Content uploaded by Raghu Garud
Author content
All content in this area was uploaded by Raghu Garud on Jul 23, 2014
Content may be subject to copyright.
Incomplete by Design and Designing for
Incompleteness
Raghu Garud, Sanjay Jain and Philipp Tuertscher
Abstract
The traditional scientific approach to design extols the virtues of completeness. However,
in environments characterized by continual change, there are challenges in adopting such
an approach. We examine Linux and Wikipedia as two exemplary cases to explore the
nature of design in such a protean world. Our observations highlight a pragmatic approach
to design in which incompleteness is harnessed in a generative manner. This suggests a
change in the meaning of the word ‘design’ itself — from one that separates the process
of design from its outcome, to one that considers design as both the medium and outcome
of action.
Keywords: platform, emergence, design, innovation community
Historically, much of the discourse on design has extolled the virtues of com-
pleteness. Completeness allows for the pre-specification of a problem, the iden-
tification of pre-existing alternatives and the choice of the most optimal
solution. Such a scientific approach to design pervades much of management
thinking, education and research (Romme 2003: 24).1
For instance, this approach is evident in the design of traditional organizations
at the turn of the 20th century. Organizations enhanced the efficiency of their
operations by systematically applying principles of scientific management to
discover ‘the one best way’ to organize (Kanigel 1997). Interchangeable parts,
division of labor, routinization — each of these were features of an organiza-
tional design capable of mass producing ‘any color car as long as it was black’
(Ford and Crowther 1922: 72).
For such an approach to work, however, there needs to be a clear and stable
boundary between the entity being designed and the context for which it is
being designed. Such a boundary makes it possible to fix the purpose of a
design based on a stable set of user preferences and performance expectations.
Clear boundaries, stable preferences and fixed goals: these form the corner-
stones of the scientific approach to design as articulated by Simon (1996).
But how does such an approach to design hold up in environments charac-
terized by continual change? What if there are multiple designers, each with
their own representation of the problem? What if users of a design are also its
designers? To further complicate matters, what if the process of discovering
new and potentially better states only takes place through a process of partici-
pation, and the unfolding of the process itself changes the problem?
article title
Organization
Studies
29(03): 351–371
ISSN 0170–8406
Copyright © 2008
SAGE Publications
(Los Angeles,
London, New Delhi
and Singapore)
Raghu Garud
Pennsylvania State
University, USA
Sanjay Jain
University of
Wisc onsin at
Madison, USA
Philipp Tuertscher
Vie nna Un iver sit y o f
Economics and
Business
Administration,
Austria
www.egosnet.org/os DOI: 10.1177/0170840607088018
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
This is the new frontier in which we find ourselves. There is no clear separa-
tion between the inside and the outside, text and context. Rather, there is only
an evolving and emerging network of associations (Barry and Rerup 2006:
267). Problems are ill-defined, preferences are fluid and solutions emerge in
action. In such situations, an emphasis on completeness is likely to result in the
creation of designs that foreclose future options.
It is useful to consider the dual meaning of the word ‘design’ within this con-
text. As a verb, ‘to design’ refers to the process of developing a plan for a prod-
uct, structure or component. As a noun, ‘a design’ is used to connote the
outcome of the process.2In traditional settings, these two meanings of design
have been separated from one another. One would engage in a process of design
(the verb) so as to emerge with a design (the noun) for a specific context. In con-
temporary settings, however, designs are more appropriately viewed as being
simultaneously noun and verb, with every outcome marking the beginning of a
new process.3Put differently, designs are like dynamic jigsaw puzzles in which
multiple actors assemble pieces within templates that change as a result of the
actors’ engagement.
It is this proposition that we develop in the paper. We suggest that, rather than
a scientific approach that tends to separate the two meanings of design, we must
embrace a pragmatic approach to design that simultaneously embraces both
process and outcome. Given this dual connotation, designs, by definition, have
to deal with incompleteness. However, rather than pose a threat, incompleteness
acts as a trigger for action. Even as actors try to complete what has been left
incomplete, they generate new problems as well as new possibilities that con-
tinually drive the design. In this way, incompleteness is both a cause and con-
sequence of the dynamics of organizing in contemporary environments.
We begin by p ro vi ding a b ri ef ov er vi ew o f th e sci en ti fi c ap pr oa ch t o desi gn
and then highlight the challenges that one confronts in applying this within con-
temporary environments characterized by continual change. To empirically
locate our observations, we examine two exemplary designs that appear to be
always in-the-making: the Linux operating system and the Wikipedia online
encyclopedia. We find that, rather than one group designing for another’s con-
sumption, designs emerge through situated use as actors co-theorize across mul-
tiple settings, and, in the process, create new options. These dynamics produce
self-perpetuating processes that further drive continual change.
To Desi gn or Not to Desi gn?
In his book, The Sciences of the Artificial (1996), Simon suggested that the
design of artificial systems — meaning man-made as opposed to natural — is
contingent upon the goals of the designer and the purposes for which the system
is designed. A key initial task for designers involves the specification of system
boundaries. As Simon pointed out,
‘An ar ti fa ct c an be t ho ug ht o f a s a m ee ti ng po in t — a n “i nt er fa ce ” in to da y’s t er ms —
between an “inner” environment, the substance and organization of the artifact itself, and
an “outer” environment, the surroundings in which it operates. If the inner environment
352 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
is appropriate to the outer environment, or vice versa, the artifact will serve its intended
purposes.’ (Simon 1996: 6)4
Once an interface has been specified and the problem has been defined in
terms of its context, form and goals, it is possible to proceed using the organiz-
ing principles to be found in the natural world. For instance, Simon (1962)
offered the principle of ‘near decomposability’ as a concept possessing clear evo-
lutionary advantages for both natural and artificial systems (see also Langlois
2002). Decomposability refers to the partitioning of a system in such a way that
the interactions of elements within a subassembly are greater than the interac-
tions between them. Such decomposition reduces the complexities confronted by
boundedly rational human beings in their efforts to design artifacts.
To i ll us tr at e t hi s po in t, S imon of fered a p arabl e of t wo w at ch make rs , Tem pu s
and Hora. Tempus organized his work in a manner that if he had ‘one (watch)
partly assembled and had to put it down — to answer the phone, say — it imme-
diately fell to pieces and had to be reassembled from the elements’.
Consequently, every time Tempus was interrupted and forced to set aside his
work, the entire unfinished assembly fell to pieces. In contrast, Hora first built
stable subassemblies that he then put together in a hierarchic fashion into a
larger stable assembly. Thus, when Hora was interrupted, only the last unfin-
ished subassembly fell apart, preserving most of his earlier work.
According to Simon, the differential cost of incompleteness that the watch-
makers confronted is a specific case of a more general challenge that individu-
als confront when addressing complex problems. The differential cost can be
explained by the interplay between the short- and long-term memories. When
individuals address complex problems, transactions are carried out in their
short-term memory. Given the limits to short-term memory, any interruption to
a task can exact a toll. This is because any intermediate outcome that might have
been accomplished before the interruption is lost. While it is possible to store
intermediary outcomes in long-term memory, this too exacts a toll as the trans-
fer between long- and short-term memory often requires considerable effort.
Scientific Design in Practice
The concept of decomposability that emerges from this parable has generated a
lot of attention from scholars who study modularity (Baldwin and Clark 2000).
Specifically, complex systems can be decomposed in a way similar to that sug-
gested by Simon, in other words, into ‘modules’ (Langlois and Robertson
1992). Each module only interacts with another through standardized inter-
faces. As a result, each module becomes a ‘black box’ (Rosenberg 1982), pos-
sessing the detail required for its functioning, but hiding such detail from other
interdependent modules (Parnas 1972).
An implicit assumption in this literature is that the overall system architec-
ture needs to be completely specified a priori (Baldwin and Clark 2000). In
terms of Simon’s parable, this means that both Tempus and Hora work with
designs that have clearly defined boundaries and pre-set user preferences (accu-
rate time-keeping, for example). In such cases, the key design decisions revolve
around issues such as detailing the elements that comprise the architecture,
Garud et al.: Incomplete by Design and Designing for Incompleteness 353
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
establishing stable interface specifications to ensure smooth functioning
between modules, and ‘black-boxing’ the modules to mask their complexity.
These decisions are integral to a design approach that values completeness.
The form and function of a system must be clearly specified. Only with such a
complete representation it is possible to identify clear and stable boundaries
between self-contained components that mask much of the complexity. These
facets are evident in much of the engineering literature with its focus on detailed
definition of requirements (Petroski 1996).
In organizational studies, such a scientific approach to design is most evident in
the work of Fredrick Taylor and his colleagues (Kanigel 1997). Once the purpose
of a corporation was fixed (for example, to produce widgets with certain predeter-
mined features in as efficient a manner as possible) this approach could be used to
identify the ‘one best way’ to organize. The design process was driven by the need
to completely understand and optimize all the cause-and-effect relationships that
could influence the outcome of an organization’s activities.5Theorizing was done
by specific individuals such as industrial engineers but not by those engaged in
ongoing operations whose job it was to ‘do and not to think’. Decomposability was
manifest in division of labor and the presence of a hierarchy, as well as the use of
interchangeable parts that were tied together through stable interface specifications
within an overall organizational architecture. Costs associated with incompleteness
that arose from interruptions were to be minimized at all costs by keeping the
assembly line running even if there were defects in the system (Kanigel 1997). The
result was the design of the quintessential lean and efficient manufacturing process
that could mass produce goods for pre-set user preferences.
Pragmatic Approach to Design
A scientific approach to design — one that requires complete representation of
the problem and identifies the optimal solution — is based on the assumption
that the environment is stable. For decades, this assumption held. Contexts
within which such designs were deployed changed infrequently, if at all.
Consequently, boundaries and preferences could be specified and stabilized,
and a design with enduring qualities established, to be tweaked as changes in
the environment took place.
However, such an approach is likely to run into problems in environments
characterized by continual change (Barry and Rerup 2006). In such contexts,
system boundaries are often unclear and user preferences are both heteroge-
neous and evolving. As a result, the goals and purpose of the design are likely
to remain a continually moving target (Rindova and Kotha 2001).
The advent of new information technologies has made such fluidity possible.
Different material artifacts and social groups can now be easily associated in a
dynamic network. Rendering the functionality of any product or service through
software enables real-time changes to a design. Indeed, these technologies
make it possible for customers to explore their preferences in use and for dif-
ferent social groups to engage with each other in an emergent fashion.
Simon appreciated the significance of such dynamic situations. For
instance, in a section titled ‘Designing without final goals’, Simon explored a
354 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
paradoxical but pragmatic view of design — to motivate activity that in turn
generates new goals. Offering the example of the rebuilding of Pittsburgh,
where new goals emerged after initial goals had been accomplished, Simon
(1996: 163) concluded:
‘Making complex designs that are implemented over a long period of time and continu-
ally modified in the course of implementation has much in common with painting in oil.
In oil painting, every new spot of pigment laid on the canvas creates some kind of pattern
that provides a continuing source of new ideas to the painter. The painting process is a
process of cyclical interaction between the painter and canvas in which current goals lead
to the new application of paint, while the gradually changing pattern suggests new goals.’
In these observations we see how a design approach need not be a static rep-
resentation of a problem, but can involve a ‘theory-design-fly-test and start-all-
over-again’ methodology (Boland et al. 2006). Despite this acknowledgment,
there has been relatively little work that explores the nature and implications of
such a design approach. Indeed, as Boland (2004: 109) states, ‘Much of
Simon’s concern centers on the local and immediate experience of an individ-
ual who faces an environment that is essentially unchangeable. It is a given to
which the managers must mold the organization.’
The challenges that arise in applying a scientific approach to design in
dynamic environments become all the more apparent when we consider the
nature of change that is upon us. If we accept Woodward’s (1965) powerful
insight that technologies of work shape the way in which we work with tech-
nologies, new information technologies not only link islands of unconnected
activities into an action net (Czarniawska 2004), but, in enabling such connec-
tions, change the very meaning of the term ‘design’ to connote continual evolu-
tion via interaction. Jelinek (2004: 115) fully understood the implications of this
change when she stated:
‘A ge nu in e r evo lu ti on o f p os si bi lity f low ed f rom th e m ov e b et we en pr io r fla t fi le sy st em s
into relational databases and, in computer systems, the hyperlinked nodes of the inter-
net. Similarly, the virtual organization — often pro tem, frequently voluntary, and
broadly distributed — is the iconic organization of our times. But how does one design
it? Or should one perhaps instead invite it to emerge? Is the issue deliberate design, or is
it design of a process of interactions that allows a collaborative process to emerge? …
How shall we theorize about such organizations? … Should we embrace the ephemeral
organization as a new norm?’
An image of an organization that is not a series of nested black boxes (March
and Simon 1958) operating in an immutable environment but, rather, a hyper-
text organization that continues to emerge is a radical shift indeed. The design
problem, then, is not one of developing a static interface (an edge in network
terms) that connects the inside and the outside; rather, it involves the creation of
multiple edges between many nodes in a dynamic network. In such an action
net, each node can potentially act as a boundary object, ‘remaining between dif-
ferent realms, belonging to all of them simultaneously, and seen from different
points of view’ (Czarniawska 2004: 104). Given this, ‘When such an organiza-
tion does emerge, it may be both transient and protean’ (Jelinek 2004: 115).
These observations further highlight the difficulties associated with designing
for completeness in a world that is continually changing. But, what does it mean
Garud et al.: Incomplete by Design and Designing for Incompleteness 355
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
to design for incompleteness? Is this an oxymoron? We think not. There are now
a number of new organizational forms that suggest that incompleteness, rather
than posing a threat, can instead be a virtue. We examine two such cases in this
paper: the Linux operating system and the Wikipedia online encyclopedia. These
cases provide us with an appreciation of the generative nature of incompleteness.
They amplify what Weick (2004: 43) astutely pointed out: ‘Life persists when
designs are underspecified, left incomplete, and retain tension.’
Research Design
Our objective is to offer a set of observations that form the basis for an ongoing
conversation among those interested in understanding the nature of design in
continually changing environments. By no means do we claim to offer a full-
fledged theory: to do so would defeat the very premise of our argument. Here,
we subscribe to the notions proposed by Boland and Collopy (2004), Romme
(2003) and others who suggest that the value of theorizing lies in the options
that are generated rather than the uncertainties that are resolved. Along these
lines, our intent is to sensitize readers to a pragmatic approach to design that
harnesses the generative forces of incompleteness.
The research approach that we adopt in this paper involves a detailed explo-
ration of two exemplary cases. Research on analogical thinking (Loewenstein
et al. 1999) suggests that individuals who are presented with multiple cases
exhibit greater ease in their abilities to identify underlying patterns. Based on
this finding, we decided to provide not one but two in-depth cases. We deliber-
ately chose to examine the Linux operating system and the Wikipedia online
encyclopedia as each represents a design that is continually evolving. We
tracked available information from a wide variety of online sources as a means
to collect raw data for our case narratives. The multiple data sources helped us
‘triangulate’ (Jick 1979) in that there were very few disagreements among the
data sources on the factual details involved. Two of the authors separately devel-
oped the case studies. The individual inferences drawn from each of the cases
were discussed and verified with the other authors. Our aim is not to reach a
state of theoretical saturation (Eisenhardt 1989; Glaser and Strauss 1967).
Rather, much like the phenomena that we are studying, our aim is to offer a set
of observations that will hopefully generate ‘theoretical tension’ and form the
basis for ongoing debate and understanding of this phenomenon.
Linux: Perpetually in the Making
Linux originated as a hobby project of Linus Torvalds, a computer science
student at the University of Helsinki. Initially, he wrote programs to under-
stand how the Minix operating system could use features of the 386 proces-
sor. Soon, he had written task-switching programs, a disk driver and a small
file system. Torvalds realized that he was actually working on a new operat-
ing system, and, as a result Linux 0.01 was born. At this stage, Linux lacked
356 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
many of the functionalities that users expected from an operating system.
Even the numbering of the version (0.01) signified its unfinished status
(Diedrich 2001).
Torv al ds ’ i ni ti al dev el opmen t ef fo rt s ca n be desc ri be d as ‘ di sc ove ri ng d es ig n
goals in action’. Upon releasing Linux version 0.02, he described the initiative
as follows:
‘This is a program for hackers by a hacker. I’ve enjoyed doing it, and somebody might
enjoy looking at it and even modifying it for their own needs. It is still small enough to
understand, use and modify, and I’m looking forward to any comments you might have
… If your efforts are freely distributable (under copyright or even public domain), I’d
like to hear from you, so I can add them to the system … Drop me a line if you are will-
ing to let me use your code.’ (Torvalds, 1991)
His announcement to the newsgroup suggests that he was unsure about what
others might want in the emergent system. Although such lack of closure could
be a problem from a traditional design perspective, it is interesting to observe
how Torvalds turned this into a virtue. His decision to allow others to adapt the
emergent system to their needs enabled him to harness the energies of the larger
programming community. Within two months of his announcement, about 30
people had contributed close to 200 reports of errors and problems using Linux.
In addition, these individuals developed new features for the nascent operating
system that, in turn, became the basis for future directions along which the
system evolved (Moon and Sproull 2000).
Traditionally, there has been a clear distinction between the designer and the
user. However, Torvalds’ actions blurred the boundaries between these two
actors: the user could also be the designer (von Hippel and von Krogh 2003).
By catalyzing such a process, Torvalds ensured that the purpose and function-
alities of Linux would now emerge from the efforts of multiple contributors.
Providing users with an opportunity to inscribe their local contexts into the
design enabled the development and linkage of components in diverse and
sometimes non-obvious ways. As these inputs were incorporated, the very pur-
pose and functionality of the platform itself changed.
Tec hn olo gy is s oci et y ma de d ur abl e, s ugge st ed La tou r (19 91 ), and in t he ca se
of Linux we can recount many instances of how technology made engagement
among contributors possible. Clearly, tools such as mailing lists and news groups
made it possible for actors to engage with the emergent platform and with one
another. Similarly, ‘installers’ that facilitated reuse and distributed modifications
contributed to the growing functionality of the system. For example, the Linux
community experienced a large growth in 1992 when the Yggdrasil distribution
made it possible for users to install Linux from a CD-ROM (Diedrich 2001).
Finally, Torvalds redesigned the Linux kernel to have one common code base
that could simultaneously support a separate tree for different hardware archi-
tectures, thereby greatly improving its portability (Torvalds 1999). Taken
together, these tools made it possible for contributors to continuously extend
Linux in a decentralized manner.
Moreover, social rules built into the technology (Lessig 1999) further fos-
tered generative engagement by actors. A key development was Torvalds’ deci-
sion to release the kernel under General Public License (GPL), which mandated
Garud et al.: Incomplete by Design and Designing for Incompleteness 357
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
that any user modifying or adding to the source code would have to make their
own contributions available to everyone else (Stallman 1999). This decision sig-
naled to the community that ownership and control of Linux was not Torvalds’
alone. Rather, it facilitated the establishment of a stable ‘generalized exchange
system’ (Kollock 1999) in which people both contributed to and benefitted from
the assistance of others. Establishing mechanisms that recognized the contribu-
tions of individuals reinforced the feeling of common ownership. A ‘credits’ file
made available with every released version of Linux listed various contributors
and their roles in developing and maintaining the kernel.
The availability of the source code played a critical part in the actors’ gener-
ative engagement with the platform. Through documentation in repositories
such as the CVS (Concurrent Version System), the source code served as a
design trace that provided a memory of how the platform had evolved until a
specific point in time. While creating options for the future, contributors could
go back to the past to find out how solutions to related problems had been devel-
oped. Moreover, the design trace provided the interpretive flexibility (Pinch and
Bijker 1984) required to recontextualize the existing platform and make it work
in new application areas without losing coherence with the original platform.
Finally, enabling developers to leave their footprint in the source code ensured
that their identities became intertwined with the platform and served as a strong
attractor for them to contribute.
Overall, these social and technical mechanisms made it easier for contribu-
tors to tinker with the system and incorporate their own notions of how it should
be further developed. In doing so, they extended Linux in ways that collectively
covered a far greater domain than any single individual could have imagined.
This manifested itself in the development of a variety of utilities and device dri-
vers that supported specific hardware and peripherals (Thiel 1991). In an inter-
view, Torvalds said about the community’s engagement with the platform:
‘Af te r I h ad p ub li she d Li nu x o n th e w eb , o th er us er s r eq ue st ed fe at ur es I ha d neve r
thought of and more and more new ideas emerged. Instead of a Unix for my own desk-
top, Linux should now become the best operating system ever. The requests by other
people and their patches and help later on made the whole thing more interesting.’
(Diedrich 2000)
Our observations from Linux, then, suggest that incompleteness is generative
in two different ways. At one level, incompleteness serves as a trigger for the cre-
ation of many diverse ideas on how a design can be extended and further devel-
oped. At another level, engagement with such a system both transforms the
design as well as creates new avenues for ongoing engagement which, in turn,
attracts a new set of contributors who bring into the fold their own contextual-
ized needs, purposes and goals. The contribution of a graphical user interface, for
example, spurred a whole new application area called ‘Linux on the desktop’, a
central aspect of current Linux versions (Diedrich 2000). This observation high-
lights a key benefit of designing for incompleteness: generative engagement with
a platform is self-reinforcing in nature, with the boundaries of both its technical
and social architecture co-evolving with one another (see also Neff and Stark
2003). While generative engagement enables developers to cope with incom-
pleteness in the present, it is also the source of incompleteness in the future.
358 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
Such generative engagement by a multitude of users, however, needs to be
channeled to maintain coherence. For instance, it is possible for individuals to
introduce security holes or malicious code intended to cause damage (Singer
2005). But even contributors with the best of intentions can sometimes inad-
vertently bring about damage to the system. More fundamentally, the lack of
control can lead to fragmentation of the platform as factions pursue different
avenues of development. As an illustration, the many different distributions of
Linux (e.g. RedHat Linux, SUSE, or Debian) all come with proprietary patches
when system updates are released. In some cases, there have been so many
changes made to the original package that the distributions have become incom-
patible (Wolf 2001). Given these challenges, what governance mechanisms can
be established to enable incompleteness to be harnessed beneficially?
Linux’s governance approach is best described as ‘centrally facilitated yet
organizationally distributed’ (Garud and Kumaraswamy 2005). A combination
of technical and social rules keeps the platform from falling apart. Ongoing
documentation of the development process, inscribed into the source code and
CVS, serve as a form of cohesive tissue. Chat rooms and bulletin boards act as
threads that order interactions among contributors across space and time
(Lanzara and Morner 2005). These mechanisms are buttressed by an overarch-
ing meritocracy within the community in which developers focus on providing
new code and modules that add features and functionality to the platform while
others, based on their reputation and prior contributions, assume maintenance
tasks such as the evaluation of code submitted by developers for inclusion into
new releases. Sitting atop this meritocracy is Torvalds, who centrally facilitates
decisions on strategic issues related to Linux. Decisions to deliberately change
the platform are entrusted to a much smaller group of individuals when com-
pared to the total number of contributors. Finally, the provisions of the GPL
provide the legal and cultural basis for interaction among community members.
These governance mechanisms facilitate access to knowledge, encourage con-
stant tinkering, and curb private appropriation and free riding. Together, these
mechanisms operate with an almost invisible touch vis-à-vis ongoing develop-
mental activity on the platform. This enables extensive experimentation to take
place on the system even as it maintains a coherent core. By contrast, a tightly
controlled design would bear the risk of falling apart, given the difficulties of
accommodating the contradictory requirements of a heterogeneous community.
A number of Linux’s governance mechanisms have themselves emerged over
time in a relatively unstructured fashion. One such convention pertains to the
successive release of Linux versions even while distributed development pro-
gresses. An even-numbered release (for instance, version 2.4) denotes that the
release is stable and ready for use, whereas an odd-numbered release (for
instance, version 2.5) signifies that the release is still being built and under eval-
uation. The announcement of an odd-numbered release serves as the call for
new code and modules that drives the current stable release forward. After a
period of development, testing, discussion and reviews, no new additions are
allowed so that the entire operating system can be tested for performance and
reliability. Once testing has ended and modules of code that compromise reli-
able performance are removed, the operating system is ready to be released as
Garud et al.: Incomplete by Design and Designing for Incompleteness 359
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
a stable even-numbered version for widespread use. Such a convention has
enabled distributed development.
Overall, our description of Linux highlights what it means to be incomplete
by design. To the extent that the system has the capacity to incorporate func-
tionality relatively easily, it is possible to harness its generative properties.
Blurring the distinction between users and producers, assuming that preferences
are heterogeneous and evolving, and maintaining abstractness in goal definition
often facilitates this generative process. These facets lie in contrast to scientific
approaches to design that favor clear boundary definitions, fixed user prefer-
ences, design closure and platform stability. While systems that are designed for
incompleteness may appear messy by conventional design evaluation metrics,
they can often outperform traditional designs by being extremely adaptable to
continually changing contexts. Adopting such a design approach, then, involves
appreciating design as the interplay between intermediary outcomes and
processes, with one influencing the other on an ongoing basis.
Wikipedia: A Lumpy Work in Progress
Initiated in 2001 by Jimmy Wales and Larry Sanger, Wikipedia’s ambitious
mission has been to ‘create and distribute a free encyclopedia of the highest
possible quality to every single person on the planet in their own language’
(Wales 2005). As of 2006, it had become one of the most visited websites in the
world, with 5 billion page views monthly. Currently, it is growing at around
6,000 edits a day. What is truly remarkable is that Wikipedia has run as a non-
profit organization since 2003 and has five employees in addition to Wales. It
meets most of its budget through donations, the bulk of these being contribu-
tions of $20 or less (Schiff 2006).
The typical encyclopedia, as represented by the Britannica,is the epitome of
a product that is designed to be complete. Its production process involves
assembling a large group of experts, working under the direction of a manager,
each performing a task on a detailed work chart to produce a work of enormous
breadth. This product is then packaged and bound in a set of volumes as an
authoritative and accurate source of knowledge. The encyclopedia, then, sub-
scribes to a scientific approach to design, with a clearly defined purpose, a pro-
duction process that is neatly modularized, clearly delineated boundaries
between producers and users, and an emphasis on stability and reliability.
Contrast this with the inner workings of Wikipedia. While the overarching
goals of this initiative are similar to that of the Britannica,participation in its
development is deliberately inclusive, blurring the boundaries between user and
producer. Any registered user can write an article that others subsequently mod-
ify and refine. This conceptual shift — along with mechanisms that enable gen-
erative engagement — has contributed to a system that has now begun to
fundamentally question the ontological basis of a traditional encyclopedia.
Tec hn ical ly, en tr ies ar e ma de usi ng a w iki, a t ool th at a llow s mult ip le use rs t o
create, edit and hyperlink pages. The word ‘wiki’ comes from the Hawaiian word
for ‘quick’, but also stands for ‘what I know is ...’, and these definitions, taken
360 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
together, provide an essence of the design approach underlying Wikipedia.
Anyone can initiate a new article by making a basic entry, a ‘stub’ in Wikipedia
parlance. If a topic that users are searching for is not covered on the site, then they
are encouraged to write it themselves. Knowledge, from this perspective, is
assumed to reside in a distributed fashion within the entire online human network.
Why might someone contribute to Wikipedia? For many, its goals — a free
encyclopedia that is a fun experience to create, and the open-source principles
on which it is based — are an important draw. This translates into a strong level
of commitment for a subset of individuals who end up spending a considerable
number of hours tending to the project. As a Wikipedian described his attach-
ment to the website, ‘You can create life in there. If you don’t know about some-
thing, you can start an article, and other people can come and feed it, nurture it’
(Pink 2005).
Given the minimal restrictions on membership, the project has turned funda-
mental assumptions about the design of knowledge systems on their head. The
system is truly democratic in that it does not favor experts over the well-read
amateur. As Wales puts it, ‘To me, the key is getting it right. I don’t care if
they’re a high-school kid or a Harvard professor’ (Schiff 2006). Over 100,000
individuals have made contributions to the website since its inception, with a
24-year-old University of Toronto graduate being the site’s premier contributor,
having edited more than 72,000 articles since 2001.
How can a system that is based on the participation of literally any individ-
ual become a useful knowledge source? On Wikipedia, every article has an open
invitation for edits, and the expectation is that most edits will be improvements.
Wikipedia operates from the presumption that any individual’s knowledge is by
definition incomplete, and that ongoing revisions enabled by mass collaboration
tools and involving a large group of ‘eyeballs’ will produce a reliable yet con-
tinually evolving knowledge repository. More fundamentally, it reflects an
appreciation of the inherently accretive nature of knowledge, one in which the
content of any article provides the generative basis for the next set of changes.
As Earl (2005) commented, ‘The philosophy of Wikipedia is that an article
gains validity and maintains currency by continuous widespread updating.
Thus, hitherto it has not declared any item finished.’
The wiki keeps track of all changes made by users and allows them to com-
pare multiple versions of an article.6This information provides a design trace
which allows actors to understand how and why an article has emerged over
time. Possessing such an overarching temporal perspective is critical to repair-
ing damaged elements in an entry as well as reverting to an older version if the
current one is inaccurate. Moreover, this trace often forms the basis for new
directions in which the article is developed. The design trace, then, both chron-
icles and initiates generative engagement with an article. In doing so, it serves
as a locus of coordination as well as a point of departure, allowing an article to
remain in a state of perpetual change.
In addition, social rules embedded in Wikipedia’s website enable further gen-
erative engagement. For example, tags such as ‘The neutrality of this article is
disputed’ represent calls for action to improve an article. Disambiguation notices
placed at the bottom of certain articles help readers accurately locate information
Garud et al.: Incomplete by Design and Designing for Incompleteness 361
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
that they are looking for. Finally, the creation of an ontological scheme via
categories aids contributors in organizing information available on the site. The
categorization of entries itself facilitates a process of adaptive structuration
(DeSanctis and Poole 1994) in which the contribution of articles with different
content change the meaning of a category that subsequently gets relabeled.
These dynamics highlights how the community structures Wikipedia in use.
An interesting by-product of such an evolving knowledge project is its abil-
ity to instantaneously respond to current events. When the Indian Ocean
tsunami erupted in late 2004, contributors produced several entries on the topic
within hours. By contrast, the Wo rl d B oo k,whose CD-ROM allows owners to
download regular updates, had not updated its tsunami or Indian Ocean entries
a full month after the devastation occurred (Pink 2005). This ability to provide
instant information is a metric on which a knowledge repository organized to
change perpetually outperforms a traditional encyclopedia.
The flip side of this design approach, however, is that it produces entries that
are often amateurish. While the facts may be sturdy, clarity and concision is
sometimes lacking. The initial contributor to an article can set its tone and may
not necessarily be highly knowledgeable in the area (Wattenberg and Viegas
2006). Disagreements on an article can lead to repeated back-and-forth editing,
with the user who spends the most time on the site prevailing. Given the obses-
sion of many users to rack up edits, simple fixes often take priority over more
complex edits. Moreover, the open access to the system has produced inci-
dences of vandalism on the site that involve individuals inserting obscenities
within entries. Given these shortcomings, what governance mechanisms ensure
that articles do not lose their integrity?
Here again, a mix of technical and social elements work together to maintain
the integrity of the website. Five webbots continually troll the site, searching for
obscenities and evidence of mass deletions and reverting text as they go. Any
editing on an article is automatically logged on a ‘recent changes’ page that par-
ticipants can monitor. Anyone who sees something false, biased or otherwise
defective can quickly and easily change the text. Moreover, every article also
carries with it a separate discussion section on which debates about what to
include on the page are encouraged. Besides this, users employ internet relay
chat to discuss ongoing issues, from article details to general policy.
An underlying community of volunteers holds these distributed contributions
together. This includes anonymous contributors, people who make a few edits.
Next, there are registered users who make edits using their by-line. Administrators
are individuals who can delete articles, protect pages and block IP addresses.
Finally, there are the super elites who make direct changes to the Wikipedia soft-
ware and database. There is also a mediation committee and an arbitration com-
mittee that rule on disputes. On this front, the inner circle of Wikipedians know
each other and value their reputations, which are themselves built bottom-up
through their past activity on the site. The consequence of such reputation-build-
ing has been the creation of a meritocracy with individuals occupying more cen-
tral positions because of their ongoing involvement when compared to others.
In terms of policies, the founders of Wikipedia have instituted a NPOV (neu-
tral point of view) rule, which urges contributors to present conventionally
362 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
acknowledged ‘facts’ in an unbiased way as well as accord space to both sides
when arguments occur (Poe 2006). Over time, Wikipedia has instituted additional
rules, reflecting the complexities involved in managing a growing decentralized
community. For example, a policy preventing certain contentious subject matters
from being openly edited by users was ratified recently (FinancialWire 2006).
This has prompted concern that Wikipedia is becoming a regulatory thicket
complete with an elaborate hierarchy of users and policies. Wales, while ambiva-
lent about the growing number of rules and procedures, recognizes them as nec-
essary. According to him, ‘Things work well when a group of people know each
other, and things break down when its a bunch of random people interacting’
(Schiff 2006). And Earl (2005) captured Wikipedia’s dilemma as follows:
‘This is a classic knowledge-creation conundrum. On the one hand there is real advan-
tage in assembling collective wisdom; on the other hand there are concerns about valid-
ity and about incentives or, more particularly, disincentives for content contributors.’
However, taking these steps raises an even more fundamental issue: should
Wikipedia abandon its current design approach in order to emulate a traditional
encyclopedia? For many, the incomplete nature of an article is valuable in its
own right. First, such an article is better than nothing at all. Second, the articles
might actually trigger more experienced participants to make contributions that
they would otherwise not have made. Instituting governance mechanisms that
make it appear more like an encyclopedia could potentially rob Wikipedia of its
unique identity and impede generative engagement by contributors. As Carr
(2005) elaborated:
‘Wikipedia is not an authoritative encyclopedia, and it should stop trying to be one. It’s
a free-for-all, a rumble-tumble forum where interested people can get together in never-
ending, circular conversations and debates about what things mean. Maybe those dis-
cussions will resolve themselves into something like the truth. Maybe they won’t. Who
cares? As soon as you strip away the need to be like an encyclopedia and to be judged
like an encyclopedia — as soon as you stop posing as an encyclopedia — you get your
freedom back.’
This observation underscores the value of incompleteness. Incompleteness
allows participants the freedom to engage with the design in a way that is mean-
ingful to them. As the quote suggests, to the extent that the goals and metrics of
traditional approaches that emphasize completeness are adopted in evaluating
designs that are inherently incomplete, there exists a danger of losing the unique
value that such designs can provide.
Discussion
The Linux and Wikipedia cases affirm what Boisvert (who built upon Deweys’
(1934) pragmatic approach) pointed out: ‘Affairs are never frozen, finished, or
complete. They form a world characterized by genuine contingency and con-
tinual process. A world of affairs is a world of actualities open to a variety of
possibilities’ (Boisvert 1998: 24). Indeed, these cases provide a deeper appreci-
ation of the title of the paper: incomplete by design and designing for incom-
pleteness. In an environment that is continually changing, designs that have
Garud et al.: Incomplete by Design and Designing for Incompleteness 363
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
been completed at a point in time are likely to become incomplete over time.
On the other hand, designs that anticipate their incompleteness are likely to
remain more up to date over time.
Design Participation within the Pragmatic Approach
The traditional scientific approach employed principles from the natural world
to design an artifact with enduring qualities that fulfilled a specific purpose in
an unchanging world (Simon 1996). From this perspective, a design was fixed
in time and space; it was opened and modified only to accommodate exogenous
environmental changes. Moreover, the locus of a design, i.e. the demarcation
between designer and user, was clear and unambiguous.
In contemporary environments, however, the distinction between designers
and users has blurred, resulting in the formation of a community of co-design-
ers who inscribe their own contexts into the emergent design, thereby extend-
ing it on an ongoing basis in diverse and non-obvious ways. Such generative
engagement by multiple co-designers is facilitated by numerous sociotechnical
mechanisms. Tools such as the wiki, licenses such as the GPL, forums such as
bulletin boards and the infrastructure provided by the internet work with one
another to facilitate participation and enable distributed development. This
dynamic action net (Czarniawska 2004), then, contributes to the design remain-
ing in a fluid state.
We c an f ur th er c on tr as t de si gn p ar ti ci pa ti on wit hi n th e sc ie nt if ic a nd pra gm at ic
approaches by returning to the parable of the watchmakers offered by Simon. In
its original version, user engagement with the design was considered an interrup-
tion resulting in watches that remained incomplete and therefore of little value.
By contrast, the Linux and Wikipedia cases demonstrate that incompleteness acts
as a trigger for generative engagement by co-designers. They are the ones who
complete what they perceive is incomplete. They discover the purpose of a design
in use. They create avenues for future development that, in turn, attract new
groups of co-designers. From a pragmatic design approach, what was considered
to be an interruption then, now becomes the basis for ongoing change.
Design Task within the Pragmatic Approach
For co-designers, the design task is very different from the one faced by
designers adopting a scientific approach. For the latter, optimization of an
objective function given constraints (as in linear programming), represented
the dominant approach to designing. By contrast, contemporary designs such
as Linux and Wikipedia can be conceptualized as an interlinked set of subjec-
tive functions, where one person’s subjective function serves as another
person’s constraints. Complicating matters, the set of interlinked subjective
functions is itself underspecified as it emerges in use over time. Under these
conditions, the design remains incomplete as the solution to the optimization
problem corresponds to more than one point in an n-dimensional space of
design parameters. It is for this reason that co-designers must learn to theorize
on the fly: i.e. they become reflective practitioners (Schön 1983) who generate
364 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
provisional workable solutions to their immediate problems, knowing fully
well that the platform that they draw upon and the context to which they apply
their solutions will inevitably change.
What is the role of modularity in such an emergent system? Modular ele-
ments allow for platform extension as is apparent in both the cases we studied.
Under these conditions, extension of the platform takes place through an accre-
tive process where modules are tacked on as ‘plug-ins’ (Garud et al. 1998). But
the two cases also demonstrate a less understood, generative facet of incom-
pleteness that is fostered by ‘partial partitioning’. On this front, we suggest that
interdependence between partitioned tasks results in situations where changes
in one task have a cascading effect on other tasks. Consequently, the system
never comes to a rest.
Modularity scholars have suggested that partial partitioning can be a problem
(Ethiraj and Levinthal 2004). The Linux and Wikipedia cases, however, indicate
that there may be benefits to be harnessed from partial partitioning. The ‘shared
division of labor’ (Nonaka and Takeuchi 1995) that emerges allows for redun-
dancy of functions, i.e. certain functions can be fulfilled by more than one com-
ponent. This enables the system to work even if some components are damaged
or left incomplete. More significantly, such redundancy of functions can be gen-
erative in that components can be deployed for a different purpose, thereby facil-
itating reconfiguration of the design in response to changes in the environment
(Garud and Kotha 1994). It also allows the system to more easily assimilate
emergent contributions from co-designers into a continually evolving platform.
Design Governance within the Pragmatic Approach
Given the distributed, emergent and protean nature of designs, the challenge
now becomes one of establishing rules that provide some stability. While an
absence of rules is likely to lead to design fragmentation, too many rules can
potentially stifle the design. It is this paradox that needs to be managed to pre-
serve the value proposition that such designs offer.
An appropriate form of governance is required to coordinate real-time dis-
tributed contributions in a way that preserves the design’s dynamic qualities, i.e.
one which allows elements of a system to inform but not determine one another
(Barry and Rerup 2006: 267). Governance mechanisms need to be underspeci-
fied (Weick et al. 1999) or semi-structured (Brown and Eisenhardt 1997); that
is, they possess minimum critical specifications (Emery 1980) to keep the
design in a state that is neither too fluid nor too crystallized (see Gehry 2004 for
this distinction).
The design trace is a key element that enables such governance. By provid-
ing widespread access to knowledge on who contributed what, when and why,
the trace makes it easier for actors to understand how a design has emerged
over time. Possessing such an overarching temporal perspective is often criti-
cal in mitigating design fragmentation. For instance, if a new contribution
turns out to be damaged or incompatible, then the trace makes it possible to
simply use an older version. Equally important, the trace can be viewed as a
boundary architecture (Star and Griesemer 1989) that co-designers draw on to
Garud et al.: Incomplete by Design and Designing for Incompleteness 365
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
develop extensions to the design. In sum, the ongoing design trace serves as a
locus of coordination as well as a point of departure for such designs.
To fu rt he r ela bo ra te , c on si der tw o diff er en t lo gi cs of en ga ge me nt des cr ib ed
by Bruner (1986). A causal logic, following the role model of the natural sci-
ences, operates when the set of variables and relationships that define the func-
tioning of a design are fully specified (see Romme 2003 for a more detailed
explication). This logic, associated with a scientific approach to design, leads to
the articulation of design rules (Baldwin and Clark 2000; Romme 2003). A nar-
rative logic, on the other hand, operates on the basis of a narrative’s internal
coherence and its external coherence with the listener’s existing knowledge. In
providing designers with the interpretive flexibility required to generate con-
textualized solutions and to imagine what might be, the narrative can help coor-
dinate distributed activities across time and space. Such an epistemology
connects designs with the emergent purposes of the social groups involved.7
The design trace allows for these two different logics to operate simultane-
ously. The trace possesses a scientific logic that offers co-designers with the rai-
son d’être of the design and the necessary details for it to be functional in real
time. From this perspective, each node of a trace is but a module with tags that
can be opened up and reused. At the same time, the trace also allows for a nar-
rative logic to operate. Co-designers are motivated to participate with a design
because of the flexibility that it offers, and the design in use that emerges is con-
vincing to these participants as the narratives recorded have verisimilitude
(Bruner 1986). A design trace, then, possesses the equivalents of both an osten-
sive and a performative dimension (Feldman and Pentland 2003). By providing
connections to assets across time and space, a design trace makes it possible for
co-designers to engage in the present by building upon the past in anticipation
of a new future (Ricoeur 1984).
In enabling the two logics to operate simultaneously, the trace is able to alle-
viate some of the problems associated with human memory and bounded ratio-
nality that Simon considered in his parable. Tempus and Hora were boundedly
rational individuals, relying on their own short- and long-term memories. A
mechanism such as the design trace could potentially have extended their mem-
ories by serving as a collective mind (Weick and Roberts 1993). As a narrative,
the trace makes it easier for individuals to recontextualize the design to their
own situations. This makes it possible for interruptions not only to be wel-
comed, but also to serve as the basis for design extension.
In offering these observations we see how future options on a design are gen-
erated even as current contributions are incorporated. Moreover, recording
changes in the design trace signals to co-designers that their contributions will
be honored in the future. If a trace was not to exist, part of the motivation to par-
ticipate might be lost: why contribute to a commons when the contributions
would not be used in the future? At the same time, if the co-designers were
over-dependent on the extant trace, then the future design would largely be
based on past experiences and become path-dependent. For a design (and its
trace) to promote diachronic processes, then, anticipations of the future and
memories of the past must jointly inform contributions in the present.
366 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
Conclusion
We h av e explo re d th e Li nu x an d Wi ki pe di a ca ses t o sk et ch o ut t he e le me nts o f a
pragmatic approach to design. In continually changing environments, adopting a
design approach that attempts to fix boundaries, goals and purposes is potentially
counterproductive. While such an approach may produce a system that is opti-
mal at a point in time, given continual change the system is likely to rapidly
become obsolete over time. Under these conditions, a pragmatic approach (one
that views design as continually evolving and essentially incomplete) may be
more appropriate. Within such an approach, boundaries between designers and
users become blurred, heterogeneous user preferences emerge in use, tasks
remain partially partitioned and the goals of the design emerge through interac-
tion. Indeed, such an approach harnesses the benefits of incompleteness in com-
parison to the scientific approach that views incompleteness as a threat.
Eventually, a pragmatic approach involves the fusing together of two mean-
ings of design, that is, as both process and as outcome. Any outcome is but an
intermediate step in an ongoing journey, representing both the completion of a
process as well as its beginning. Whereas the scientific approach emphasizes
the need to crystallize designs, the pragmatic approach highlights the value of
retaining fluidity. The essence of this approach is well captured by Hedberg
et al. (1976: 43) who noted, ‘Designs can themselves be conceived as processes
— as generators of dynamic sequences of solutions, in which attempted solu-
tions induce new solutions and attempted designs trigger new designs.’
Some of the ideas in this paper are based on prior discussions with Arun Kumaraswamy. We thank
him. We also thank the three reviewers of this paper and the editors of this Special Issue for their
valuable feedback.
1In this paper,we make a distinction between a ‘scientific approach’to design that applies ana-
lytic thinking to address clearly defined problems to discover an optimal solution (following
the natural sciences as a role model) and a ‘pragmatic approach’ that applies synthetic think-
ing to address ill-structured problems. Others have used the terms ‘science’ vs ‘design’
(Romme 2003) and ‘decision science’ vs ‘design science’ (Boland and Collopy 2004) to make
such a distinction.
2Dewey’s (1934) approach to pragmatism informed our understanding of design as noun and verb,
an understanding reinforced by the entry on design in Wikipedia (http://en.wikipedia.org/wiki/
Design).
3This is derived from a structurational perspective (Giddens 1979) where structure is both
medium and outcome of action.
4Specifying the boundaries is by no means a trivial task. Alexander’s insightful analysis (1964)
of the redesign of a tea kettle illustrates this point. At first blush, such a redesign seems sim-
ple, given that the kettle is a clearly defined object and the boundaries between the kettle and
its environment are obvious. However, Alexander goes on to demonstrate that by changing the
nature of the problem to be addressed — for instance, by asking if it is the method of heating
kettles rather than the kettle itself that needs redesigning — the solution obtained can change
drastically. In reframing the question, the kettle becomes part of the outer environment and the
stove becomes the inner environment.
5See Boland and Collopy (2004) for the use of linear programming methods in inventory con-
trol, which is based on the maximization of a clearly specified objective function given a set of
pre-specified constraints.
6We invite the reader to experiment with the design trace for the Wikipedia entry on ‘Design’
http://en.wikipedia.org/w/index.php?title=Design&action=history.
7We thank Georges Romme for this insight.
Garud et al.: Incomplete by Design and Designing for Incompleteness 367
Notes
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
Alexander, C.
1964 Notes on the synthesis of form.
Harvard University Press, Cambridge,
MA.
Baldwin, C. Y., and K. B. Clark
2000 Design rules: The power of
modularity. Cambridge, MA: MIT
Press.
Barry, D., and C. Rerup
2006 ‘Going mobile: Aesthetic design
considerations from Calder and the
constructivists’. Organization Science
17/2: 262–276.
Boisvert, R. D.
1998 John Dewey: Rethinking our time.
Albany, NY: State University of New
Yo rk P r e ss .
Boland, R. J.
2004 ‘Design in the punctuation of
management action’ in Managing as
designing. R. J. Boland and F.
Collopy (eds), 106–120. Stanford,
CA: Stanford University Press.
Boland, R. J. and F. Collopy, editors
2004 Managing as designing. Stanford,
CA: Stanford University Press.
Boland, R. J., M. Jelinek, and A. G. L.
Romme
2006 Call for Papers: Organization Studies
as a Science of Design. Organization
Studies.
Brown, S. L., and K. M. Eisenhardt
1997 ‘The art of continuous change:
Linking complexity theory and time-
paced evolution in relentlessly
shifting organizations’.
Administrative Science Quarterly
42: 1–34.
Bruner, J. S.
1986 Actual minds, possible words.
Cambridge, MA: Harvard University
Press.
Carr, N.
2005 ‘Let Wikipedia be Wikipedia’
(http://www.roughtype.com/archives/
2005/12/let_wikipedia_b.php).
Czarniawska, B.
2004 ‘Management as designing for an
action net’ in Managing as designing.
R. J. Boland and F. Collopy (eds),
102–105. Stanford, CA: Stanford
University Press.
DeSanctis, G., and M. S. Poole
1994 ‘Capturing the complexity in
advanced technology in use: Adaptive
structuration theory’. Organization
Science 5/2: 121–147.
Dewey, J.
1934 Art as experience. New York: Minton
Balch & Company.
Diedrich, O.
2000 ‘Was Anwender tun, ist niemals
falsch’ c’t, 90.
Diedrich, O.
2001 ‘Happy Birthday, Tux!’c’t, 162.
Earl, M.
2005 ‘Wikipedia’s struggle to govern a
knowledge democracy’. Financ ial
Times, London.
Eisenhardt, K. M.
1989 ‘Building theories from case study
research’. Academy of Management
Review 14/4: 532–550.
Emery, F.
1980 ‘Designing socio-technical systems
for “greenfield’ sites”. Journal of
Occupational Behaviour 1: 19–27.
Ethiraj, S. K., and D. Levinthal
2004 ‘Modularity and innovation in
complex systems’. Management
Science 50/2: 159–173.
Feldman, M. S., and B. T. Pentland
2003 ‘Reconceptualizing organizational
routines as a source of flexibility
and change’. Administrative
Science Quarterly 48/1: 94–118.
FinancialWire
2006 ‘Wikipedia to change revision
policy’. June 19 2006. Forest Hills,
MD: FinancialWire.
Ford, H., and S. Crowther
1922 My life and work. Garden City,
NY: Doubleday, Page &
Company.
Garud, R., S. Jain, and C. Phelps
1998 ‘Technological linkages and
transience in network fields: New
competitive realities’ in Advances in
strategic management, Vol. 14, J.
Baum (ed.), 205–237. Greenwich,
CT: JAI Press.
Garud, R., and S. Kotha
1994 ‘Using the brain as a metaphor to
model flexible production systems’.
Academy of Management Review
19/4: 671–698.
Garud, R., and A. Kumaraswamy
2005 ‘Vicious and virtuous circles in the
management of knowledge: The case
368 Organization Studies 29(03)
References
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
of Infosys Technologies’. MIS
Quarterly 29/1: 9–33.
Gehry, F. O.
2004 ‘Reflections on designing and
architectural practice’ in Managing as
designing. R. J. Boland and F.
Collopy (eds), 19–35. Stanford, CA:
Stanford University Press.
Giddens, A.
1979 Central problems in social theory:
Action, structure, and contradiction
in social analysis. Berkeley, CA:
University of California Press.
Glaser, B. G. and A. L. Strauss
1967 The discovery of grounded theory:
strategies for qualitative research.
Hawthorne, NY: Aldine de
Gruyter.
Hedberg, B. L., P. C. Nystrom, and W. H.
Starbuck
1976 ‘Camping on seesaws: Prescriptions
for a self-designing organization’.
Administrative Science Quarterly
21: 41–65.
Jelinek, M.
2004 ‘Managing design, designing
management’ in Managing as
designing. R. J. Boland and F.
Collopy (eds), 113–120. Stanford,
CA: Stanford University Press.
Jick, T. D.
1979 ‘Mixing qualitative and quantitative
methods: Triangulation in action’.
Administrative Science Quarterly
24/4: 602–611.
Kanigel, R.
1997 The one best way: Frederick Winslow
Tay l or an d t he en ig ma of ef fi ci e nc y.
New York: Viking.
Kollock, P.
1999 ‘The economies of online
cooperation: Gifts and public goods
in cyberspace’ in Communities in
cyberspace. M. A. Smith and
P. Kollock (eds), 220–239. London:
Routledge.
Langlois, R. N.
2002 ‘Modularity in technology and
organization’. Journal of Economic
Behavior & Organization 49: 19–37.
Langlois, R. N., and P. L. Robertson
1992 ‘Networks and innovation in a
modular system: Lessons from the
microcomputer and stereo component
industries’. Research Policy 21/4:
297–313.
Lanzara, G. F., and M. Morner
2005 ‘Artefacts rule! How organizing
happens in open source software
projects’ in Actor-Network-Theory
and Organizing. B. Czarniawska,
and T. Hernes (eds), 67–90.
Copenhagen: Copenhagen Business
School Press.
Latour, B.
1991 ‘Technology is society made durable’
in A sociology of monsters: Essays on
power, technology and domination.
J. Law (ed.), 103–131. London:
Routledge.
Lessig, L.
1999 Code and other laws of cyberspace.
New York: Basic Books.
Loewenstein, J., L. Thompson, and
D. Gentner
1999 ‘Analogical encoding facilitates
knowledge transfer in negotiation’.
Psychonomic Bulletin and Review
6/4: 586–597.
March, J. G., and H. A. Simon
1958 Organizations. New York: Wiley.
Moon, J. Y., and L. Sproull
2000 ‘Essence of distributed work: The
case of the Linux kernel’. Firs t
Monday 5/11.
Neff, G., and D. Stark
2003 ‘Permanently beta: Responsive
organization in the internet era’ in
The internet and American life.
P. E. N. Howard and S. Jones (eds).
Thousand Oaks, CA: Sage.
Nonaka, I., and H. Takeuchi
1995 The Knowledge-creating
company. New York: Oxford
University Press.
Parnas, D.
1972 ‘On the criteria to be used in
decomposing systems into modules’.
Communications of the ACM 15/12:
1053–1058.
Petroski, H.
1996 Invention by design: How engineers
get from thought to thing. Cambridge,
MA: Harvard University Press.
Pinch, T. J., and W. E. Bijker
1984 ‘The social construction of facts and
artefacts: or How the sociology of
science and the sociology of
technology might benefit from each
other’. Social Studies of Science
14/3: 399–441.
Garud et al.: Incomplete by Design and Designing for Incompleteness 369
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
Pink, D. H.
2005 ‘The Book Stops Here’. Wire d
13.03.2006
(http://www.wired.com/wired/archive
/13.03/wiki.html).
Poe, M.
2006 ‘The Hive’. The Atlantic Monthly
September 2006: 86–96.
Ricoeur, P.
1984 Time and narrative. Chicago:
University of Chicago Press.
Rindova, V. P., and S. Kotha
2001 Continuous ‘morphing’: Competing
through dynamic capabilities, form,
and function. Academy of
Management Journal 44/6:
1263–1280.
Romme, A. G. L
2003 ‘Making a difference: Organization
as design’. Organization Science
14/5: 558–573.
Rosenberg, N.
1982 Inside the Black Box: Technology and
economics. Cambridge: Cambridge
University Press.
Schiff, S.
2006 ‘Know It All’. The New Yorker 31
July 2006
(http://www.newyorker.com/archive/2
006/07/31/060731fa_fact).
Schön, D. A.
1983 The reflective practitioner: How
professionals think in action. New
Yo rk : B a s ic B o ok s .
Simon, H. A.
1962 ‘The architecture of complexity’.
Proceedings of the American
Philosophical Society 106/6:
467–482.
Simon, H. A.
1996 The sciences of the artificial, 3rd edn.
Cambridge, MA: MIT Press.
Singer, M.
2005 ‘Linux creator: A little fragmentation
is good’. Internetnews.com
(http://www.internetnews.com/dev-
news/article.php/3467241).
Stallman, R.
1999 ‘The GNU Operating System and the
free software movement’ in Open
sources: Voices from the open source
revolution. C. DiBona, S. Ockman,
and M. Stone (eds), 53–70.
Sebastopol, CA: O’Reilly &
Associates.
Star, S. L., and J. R. Griesemer
1989 ‘Institutional ecology, “translations”
and boundary objects: Amateurs
and professionals in Berkeley’s
Museum of Vertebrate Zoology,
1907–39’. Social Studies of Science
19/3: 387–420.
Thiel, W.
1991 ‘Keyboard.S with German
Keyboard Linux-Activists’
(ftp://tsx-11.mit.edu/pub/linux).
Tor va ld s, L.
1991 ‘Free Minix-like Kernel Sources
for 386-AT’. comp.os.minix
(http://listserv.nodak.edu/cgi-
bin/wa.exe?A2=ind9110a&L=minix-
l&T=0&P=8303).
Tor va ld s, L.
1999 ‘The Linux edge’ in Open sources:
Vo ic e s f ro m th e o p en s o u rc e
revolution. C. DiBona, S. Ockman,
and M. Stone (eds), 101–111.
Sebastopol, CA: O’Reilly &
Associates.
von Hippel, E., and G. von Krogh
2003 ‘Open source software and the
“private-collective” innovation
model: Issues for organization
science’. Organization Science
14/2: 209–223.
Wal es , J.
2005 ‘Wikipedia is an encyclopedia’
(http://mail.wikipedia.org/
pipermail/wikipedia-l/2005-
March/038102.html).
Wat te nb er g, M. , an d F. B. Vi ega s
2006 ‘The hive mind ain’t what it used
to be’. Edge (http://www.edge.
org/discourse/digital_maoism.html).
Wei ck , K. E .
2004 ‘Rethinking organizational design’
in Managing as designing. R. J.
Boland and F. Collopy (eds),
36–53. Stanford, CA: Stanford
University Press.
Wei ck , K. E ., an d K . H. R ob er ts
1993 ‘Collective mind in organizations:
Heedful interrelating on flight
decks’. Administrative Science
Quarterly 38: 357–381.
Wei ck , K. E ., K . M. S ut cl if fe , an d D. O bs tf el d
1999 ‘Organizing for high reliability:
Processes of collective mindfulness’ in
Research in organizational behavior.
R. I. Sutton, B. M. Staw (eds),
81–123. Stamford, CT: JAI Press.
370 Organization Studies 29(03)
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
Raghu Garud is Alvin H. Clemens Professor of Management and Organization and the
Research Director of the Farrell Center for Corporate Innovation and Entrepreneurship,
Pennsylvania State University. Before joining the Smeal College, Raghu Garud was at
the Stern School of Business, New York University. Raghu earned a PhD degree in
strategic management and organization from the University of Minnesota. Raghu’s
research currently explores the emergence of novelty. Specifically, he is interested
in understanding how new ideas emerge, are valued, and become commercialized. He is
a PI on an NSF-funded project titled ‘The architecture of collaboration in scientific
communities’.
Address:Farrell Center for Corporate Innovation and Entrepreneurship,Smeal College
of Business, The Pennsylvania State University, 451 Business Building, University Park,
PA 16802, USA.
Email: rug14@psu.edu
Sanjay Jain is an assistant professor of management and human resources and a princi-
pal faculty member of INSITE (Initiative for Studies in Technology Entrepreneurship)
at the University of Wisconsin–Madison. He received his PhD from the Stern School of
Business at New York University. His current research focusses on technology strategy,
institutional entrepreneurship, innovation management and university technology com-
mercialization. His work has been published in the Academy of Management Journal
and Industrial and Corporate Change and covers such technologies as Java and embry-
onic stem cells.
Address:4271 Grainger Hall,975 University Avenue,Madison,WI 53706,USA.
Email:sjain@bus.wisc.edu
Philipp Tuertscher is a researcher at the E&I Institute for Entrepreneurship and
Innovation at the Vienna University of Economics and Business Administration, Austria,
and PhD candidate at University of St Gallen, Switzerland. Before joining E&I, Philipp
was a visiting researcher at the Farrell Center for Corporate Innovation and
Entrepreneurship, Pennsylvania State University. His current research explores how
technological architectures emerge in complex innovation projects. He is particularly
interested in the role of technological controversies and how these get resolved as dif-
ferent groups in a project interact with each other.
Address:Vienna University of Economics and Business Administration,Augasse 2-6,
A-1090 Vienna, Austria.
Email:philipp.tuertscher@wu-wien.ac.at
Wolf, C.
2001 The ROCK Linux Philosophy
(www.linuxdevcenter.
com)
Woo dw ar d, J.
1965 Industrial organization: Theory and
practice. New York: Oxford
University Press.
Garud et al.: Incomplete by Design and Designing for Incompleteness 371
Raghu Garud
Sanjay Jain
Philipp
Tuer tsc he r
© 2008 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.
by Raghu Garud on April 24, 2008 http://oss.sagepub.comDownloaded from
