, Q

apply boundary
condition
SL2
P

F,
rigid-body
motion
0

Figure 78. Abstract weaving grammar.

142

F
P

F. SB
FF,
s
F.
rigid-body
motion
Sb2
see points

inverse-boundary
SL
M l>
F,
see lines
RF,
9F,
SL

Fa
Kenneth Snelson Tensegrity Module apply flexible Sb3
material A

see lines

SP
inverse-boundary

PF M1 see plane

F,
F, SP
a
+F,
Paul Jackson Paper Folding Module apply flexible
material B
V,1 M1
see plane
:Fs
,F.

Diana's Pena Tensile-Textile Module apply flexible

material C

Nechanical Seeing Visual

Embedding Embedding

Figure 79. Abstract weaving grammar (continued.)1

"

1" Image credit for tensegrity and paper-folding module: Fernando Sierra, The Bionics Research Group,
Pontificia Bolivariana University, Colombia. From: http://www.sculptors.com/~salsbury/Synergy/. Image
credit for Tensile-Textile Module: Diana Pefia, Ignasi Llorens, and Ramon Sastre, "Application of the
 143

The abstract weaving grammar can further generate more designs from
this condition. From the tensile-rigid configuration, rule B highlights the tensile
lines for rule SL to apply inverse-boundary rules once again onto the lines. As a
result, the lines are retrieved into a plane. Rule SP then highlights this new plane,
and subsequently, rule M embeds an anisotropic material property into the
plane.'5 The knot then behaves like a module of folded sheet, where in this
module, the threads become the folding lines.

Repeating rule M1 in this form would change the material properties once
again. From the folding module, rule SP highlights the plane and then rule M1
applies a different material property onto the plane: isotropic material (e.g., glass,
metal, fabrics). The threads remain rigid, yet the plane is flexible. Examples in
Figure 7977 show the same folding module with different material properties, i.e.,
sticks and fabrics.

Thus, within the same over-and-under configuration, the knot has been
transformed into a different spatial and mechanical configuration.

Case Studies
The examples in the previous section show the way a knot's mechanical
reconfiguration derives different kinds of knot designs that exist. In the following
case studies, I demonstrate the same strategies to reconceptualize weaving into a
novel architectural component.

Study 1: Woven Brick

This case study illustrates the way a weaving material's reconfiguration translates
elements of handicraft weaving into an architectural component. In this case, I
used a mat-weaving surface as a basis to generate the component's design. I
describe the design generation in three phases: (1) visual embedding, (2) shape
materialization, (3) and shape computation.

ofSpace Structures25, no. 1

Tensegrity Principles on Tensile Textile Constructions," InternationalJournal
(March 1, 2010): 57-68, doi:10.1260/0266-3511.25.1.57.
' Anisotropy is the property of being directionally dependent in a material, such as in the case of wood,
which is easier to split along its grain than against it. Isotropic material has properties that behave in the
same manner in all directions, such as glass or metal.
144

In the visual embedding phase, a weaving surface is considered as an

initial condition to embed new shapes. In this initial condition, shapes on the
surface are explored as a two-dimensional pattern, from which a 2-D rectangular
shape is chosen (Figure 8078).

Figure 80. 2D - 3D Weave

With a woven mat as an initial design, the first step considers the mat-weave as a 2-D
pattern consisting of many shapes. After a shape is defined, the material properties are
embedded in the shape. In this case, the shape is assumed to be rigid instead of flexible.
With the rigid shape, I then rememberthe weaving mechanism by applying the
overlapping and reciprocal configuration of weaving.

In the shape materialization phase, the chosen rectangular shape is

retrieved back to its 3-D appearance-a thread with a certain thickness. In this
dimension, the shape is embedded with rigid material property and optimized to
withstand external loading. The optimized shape is rendered as a brick. In
particular, three kinds of brick were defined to reproduce a traditional woven
motif in architectural scale: (1) the "edge brick" for the "erasing" rule, (2) the
"checkered brick" for the checkered motif, and (3) the "continuous bricks" for the
continuous straight line (Figure 8179). The pixelated pattern in the traditional
woven pattern corresponds to the surface of the brick.

In the shape computation phase, the method of assembling the brick into
a point's composition is formulated into computational rules. Three rules guide
the brick assembly process by focusing on two cells in the pattern (black and
white):
 145

* Continuous rule forms a continuous line: IF the next cell is of the same
color as the previous cell, THEN add a continuous brick to form a
continuous line;
" Checkered rule forms a checkered pattern: IF the next cell has
different color from the previous cell, THEN use a checkered brick;
" Edge rule stops the iteration at the edge of the surface: IF there is no
cell found, THEN use an edge brick.

Each brick is connected to the other bricks via notches and tabs.
Simultaneously, the bricks create a self-supported surface and a mosaic pattern
that requires a minimal amount of binding component such as cement or
15 7
mortar. ,1

Figure 8280 shows an illustration of the brick application. As flexible

material (e.g. bamboo) in traditional weaving is considered less permanent for
architectural function, the brick, as an alternative embodiment of weaving
principles, shows a promising weaving application in architectural scale for a
long-term purpose.

Edge
brick
Continuous
* brick
1__ Checkered
brick

No color
detected

same color
detected

different color
detected

' Also note that in this project, the brick module has a rectilinear grid, which creates a rectilinear
surface. Likewise, different brick modules, such as a diamond, circular, or triangular shape, will create
different geometrical patterns.
' Similarly, to eliminate the need for mortar, create low-cost patterns using Meander Grammar.as in Larry
Sass's and Terry Knight's "Meandering Brick." The "Meandering Brick" has integrated alignment features
(i.e., two horizontal notches and one vertical notch serve as aligners) that can be easily matched and
manually locked together without mortar [157].
146

1. Projecting the Grid

2. Applying Checkered Rule

3. Applying Linear and Edge Rule

4. Continue on slated surface

Figure 81. Weaving a brick

 147

Figure 82. Woven house scaled model (top) and application illustration (bottom)
148

Study 2: Bead-Brick
In this case study, I investigate an alternative approach to generating a shape by
considering different sources for the shape's geometry and different materials for
the shape's physical properties compared to the previous case. In terms of the
source for the shape's geometry, the previous case study (woven brick) shows the
way a shape is embedded from a thread's surface. In this case, I show another
example where a shape is embedded from the space between the interwoven
threads. In terms of material, previous case studies (woven timber and woven
brick) demonstrated how weaving is applied to rigid material to replace the
flexible material. This case study investigates ways to incorporate both rigid and
flexible materials into an architectural weaving component.

In the following, I illustrate my investigation through the phases of visual

embedding, shape materialization, and shape computation..

In the visual embedding phase, the initial condition in this case study is
the shadow of a triaxial-weave basket. From this shadow pattern, a hexagon and
the overlapping triangles are chosen.

Figure 83. Visual Embedding in the Shadow patterns

In the shape materialization phase, I develop the hexagon and triangles

into 3-D shapes through mechanical and visual embedding (Figure 8482). To
generate rigid material, I retrieve the point as the boundary of the shapes (step 2
to 3). Using these points, I then apply force to pull the triangles away from each
 149

other (step 4 to 5). As the triangles are separated, rule Bi retrieves the inverse
boundary of the triangles' points into a 3-D shape and embeds the shape into rigid
material (step 6 to 7). To optimize this rigid shape, parts that are at risk of cracking
(at the tip of each corner) and parts that do not have large stress distribution
(center of the shape) are removed (step 13 to 15).

To generate the flexible material, I use inverse boundary rules at the same
initial points at the beginning (step 10). At these points, I generate three lines that
intersect in the middle of the hexagon (step 11). Into these lines, I apply flexible
material property (step 12). These lines go through the 3-D shapes similarly to the
way string goes through a bead in traditional beading. Based on this, the 3-D shape
hereafter is referred to as a beadandthe lines as a string.

In the shape computation phase, the way this bead is assembled is

formulated into a transformation rule-in particular, the transformation that
benefits the inherent geometrical properties of the bead.

The bead has three sides to support its adjacent beads and three sides to be
supported by its neighboring beads. To connect and hold the bead together with
the other beads, the strings go through the bead and compact the bead similarly to
the way traditional beads are tightened and bound together with tension. The
combination of compression from the sides of the bead and tension from the
thread holds the shape together to withstand both axial and lateral loads from
external forces.1 8

' According to the FEM analysis, the center of this shape does not share the same stress as the other sides;
therefore, the center part is subtracted from the shape. The empty space allows the hand to maneuver the
string to tighten the bead. This also helps reduce the weight of the material.
150

1 2 3

G15 B3

Initial Sfape

4 5 6

Li L2 Sb2

8 9
I

sb3 Ml

4,

1 11 12

83 Sb2 M2

14 15

optimization optimization

Figure 84. Bead Transformation

 151

The bead has the order symmetry of 12 -- that is, there are 12 different

ways the beads can be placed in a shape relationship. With these orders of
symmetry, two beads can be arranged in various 2-D and 3-D shape relationships.

In the 2-D relationship, two beads meet as shown in Figure 8583.19 In a 3-D
relationship, two beads are connected so that the two share a large surface to
support each other in 3-D composition (Figure 8785). With this type of connection,
the resulting shape computation generates some emergent design possibilities
that could produce new shapes, such as a tetrahedron, which could later serve in a
new module as an initial shape.

12
syet ry

AL

MOW

Bd-1 Bd-3 Bd-l

""NOW

AIL\ 1A,

Bd-3 ,d-i /Bd-3

/A\

Figure 85. 2D shape relationships

I By comparison, 24 basic designs can be derived from its basic shape relationships with two beads (12
symmetry x 2 beads). To develop a basic loop of triaxial beading with three beads, we can make 36 designs
(12 symmetry x 3 beads), and to develop a hexagonal beaded loop with seven beads, we can make at least 84
basic designs (12 symmetry x 7 beads).
152

-,J r v

Figure 86. Bead as an initial shape and ornamental variations

Four ornamental designs at the bottom show the variation of ornaments based on the
ornament's symmetry.

NI pt

k~

.,~< ~*~>;~ I 9
Figure 87. 3D shape relationships
 153

Figure 88. Illustration of architectural application

Figure 89. 2D Composition

154

Figure 90. A Beaded House (Scaled Model)

 155

3.4 Discussion

In this chapter, I have formulated a series of computational design methods to

reinvent weaving in architectural design using both basic weaving grammar and
abstract weaving grammar.

Basic weaving grammar provides a set of rules to explicitly express

weaving design intentions, to generate designs from such intentions, and to
evaluate the weaving condition from the generated designs. Abstract weaving
grammar provides a set of rules to reconfigure the mechanics of weaving in a
different way, through which a new kind of weaving design can emerge.

In the following, I summarize the way that traditional and computational

methods are hybridized within the grammar, and the way that the grammar
allows an expansion of the language of weaving design to contextualize
computational weaving across cultures.

Traditional Computation in the Computational Tradition

This study shows that in developing the grammars, informal knowledge that is
represented by mental shape and physical shape from traditional weaving is
equally as important as formal theories from computational design and
computational mechanics. The weaver's sensorial experience was not used merely
as a metaphor but more as a tangible model that can be incorporated within the
formal theories.

When developing the basic weaving grammar, I incorporated the polygon

from the weaver's visual perception in formulating a set of rules for weaving
intention and weaving action. Furthermore, the grammar also takes into account
the knot counting process from the way the weaver visually evaluates the
woven/nonwoven condition of traditional weaving. As a result, the revived mental
shapes in the grammar provide an essential visual assistance to generate the
weaving design in a similar fashion the mental shape may help the weaver to
visually engage with the weave.
156

In developing the abstract weaving grammar, I have incorporated findings

from the weaver's haptic experiences into a set of rules for the mechanical
embedding process. The interpretation of the weaver's haptic events in sensing
the weave helped me to define strategies to perceive and design the weave
computationally, particularly with regard to the way in which external forces and
material properties are embedded into the knot.

The Expanding Weaving Language

Several design exercises from the grammars have allowed me to analyze existing
architectural weaving designs as well as synthesize a new kind of weaving for
architectural design.

In exercising the basic weaving grammar, the generated designs have

helped me to revisit several interwoven designs from the past, such as the
reciprocal frame system. By encapsulating the existing designs as part of the
computational weaving design language, I was able to reinvent the designs
algorithmically for different contexts, forms, and materials.

In computing the abstract weaving grammar, the design iteration enabled

me to extend my horizons of weaving design language across crafts such as paper
folding and tensegrity. With this new horizon, I was able to invent new types of
weaving that incorporate methods from different craft disciplines, such as
beading and masonry.

Grammar Assimilation across Cultures

The field study shows that through basic weaving grammar, weaving principles
that appeared to be exclusive to a certain profession are transferable across
disciplines.

I have demonstrated grammar application across media with engravers

and carpenters in Toraja. With paper and pens as visual media, the engravers
easily comprehended the weaving grammar. With wooden beams as physical
media, the grammar helped the carpenter to build a woven-timber installation
within one day. Moreover, grammar comprehension allowed people across age
 157

groups, from craftsmen in Toraja to students from the fifth to ninth grades in
Cambridge, to learn and generate rules from computational weaving.
158
 159

4 CONCLUSIONS

4.1 Summary

In this dissertation, I have outlined my strategies for reinventing architectural

weaving through in situ observation and computational design formulation.

Via a computational lens, traditional weaving that seems deceptively

simple appears to be more complex, collaged with mental shapes that are tacitly
embedded in the weaver's perceptions. Analysis of the weaver's mental shapes
highlights some aspects of weaving that are not bound by the weave's physicality.
This analysis allows these aspects of weaving to be used for other applications
across scale and disciplines.

With explicit representation, the mental shapes are better able to be

incorporated with computational design methods to develop a Weaving Grammar.
The set of rules in the Weaving Grammar is not only able to analyze existing cases,
but also to synthesize new weaving designs for architectural use. With the
increase in new designs, the language of architectural weaving is expanded.

EthnoComputation as a Fieldwork Companion

I titled this thesis 'EthnoComputation' to underline the pivotal role of the human
being in sensing and computing. EthnoComputation helped me, the observer, to
interpret weaver-weave interaction by looking at the shapes that are exchanged
within the weaver's inner-outer environment during his weaving process.

In brief, EthnoComputation helped me to learn from a craftsperson

through the following sequence:

First, I defined an interface between the weaver and the weave. In this
interface, shapes from the weaver's inner environment - mental shapes - interact
with shapes from his outer environment - physical shapes. In the case of weaving,
the interface is the weaver's seeing and touching activities. In the case of bamboo
fabrications, the interface is the weaver's bamboo cutting activities.
160

Second, I interpreted the weaver's mental shapes as an explicit

computational iteration. The iteration is discretized based on changes in the
weaver's sensory attention - e.g., gaze shifting in the weaver's vision. When the
craftsperson is not present, the iteration is discretized based on the craft object's
variations on the site - e.g., differences in the shape of the bamboo on the site.

Third, I examined the reasons underlying the use of mental shapes in the
weaver's action. The examination is subject to several aspects in the weave (e.g.,
form and material) and to theories that pertain to the weaver's sensorial
experience (e.g., Gibson's Perceptual System). The mental shapes in each step
were analyzed individually to elucidate each one's relationship with the physical
shape. To make it possible to analyze several mental shapes simultaneously, the
shapes were superimposed in layers so as to reveal the most repeated thread's
position within several steps.

The Roundtable in Mind

From EthnoComputation study, shapes from the weaver's sensorial experience
are incorporated in the process of developing the Weaving Grammars. A set of
rules provided in the grammars allows an architect to negotiate design
opportunities and consequences during the early design conceptualization that
could anticipate potential issues between shapes and conflicting descriptions.

A design contains shapes and an intention underlying design contains a

certain description. In architectural design, a shape can embody different
descriptions and a certain description can fit different shapes. The two can coexist
or be independent from one another. This, however, is not always the case
between design and intention in practice. Sometimes the relationship between the
two is tightly bonded in that the design becomes the intention. In this tight bond,
shapes in the design are fixed by the intention's descriptions. For instance, a
project brief may say, 'a design is supposed to be effective'. If many precedents
show that vertical shapes are efficient, then the vertical design could become the
intention, regardless of other possible 'efficient' shapes. Here, shapes become
fixated by description, and too rigid for further manipulation (see 'design fixation'
in chapter 1).
 161

In the weaving grammars, this research minimized such fixation by

untying shape from weaving's description. In the grammar, "weaving" is
conceived as an umbrella term that describes an over-and-under shape-
relationship that is composed into a cyclical configuration. Any designs that fit
this description can be encapsulated into the language of weaving, regardless of
scale, shape and material.

This conception allows a designer to synthesize many shapes for weaving

design while at the same time providing guidelines to ensure that the synthesized
design fits with the weaving description. To fit the description, rules in Basic
Weaving Grammars work in tandem: the additive rules generate various over-and-
under design compositions, while the erasing rules confirm that the designs
satisfy the cyclical configurations.

4.2 The Limits and the Prospects

In this section, the limitations of this research and suggestions to leverage these
limitations for future studies are discussed.

Augmented-Me: The Future of EthnoComputation

The field study in this research was conducted by simulating the way I learned
from a weaver in a tacit environment. The data from the weaver-weave interaction
was interpreted using my own perceptual apparatus. As a result, the data
sampling from this interaction has limited precision.

Nevertheless, the process of the weaver-weave interaction was interpreted

through the EthnoComputation framework that organizes how to represent the
weaver's sensorial experience explicitly. This framework could lay the foundation
for future use of cutting-edge technology in capturing more accurate data from
observation. For instance, the EthnoComputation framework could help by
channeling the data gathered by sensory devices, such as eye-tracking devices,
body-motion sensors, and thermal-imaging sensors, into a meaningful
representation for design study. Furthermore, the EthnoComputation framework
could help organize other types of data that are enabled by this technology. For
example, the framework could to used to investigate the shape-and-time
162

relationship between a shape that the weaver sees on the thread and the exact
moment he starts applying muscular force to move the thread. To capture the
data, the weaver's gaze at the thread could be captured by an eye-tracking device,
and the weaver's muscular force could be detected by a thermal-imaging sensor
and body-motion sensor. (However, to preserve the weaver's natural activities, the
use of sensors should intrude as little as possible.)

For better EthnoComputation studies, it is imperative to view the sensorial

technology as an extension of the observer's sensorial apparatus, instead of as a
replacement for the observer's intuition in interpreting the sensorial data. The
value of EthnoComputation rests upon non-deterministic data interpretation. In
terms of design, different data interpretation would lead to new design
reinvention strategies, and in terms of research, different interpretation methods
would enrich the repertoire of computational studies on learning design in tacit
environments.

Digital Shape: On Weaving and Automation

The building workshop in Toraja focused on examining the cycle of design
analysis - design synthesis - design construction. It has yet to examine the
subsequent cycle of building construction - building occupancy - building
demolition/reuse. Thus, although the grammar of the architectural component
was easy for local craftspeople to learn, fabricate and assemble, there are other
architectural aspects of the component that need to be inspected. To meet desired
structural requirements and environmental constraints, the new designs (e.g.,
woven-timber, woven-brick and bead-brick) would need more rigorous scrutiny.

This indicates the need for effective methods, such as digital simulation, to
examine structural or other requirements. Here, the algorithmic form of the
Weaving Grammar comes into play. As an algorithm, the rules of the grammar
can be encoded directly into a computer program, through which the digital
 163

grammar can instantly generate, evaluate and optimize the generated weaving
designs. 161

Having said that, the pair of human and computer in computing the
Weaving Grammar should be symbiotic, as is the pair of aiming-shooting an arrow
and the flight of an arrow in archery. The aiming and shooting depend on the
archer's mind and sensory-motor skill. The arrow's flight is entrusted to the laws
of nature.16 2 The former process is slower, yet crucial and mandatory for the
latter. The latter process is faster and self-regulating, yet computable by the
former process.

This study acknowledges the significance of human sensing and

computation as the predecessor to digital computation. While digital computation
offers high-speed and precise processing, the benefit of human computation lies
in the absence of speed and a determinate answer. In human computation, a
slower process leads to more designs being contemplated, and indeterminate
answers may lead to more design solutions.'3

In EthnoComputation, I have iterated the slow computation process

wherein a mental shape that the weaver perceived for less than a second was
interpreted over a several-month period during this thesis process into many
alternative shapes. In the Weaving Grammars, I demonstrated an indeterminate
computation in which different ways of looking at the shape generates many
weaving designs.

1 For more example on how the grammar is translated into a script language, see Rizal Muslimin,

"Interweaving Grammar: Reconfiguring Vernacular Structure Through Parametric Shape Grammar,"

ofArchitectural Compudng8, no. 2 (September 1, 2010): 93-110, doi:10.1260/1478-
InternadonalJournal
0771.8.2.93.
' The interaction between the arrow's outer environment, e.g. air speed, wind-direction, gravity, and inner
environment e.g., mass, materials, force and velocity.
' See more about 'Slow Theories' and the benefit of Slow Computation in Terry Knight, "Slow Computing:
Teaching Generative Design with Shape Grammars," in ComputadonalDesign Methods and Technologies:
ApplicationsinCAD, CAMand CAEEducadon, ed. Ning Gu and Xiangyu Wang (IGI Global, 2012),
http://www.igi-global.com/chapter/slow-computing-teaching-generative-design/62940., p.35
164

4.3 Borderless Weaving

It would be misleading to claim that the EthnoComputation study establishes how

the weaver thinks about weaving, or that the rules in the Weaving Grammars
demonstrate how past designers reinvented weaving. There are weaving designs
produced without the Weaving Grammars. There are designers who might
develop similar weaving design-rules without observing the traditional weaver.

The coexistence of traditional weaving and computational weaving in this

study is not about nostalgic preservation of traditional craft. Nor is it about
cultural justification of a computational process. Here, their coexistence is part of
a strategy of systematizing design resources (in traditional craft) and design
conceptualization (in architectural design). In this system, inspirations and ideas
become computable, regenerative and renewable. EthnoComputation and
Weaving Grammar have linked traditional weaving and architectural weaving
with regard to the way in which weaving principles can be applied across
disciplines, hence blurring the boundary between crafts and architecture.

Within this borderless setting, the strong affiliation of weaving and

architecture in their origins was revisited and revitalized in this thesis.
 165

4.4 Epilogue

Toward the end of work on this dissertation, several headlines demonstrated the
emerging recognition of digital manufacturing as a national asset. In Zurich, the
architecture and digital fabrication lab at ETH, led by Matthias Kohler, received a
substantial grant from the Swiss government to advance architectural building
methods through research in robotics (about US$14.7 million as part of the
National Centre of Competence in Research 2014-2017 programs). 16 In June 2014,
in Washington, D.C., Maker Faire, an annual event for the digital fabricator
community, was held for the first time on the White House lawn. The exhibition
revealed the democratization of digital technology and the emerging Do-It-
Yourself (DIY) mentality within American society.

The headlines from Zurich and Washington, D.C., suggest that rethinking
the production system for building architecture is becoming equally as important
as valuing the built architecture as a product.

Government support for digital manufacturing, either by the Swiss or the

U.S. governments, indicates a political effort to assimilate manufacturing
technology as part of a nation's culture. For Switzerland, this strategy is pragmatic
and rational. The high hourly wage in their manufacturing industry (i.e., the
second-highest in the world at $34.29/hr.)'66justifies their research budget to
increase the capacity of their robotic technology to minimize labor costs. For the
U.S. government, hosting the Maker Faire event is an endorsement that supports
the "American Manufacturing Renaissance". 167 Disney's popular mantra in Maker
Faire that "if you can dream it, you can make it", is well aligned with the American
motto that "if you work hard, you can get ahead."

'1 As part of the new National Centres of Competence in Research (NCCR). Source available from:
htt://www--sbfi-admin-ch/themen/01367/01677/lndex.htmt?langaen
1
Based on Global Wage Report 2012/13: Wages and equitable growth: (International comparison of hourly
direct pay for time worked in manufacturing, 2010), published by the International Labour Organization
(ILO). Available from:
/

1 Becky Fried and Becky Wetstone, "President Obama at the White House Maker Faire: 'Today's
D.I.Y. Is
Tomorrow's "Made in America"' I The White House," June 18, 2014,
http://www.whitehouse.gov/blog/2O14/06/18/president-obama-white-house-maker-faire-today-s-diy-
tomorrow-s-made-america.
166

The national context of robotic technology and 3D printing justifies efforts

at technological advancement. This justification sheds light on a larger
implication of reinventing traditional crafts via computational methods.
Compared to Switzerland and the U.S., the context of this study is more modest,
yet nevertheless unique.

In terms of labor cost, developing countries such as Indonesia have a low

170
hourly wage (around US$75/month. to US$200/month.). Additionally, the
country has a large labor force (about 116,500,000). Thus, it presents an opposite
situation to that of Switzerland, which has a much smaller labor force.

In terms of knowledge assimilation, the nature of crafting knowledge

distribution is tacit. In contrast, one important source of the Maker Faire's success

is that the knowledge is explicit and accessible. This explicit knowledge, along
with the affordability of buying the components, has helped to accelerate the
democratization of digital technology in the U.S.

However, in terms of craftsmanship, the Torajan community has many

crafting disciplines, each of which expresses diverse techniques and designs. The
craftspeople are supported by the abundance of local materials. In this fertile
environment, crafting is pervasive and diverse, manifesting itself in the
production of accessories, handicrafts, clothes musical instruments and buildings
themselves.

Correspondingly, crafting activities in Toraja, as local context, justified my

computational design strategy. In adapting this strong crafting culture, the issues
of labor and knowledge distribution were contextualized two strategies.

First, the assembly techniques for the new designs were tailored for
human hands rather than robotic arms. To weave the timber beams, one needs to
fit the notches of one beam with the tabs from the neighboring beams. To bead

' Based on the Ministry of Manpower and Transmigration of the Republic of Indonesia report on the 2014
Regional Minimum Wage (UMR) . Source available from:
http://infokerijadepnaker.blogspot.com/2013/11/Daftar-Gaji-Terbaru-UMR-UMK-Regional-Kota-Seluruh-
Indonesia.html
 167

the bricks, one needs to insert the strands through a small hole in a brick, weave
the threads inside the brick's hole, and compact the bricks by pulling the strands
and packing the bricks together. These maneuvers are relatively complex to be
executed by robots.

Second, the visual representation of the Weaving Grammar allows the

craftsman to learn the rules tacitly. Furthermore, the grammar is not tightly
bonded with a certain medium. The media can be visual or physical. As a result,
the grammar can be computed across crafts disciplines in Toraja, such as in
drawing by the engravers and in assembling the timber by the carpenter. By
having the grammar used in different crafts, the nature of the design in each craft
can enrich the design language and the versatility of the grammar.

The above strategy shows that the absence of computer-aided

manufacturing at the traditional site does not mean a computational fabrication
cannot be performed. The case studies show that without computers, the
craftsman can nonetheless employ computational thinking.

Thus, introducing a machine to replace the labor force in traditional crafts

may not be culturally sustainable. Teaching computational design methods to
craftspeople so that they can develop their own rules and designs may be more
empowering. The Indonesian crafting renaissance has yet to come. But if it is
developing, it is unlikely to resemble what was showcased at the Maker Faire
festival. The renaissance may be quiet and tacit, yet perceptible and computable.
168
 169

BIBLIOGRAPHY

Adams, Kathleen M. Art asPolidcs:Re-CrafngIdentities, Tourism, andPowerin

Tana Toraja, Indonesia. Honolulu: University of Hawaii Press, 2006.
Albers, Anni. On Weaving. Wesleyan University Press, 1965.
Alexander, J. W. "Topological Invariants of Knots and Links." Transactionsofthe
American MathematicalSociety30, no. 2 (1928): 275-306. doi:10.1090/S0002-
9947-1928-1501429-1.
American, Mathematical Society. ApplicationsofKnot Theory: American
MathematicalSociety Short Course, Januay4-5, 2008, San Diego,
California.Providence, R.I: American Mathematical Society, 2009.
Ascher, Marcia. Mathematicsof the Incas:Code ofthe Quipu. Dover Publications,
1997.
Ateshian, Gerard A. "Artificial Cartilage: Weaving in Three Dimensions." NatMater
6, no. 2 (February 2007): 89-90. doi:10.1038/nmatl830.
Barker, Henry. Egyptian Gods & Goddesses. Grosset & Dunlap, 1999.
Barlow, Alfred. The History andPrinciplesof Weaving, by HandandbyPower.
London: Low, Marston, Searle, & Rivington, 1878.
Bathe, Klaus-Jiirgen. FiniteElement Procedures.Englewood Cliffs, N.J: Prentice
Hall, 1996.
Bell, T. F. JacquardWeaingandDesigning. Longmans, Green, 1895.
Boas, Franz. PrimitiveArt. 1st Dover Edition. Dover Publications, 1955.
Bucalem, Miguel Luiz, and Klaus-Jurgen Bathe. The Mechanics ofSolids and
Structures-HierarchicalModelingand the FiniteElement Solution. 2011
edition. Berlin; New York: Springer, 2011.
Deleuze, Gilles. A ThousandPlateaus:CapitalismandSchizophrenia.Minneapolis:
University of Minnesota Press, 1987.
Bergsonism. New York: Zone Books, 1988.
Dietz, Ada K. Algebraic Expressionsin Handwoven Textiles. Louisville, Kentucky:
The Little Loomhouse, 1949.
Dunkelberg, Klaus. IL 31, Bambus. IL 31, Bamboo. BambusAls Baustoff Bauen Mit
PflanzlichenStiben. Kramer, Stuttgart, 2000.
Eglash, Ron. African Fractals:Modern CompudngandIndigenousDesign. 3rd
edition. New Brunswick, N.J: Rutgers University Press, 1999.
"Culturally-Situated Design Tools." CulturallySituatedDesignTools:
TeachingMath and Computingthrough Culture. Accessed May 16, 2014.
http://csdt.rpi.edu/.
170

Eglash, Ron, Audrey Bennett, Casey O'donnell, Sybillyn Jennings, and Margaret
Cintorino. "Culturally Situated Design Tools: Ethnocomputing from Field
Site to Classroom." American Anthropologist108, no. 2 (June 1, 2006): 347-
62. doi:10.1525/aa.2006.108.2.347.
Elkins, James. WhatPaintingIs.1 edition. Routledge, 2004.
Erkki Sutinen, Mikko Vesisenaho. "Ethnocomputing in Tanzania: Design and
Analysis of a Contextualized ICT Course." Research andPracticein
Technolog EnhancedLearning1 (2006): 239-67.
doi:10.1142/S1793206806000238.
Essinger, James. Jacquard'sWeb: How a Hand-Loom Led to the Birth of the
InformationAge. Oxford: Oxford University Press, 2004.
Evans, Miriama. The Art ofMgori Weaving: The EternalThread Wellington N.Z.:
Huia Publishers, 2005.
Fried, Becky, and Becky Wetstone. "President Obama at the White House Maker
Faire: 'Today's D.I.Y. Is Tomorrow's "Made in America'" IThe White
House," June 18, 2014.
http://www.whitehouse.gov/blog/2014/06/18/president-obama-white-house-
maker-faire-today-s-diy-tomorrow-s-made-america.
Geertz, Clifford. Works andLives: The AnthropologistasAuthor. 1 edition.
Stanford, Calif.: Stanford University Press, 1989.
Gerdes, P. "Fivefold Symmetry and (basket) Weaving in Various Cultures." Fivefold
Symmetry, 1992, 245.
Gibson, James Jerome. The EcologicalApproachto VisualPerception. Routledge,
1986.
The Senses Consideredas PerceptualSystems. Boston: Houghton Mifflin,
1966.
Gillow, John, and Bryan Sentance. World Textiles:A Visual Guide to Traditional
Techniques. Thames & Hudson, 2005.
Goldstein, E. Bruce. "The Ecology of J. J. Gibson's Perception." Leonardo14, no. 3
(Summer 1981): 191-95.
Greenfield, Patricia Marks. Weaving GeneradonsTogether:Evolving Creativityin
the Maya of Chiapas. Illustrated edition. School of American Research
Press, 2004.
Hamedi, Mahiar, Robert Forchheimer, and Olle Inganas. "Towards Woven Logic
from Organic Electronic Fibres." NatMater6, no. 5 (May 2007): 357-62.
doi:10.1038/nmatl884.
Heartney, Eleanor. Kenneth Snelson:ForcesMade Visible. Lenox, MA: Hard Press
Editions, 2009.
Hockings, Paul. Principlesof VisualAnthropolog. 3 edition. Berlin: De Gruyter,
2003.
 171

Ingold, Tim. Bringing ThingsBack to Life: CreativeEntanglements in a World of

Materials. Working Paper. Realities / Morgan Centre, University of
Manchester, July 20, 2010.
http://www.manchester.ac.uk/realities/publications/workingpapers/.
Making:Anthropology, Archaeology, Art andArchitecture.Routledge, 2013.
"OfBlocks and Knots: Architecture as Weaving." Accessed May 17, 2014.
http://www.architectural-review.com/comment-and-opinion/of-blocks-
and-knots-architecture-as-weaving/8653693.article.
Jansson, David G., and Steven M. Smith. "Design Fixation." Design Studies 12, no. 1
(January 1991): 3-11. doi:10.1016/0142-694X(91)90003-F.
Kawabata, S., Mari Inoue, and Masako Niwa. "Non-Linear Theory of the Biaxial
Deformation of a Triaxial-Weave Fabric." Journalofthe Texdle Institute83,
no. 1 (1992): 104-19. doi:10.1080/00405009208631180.
Kawabata, S., Masako Niwa, and H. Kawai. "3-the Finite-Deformation Theory of
Plain-Weave Fabrics Part I: The Biaxial-Deformation Theory." Journalof
the Textile Institute 64, no. 1 (1973): 21-46. doi:10.1080/00405007308630416.
"Kenneth Snelson." Accessed November 23, 2009.
http://www.kennethsnelson.net/new-structure/structurel.htm.
Kent, Kate P. "West African Decorative Weaving." African Arts 6, no. 1 (October 1,
1972): 22-88. doi:10.2307/3334637.
Kis-Jovak, Jowa Imre. Banua Toraja:ChangingPatternsin Archftecture and
Symbolism Among the Sa'dan Toraja, Sulawesi, Indonesia. Amsterdam:
Royal Tropical Institute, 1988.
Knight, T W. "Transformations of the Meander Motif on Greek Geometric Pottery,"
Design Computing, no. 1 (1986): 29-67.
Knight, T. W. "Constructive Symmetry." EnvironmentandPlanningB:Planning
andDesign 22, no. 4(1995): 419-50. doi:10.1068/b220419.
Knight, T. Weissman. "Languages of Designs: From Known to New." Environment
andPlanningB:Planning andDesign8,no. 2 (1981): 213-38.
doi:10.1068/b080213.
Knight, Terry. "Computing with Ambiguity." EnvironmentandPfanningB:
PlanningandDesign 30, no. 2 (2003): 165-80. doi:10.1068/bl2915.
"Computing with Emergence." Environmentand PlanningB: Planningand
Design30, no. 1 (2003): 125-55. doi:10.1068/bl2914.
"Slow Computing: Teaching Generative Design with Shape Grammars." In
ComputationalDesignMethods and Technologies:Applicationsin CAD,
CAMand CAEEducadon, edited by Ning Gu and Xiangyu Wang. IGI Global,
2012. http://www.igi-global.com/chapter/slow-computing-teaching-
generative-design/62940.
172

Knight, Terry, and George Stiny. "Classical and Non-Classical Computation." Arq:
ArchitecturalResearch Quarterly5, no. 04 (2001): 355-72.
doi:10.1017/S1359135502001410.
"Making Grammar: From Computing with Shapes to Computing with
Things." In ComputationalMaking. London, 2014.
Kraus-Boelte, Maria, and John Kraus. The KindergartenGuide: The Gifts. - V.8. the
Occupadons. Nabu Press, 2010.
Larsen, Olga Popovic. ReciprocalFrameArchitecture. 1st ed. Architectural Press,
2008.
Latour, Bruno. "Gaston, A Little Known Successor of Daedalus." translated by
Charis Cussins, n.d.
""'Where Are the Missing Masses? The Sociology of a Few Mundane
Artifacts.""' In Shaping Technology/BuildingSociety: Studies in
SociotechnicalChange, by Wiebe E. Bijker and John Law. Cambridge,
Mass.: The MIT Press, 1994.
Lewis, Frank A., and Robert Bolton. Form, Matter, andMixtureinAristode.
Oxford, UK; Malden, MA: Wiley-Blackwell, 1997.
Liu Suyi. "Fabric Weave Design Based on Cellular Automata Theory," May 15, 2009,
145-47. doi:10.1109/IUCE.2009.88.
MacDougall, David. The CorporealImage: Film, Ethnography, and the Senses.
Princeton, N.J: Princeton University Press, 2005.
Macrae, Norman. John Von Neumann: The Sciendfc Genius Who Pioneeredthe
Modern Computer, Game Theory, NuclearDeterrence,andMuch More.
AMS Bookstore, 2000.
Marmodoro, Anna. "Aristotle's Hylomorphism without Reconditioning." Accessed
May 17, 2014.
https://www.academia.edu/4195031/Aristotles-hylomorphism-without-rec
onditioning.
Marr, David, Shimon Ullman, and Tomaso A. Poggio. Vision:A Computational
Investigationinto the Human RepresentationandProcessingof Visual
Information. MIT Press, 2010.
Matti, Tedre. "Ethnocomputing: A Multicultural View on Computer Science." M.Sc
Thesis, University of Joensuu, 2002.
Moutos, Franklin T., Lisa E. Freed, and Farshid Guilak. "A Biomimetic Three-
Dimensional Woven Composite Scaffold for Functional Tissue Engineering
of Cartilage." NatMater6, no. 2 (February 2007): 162-67.
doi:10.1038/nmat1822.
Murasugi, K. "Knot Theory in Molecular Biology." Knot Theory &Its Applicadons,
2008,267-83.
 173

Muslimin, Rizal. "Decoding Passura' - Representing the Indigenous Visual

Messages Underlying Traditional Icons with Descriptive Grammar." In
Open Systems, 781-90. Singapore: Southeast University Press Pub.
Date :2007-10-01, 2013.
"Gestalt Grammar in Weaving Design: Leveraging Visual Affordance in
Analysing and Synthesizing Weaving Design Based on the Laws of Prdgnanz
and Shape Grammar (poster)." Texas, USA, 2012.
"Interweaving Grammar: Reconfiguring Vernacular Structure Through
Parametric Shape Grammar." InternadonalJournalofArchitectural
Computing8, no. 2 (September 1, 2010): 93-110. doi:10.1260/1478-
0771.8.2.93.
"Learning from Weaving for Digital Fabrication in Architecture." Leonardo
43, no. 4 (July 16, 2010): 340-49. doi:10.1162/LEON-a_00007.
"One-Piece Weaving: Reconfiguring Folding and Knotting Algorithm in
Computational Design," 9-18. Newcastle, Australia, 2011.
"Recursive Embedding of Gestalt Laws and Shape Grammar in the Weaving
Design Process," 1:443-49. Prague, Czech Republic, 2012.
O'Reilly, Karen. Key Concepts in Ethnography. SAGE Publications Ltd, 2014.
Otto, Frei. IL-37AncientArchitects. Universitdt Stuttgart, 1994.
Otto, Berthold Burkhardt, Jugen Hennicke Frei. IL 10 - GridShells. Institute for
Lightweight Structures, University Stuttgart, n.d.
Pefia, Diana, Ignasi Llorens, and Ramon Sastre. "Application of the Tensegrity
Principles on Tensile Textile Constructions." InternadonalJournalofSpace
Structures25, no. 1 (March 1, 2010): 57-68. doi:10.1260/0266-3511.25.1.57.
Pink, Sarah. DoingSensory Ethnography. SAGE Publications Ltd, 2012.
Doing VisualEthnography. Second Edition edition. SAGE Publications Ltd,
2012.
Doing VisualEthnography. Third Edition edition. SAGE Publications Ltd,
2013.
Polanyi, Michael. Knowing andBeing: Essays byMichaelPolanyi. First Edition
edition. Chicago: University Of Chicago Press, 1969.
Prechtel, Martin, and Robert S. Carlsen. "Weaving and Cosmos Amongst the
Tzutujil Maya of Guatemala." RES: AnthropologyandAesthetics, no. 15
(April 1, 1988): 122-32.
Purcell, A. Terry, and John S. Gero. "Design and Other Types of Fixation." Design
Studies, Special Issue: Design Cognition and Computation, 17, no. 4
(October 1996): 363-83. doi:10.1016/S0142-694X(96)00023-3.
Salvadori, Mario. The Art of Construction:Projectsand PrinciplesforBeginning
EngineersandArchitects. 3rd ed. Chicago: Chicago Review Press, 1990.
174

Semper, Gottfried. The FourElements ofArchitecture and Other Writings. Reissue.

Cambridge University Press, 1851.
Sentance, Bryan. Basketry:A World Guide to TradidonalTechniques. Thames

&
Hudson, 2007.
Shields, Christopher. "Aristotle." In The StanfordEncyclopeda ofPhilosophy,
edited by Edward N. Zalta, Spring 2014., 2014.
http://plato.stanford.edu/archives/spr2014/entries/aristotle/.
Simon, Herbert A. The Sciences ofthe Arificial- 3rdEdidon. 3rd ed. The MIT
Press, 1996.
"SmartBird." Accessed May 16, 2014.
http://www.festo.com/cms/en.corp/11369.htm.
Sossinsky, Alexei. Knots: Mathematicswith a Twist. Harvard University Press,
2004.
Stiny, G. "Ice-Ray: A Note on the Generation of Chinese Lattice Designs."
EnvironmentandPlanningB: PlanningandDesign4, no. 1(1977): 89-98.
doi:10.1068/b040089.
"Weights." EnvironmentandPlanningB: PlanningandDesign 19, no. 4
(1992): 413-30. doi:10.1068/b190413.
Stiny, George. Shape: Talking aboutSeeing andDoing. The MIT Press, 2008.
Stiny, George, and James Gips. "Shape Grammars and the Generative Specification
of Painting and Sculpture." InformationProcessing71, no. 1460-65 (1972):
125.
"Shape Grammars and the Generative Specification of Painting and
Sculpture." In Information Processing71. Ljubljana, Yugoslavia, 1971.
Todd, Mattie Phipps. Hand-Loom Weaving:A ManualforSchoolandHome. Rand,
McNally & company, 1902.
Waterson, Roxana. PathsandRivers: Sa'dan Toraja Societyin Transformation.
Leiden: KITLV Press, 2009.
The Living House:An AnthropologyofArchitecturein South-EastAsia.
Tuttle Publishing, 2010.
 175

LIST OF FIGURES

Figure 1. Shape Grammar Computation 23

Figure 2. Pseudo-Woven 28
Figure 3. Woven Bentwood and Sculptural Weaving 29
Figure 4. Street photographer in Dehra Dun, India, 1989 37
Figure 5. The Disappearing European man, 1905 37
Figure 6. Women in Bullfighting 38
Figure 7. Introduction to Java Programming Language in Tanzania 41
Figure 8. African Fractal simulation 41
Figure 9. Algebra of Perceptual Shape 45
Figure 10. Rules and Computation in Shape Grammar 46
Figure 11. Rules A -4 B and Computation 47
Figure 12. Extended Rules Iteration 47
Figure 13. Inverse-computation from existing shapes into interpretive rules 49
Figure 14. Kete Kesu, Toraja, Indonesia 50
Figure 15. The initial setting for EthnoComputation 50
Figure 17. Weighted objects on the Site 52
Figure 18. Object associations 53
Figure 19. Reconfigured Sites from Object Association: Nature-Craft and Human-Craft 53
Figure 20. Abundant Bamboo trees (top) and various bamboo crafts in Kete Kesu (bottom) 55
Figure 22. Cutting rules on the surface 58
Figure 23. Lattice of Cutting Rules on Bamboo surface 60
Figure 24. Cutting rules on cross-sectional profile 61
Figure 25. Lattice of Cutting Rules on Bamboo Cross-Section 63
Figure 26. Recursive Causalities 64
Figure 28. Sensorial Weight 68
Figure 29. Sensorial Computation 69
Figure 30. Continuous Attention Span 70
Figure 31. Ground shapes as mental shape in weaving 71
Figure 32. Visual-Haptic Coordination 72
Figure 33. Visual Assessment 73
Figure 34. Embedded and Embodied Shape 74
Figure 35. Emergent perceptual shapes 74
Figure 37. Weaver-Weave Sensorial Exchange 78
176

Figure 38. Weaver's motion from the observer's eyes 82

Figure 39. Weaver's - Observer's Orientation Alignment 83
Figure 40. Weave position from the weaver's eyes 84
Figure 41. Asymmetrical Making 85
Figure 42. physical shape and mental shape 95
Figure 43. Rules to define the left-hand shape Weaving Intention 96
Figure 44. the development of weaving Intention 97
Figure 45. Additive Rule to define the right-hand shape for weaving Action 98
Figure 46. Shape's order of symmetry as the basis of transformation 98
Figure 47. A new rotation point for the new shape 99
Figure 48. Knot Counting 102
Figure 49. Rotation Point Interpolation 103
Figure 50. Point's Determination 104
Figure 51. Liberation of the Point 105
Figure 52. Ratio in point and line relationship in embedding shape on the thread 107
Figure 53. What shape to compute? 107
Figure 54. Where to add a new shape? 109
Figure 55. Rule Rhythms 109
Figure 56. Analysis of Existing Design 110
Figure 57. 2D Computational Weaving Workshop 113
Figure 58. 2D shape is translated into physical 3D shape. 114
Figure 59.3D Weaving Workshop -Session 1: Weaving with rule 116
Figure 60. 3D Weaving Workshop - Session 2: The No-Rule Session 116
Figure 61. Idealized model for structural simulation 118
Figure 62. Displacement and Stress Distribution from five notch locations 119
Figure 63. The diagram shows the effect of load distribution on the design pattern generated
with one rule: R7A. 119
Figure 64. Three models of interwoven beamspof equal volume and weight are generated from
three rules in the grammar. 120
Figure 65. Optimization Test and Weighting (vertical and horizontal) 121
Figure 66. Results from Compressive Test 122
Figure 67. A carpenter holds the scaled model during the fabrication process to verify where to
place the timber. 125
Figure 68. The Mock Up Installation 125
Figure 69. Rules for FEM idealization in shape grammar. 129
Figure 70. Kawabata's Schema Reiteration 130
Figure 71. Variables for optimization of the knot's rigidity.in the Kawabata's schema. 131
 177

Figure 72. Boundary rule iterations. 132

Figure 73. Different focus in in optimization between internal optimization and external
optimization. 133
Figure 75. Loading ambiguity-new shape-relationships. 137
Figure 76. Boundary ambiguity-new external forces. 138
Figure 77. Material embedding-new shape behavior. 139
Figure 78. Abstract weaving grammar. 141
Figure 80. 2D - 3D Weave 144
Figure 81. Weaving a brick 146
Figure 82. Woven house scaled model (top) and application illustration (bottom) 147
Figure 83. Visual Embedding in the Shadow patterns 148
Figure 84. Bead Transformation 150
Figure 85. 2D shape relationships 151
Figure 86. Bead as an initial shape and ornamental variations 152
Figure 87. 3D shape relationships 152
Figure 88. Illustration of architectural application 153
Figure 89. 2D Composition 153
Figure 90. A Beaded House (Scaled Model) 154