Transactive Memory System in Clusters
The Knowledge Management Platform Experience
Pascal Amandine
and Thomas Catherine
Aix Marseille Université, CNRS, LEST UMR 7317, 13626, Aix-en-Provence, France
Université Nice Sophia Antipolis, CNRS, GREDEG, UMR 7321,
250, rue Albert Einstein- Sophia Antipolis 06050 Valbonne, France
Keywords: Transactive Memory System, Cluster, Knowledge Creation, Information Technology.
Abstract: Clusters produce social and cognitive proximities that support knowledge flows and combination. As such,
clusters affect both members’ motivation to engage in collective knowledge creation processes and their
ability to identify actors’ expertise and then to exchange and combine distributed pieces of knowledge. In
other words, social and cognitive interactions within a cluster should affect the development of Transactive
Memory Systems (TMS). This paper proposes to extend the TMS concept at the cluster level. Based on an
integrative design science methodology, this paper makes theoretical contributions on how a TMS within a
cluster functions with the assistance of social interactions and information technologies. The study builds on
the design of a semantic web service of competencies within the Sophia Antipolis telecom cluster. This
study provides empirical support to the potential benefits of the TMS approach at the cluster level and
specifically the ability of an IT to support the development of an effective TMS in a cluster.
A vast literature emphasises the crucial role of
cluster, defined as localized network, in building
learning and innovative capabilities (Kogut, 2000;
Maskell and Lorenzen, 2004; Nooteboom, 2005;
among others). From a Schumpeterian perspective,
Nelson and Winter (1982) treat innovation as a
search process that explores the space of possible
combinations of pieces of knowledge to create new
or better alternatives. Cluster and network favour
interactions among actors, which in turn produce
social and cognitive proximities supporting
knowledge exchange and combination (Kogut, 2000;
Maskell and Lorenzen, 2004). As such, clusters
affect both members’ motivation to engage in
collective knowledge creation processes and their
ability to identify actors’ expertise and then to
exchange and combine distributed peaces of
knowledge. In other words, social and cognitive
interactions within a cluster should affect the
development of Transactive Memory Systems
We propose to extend the notion of TMS at the
cluster level. TMS refers to a collective system that
individuals in closed relationships use to encode,
store, and retrieve knowledge about different
substantive domains (Wegner, 1987). A commonly
used of TMS is a shared system that provides hints
about “who knows what” (Ren and Argote, 2011).
Whereas TMS were originally observed in small
groups, some recent researches wondered if the
TMS concept would “scale up” to fit well within
organizational settings (Nevo and Wand, 2005;
Jackson and Klobas, 2008). If the TMS concept
shows promise for being generalized to the
organizational level, our understanding of how an
organizational TMS functions remains limited (Ren
and Argote, 2011). In addition, extending the TMS
at the organizational level introduces the question of
how IT can support the development of effective
TMS (Nevo and Wand, 2005).
Based on an integrative design science
methodology, this paper makes theoretical
contributions on how a TMS within a cluster
functions with the assistance of social interactions
and information technologies. The study is based on
the Knowledge Management Platform (KMP)
project within the Sophia Antipolis telecom cluster,
one of the main European centres of high tech. We
explore technological interventions to assist clusters’
members in building R&D collaborative projects
Amandine P. and Catherine T..
Transactive Memory System in Clusters - The Knowledge Management Platform Experience.
DOI: 10.5220/0005026200050014
In Proceedings of the International Conference on Knowledge Management and Information Sharing (KMIS-2014), pages 5-14
ISBN: 978-989-758-050-5
2014 SCITEPRESS (Science and Technology Publications, Lda.)
through the design of a map of competencies. The
KMP project is an experimental IT infrastructure
based on a semantic web service of competencies.
The KMP experience provides empirical support
to the potential benefits of the TMS approach at the
cluster level. Our study provides specific guidelines
to create an IT to support the development of an
effective TMS in a cluster. Based on an appropriate
formalization of knowledge about “who knows
what” (the map of competencies), this IT creates an
artificial directory to link different organisations and
subgroups in order to facilitate knowledge exchange
and combination.
The argument in this paper is organized as
follows. First, we articulate the literature on
knowledge creation within a cluster with TMS
discourse. Second, we describe the methodology
adopted in this paper. The argument then turns to a
case study of designing and developing a map of
competencies for fostering knowledge creation in a
telecom cluster. Finally, the findings from this case
study are discussed and implications for future
research are explored.
2.1 Knowledge Creation inside a
According to Kogut and Zander (1992), Nahapiet
and Ghoshal (1998) and Shawney and Prandelli
(2000), organizational knowledge creation is above
all a social process. Moran and Ghoshal (1996) and
Nahapiet and Ghoshal (1998) enhance in a
schumpeterian perspective that organizational
knowledge creation is based on two key
mechanisms: exchange and combination. Creating
new knowledge therefore requires combining
elements previously unconnected or developing
novel ways of combining elements previously
associated. When various agents hold resources,
exchange is a prerequisite for resources
combination. Nahapiet and Ghoshal (1998)
identified four conditions that would favour
knowledge exchange and combination: 1) the
opportunity to engage in exchange and / or
combination; 2) the capability to anticipate
knowledge combination value; 3) the motivation to
engage in exchanging and combining knowledge
and 4) the capability to combine knowledge.
This approach of social knowledge creation
emphasizes the necessity for organizations to open
themselves to the outside in order to reach new
knowledge (Van de Ven, 2005). The network then
represents a privileged source of knowledge
exchange, and provides structures and stability that
can be used for collective learning (Kogut, 2000).
Hence, the network appears like an organizational
configuration able to create, accumulate and transfer
collective knowledge (Baskerville and Dulipovici,
2006). Moreover, cluster capacity to formulate
collective entrepreneurial projects becomes
nowadays crucial (Crevoisier and Jeannerat, 2009).
This gives way to combinatorial territorial dynamics
that are mainly based on the anchoring of composite
fields of knowledge (Antonelli and Calderini, 2008).
Beyond that, the key factor is the capacity to act
collectively (Crevoisier and Jeannerat, 2009). These
capabilities emerging from interactions of actors
within networks are named “network capabilities”
and included two main aspects: the architecture and
the identity (Kogut, 2000). Network architecture
refers to the links structure, the types of actors and
the coordination mechanisms. Through identity,
individual anchors their perception of self and other
and attach meaning to membership, as well as in the
categories of skill and knowledge that define a
spatial and cognitive division of labour. In addition,
Dyer and Nobeoka (2000) emphasise that if agents
are able to represent a shared space, knowledge
combination can be easily generated. Architecture
and identity thus allow to coordinate specialized and
distributed knowledge (Kogut, 2000). As such, they
enable the flow and the combination of knowledge
across organizational boundaries (Mitchell and
Nicholas, 2006). In sum, network innovation
capabilities rely not only on the existence on a broad
range of knowledge (cognitive variety), but also on
the ability of cluster’s members to access and
combine this knowledge, knowing ‘who knows
what’ and sharing mental models (transactive
However, network capabilities, and more
specifically the capacity to act collectively, are
rarely formed by design but rather “arise from
inherent characteristics of technologies that populate
an industry, as well as social norms and institutional
factors that favour the operation of particular rules”
(Kogut, 2000: 410). Thus, the question of how to
build effective network capabilities in order to foster
innovation through knowledge exchange and
combination is still open. In other words, can a
technological system support the development of a
transactive memory within a cluster defined as a
localized network in cases where it does not develop
2.2 Transactive Memory System at the
Cluster Level
A Transactive Memory System (TMS) is a shared
system that people in closed relationships develop
for encoding, storing, and retrieving information
about different substantive domains (Wegner, 1987;
Ren and Argote, 2011). The basic idea is that
individual knowledge in a group consists of internal
knowledge (held in his mind) and external
knowledge (which the individual can access using
the TMS) (Jackson and Klobas, 2008). As such,
TMS supposes that individuals play the role of
external memory for other individuals who in turn
encode meta-memories (i.e., the label or subject of
the knowledge as well as its location, but not the
knowledge itself) (Nevo and Wand, 2005).
TMS at the group level
Originally, researches on TMS were developed
at dyad or team group level of analysis. TMS
implies a cooperative division of learning,
remembering and communicating knowledge within
the group (Wegner, 1987). Over time, knowledge in
TMS becomes more specialized or differentiated
among members as a result of the division of
learning; at the same time, shared or integrated
knowledge increases as individuals develop a shared
cognitive representations of “who knows what”
(Brandon and Hollingshead, 2004; Ren and Argote,
As a result, three components are crucial to
TMS: cognitive interdependence, expertise related to
task and people, and shared mental models (Brandon
and Hollingshead, 2004). First, cognitive
interdependence describes the extent to which team
member’s work outcomes depend on a combination
of their own input and the input of others members.
As such, it motivates and sustains the development
of TMS. Task interdependence led thus to a higher
level of TMS, which in turn led to improved team
performance (Ren and Argote, 2011). Second,
Brandon and Hollingshead (2004) expand the basic
notion of labels and location (who knows what) into
a more explicit portrait of relations between Task,
Expertise and People (TEP). Third, shared mental
model concern not only a shared representation of
“who knows what” or “TPE” units, but also macro-
organizations of those “TPE” units. These shared
mental models have implications for the
effectiveness of the TMS. Brandon and
Hollingshead (2004) propose to evaluate their
development along three dimensions: accuracy,
sharedness and validation.
TMS at the organizational level
Only four studies have recently extended the
TMS concept to the organizational level (Ren and
Argote, 2011), including one case study (Jackson
and Klobas, 2008). Generalizing TMS to the
organizational level raises several challenges (Ren
and Argote, 2011). First, members might have more
trouble identifying who knows what in large
organizations than in small groups. Second,
organizations are composed by multiple subgroups
increasingly geographically distributed with less
communication and knowledge sharing across these
subgroups. Finally, when tacit knowledge is
available in a distal part of organization, retrieval
becomes difficult. Because organizations are larger
than work groups and geographically distributed,
Nevo and Wand (2005) shows that TMS might rely
upon advanced technology to locate and shared
information. They suggest that a general directory of
meta-memories should be formed, linking the
different communities and supporting knowledge
transfer between individuals in different
communities. In this case, knowledge transfer is not
provided through repositories but rather through
technology mediated connections.
Extending TMS to a large group requires the use
of artificial directories based on formalized meta-
knowledge integrating three main dimensions (Nevo
and Wand, 2005): conceptual, descriptive and
persuasive. In fact, a set of concepts is needed to
describe the subject of knowledge (ontology can be
used here). Descriptive knowledge can be
formalized to describe the author of knowledge
(location) and to characterize the knowledge (date,
format…). At the persuasive level, source of
credibility and perceptions of expertise should be
formalized. In this line, Jackson and Klobas (2008)
add two insights: in an organizational TMS people
access each other’s knowledge through a
combination of personal and codified directory
system; maintaining these directories which can be
activated for retrieving knowledge when it’s needed
(passive allocation) is more efficient than storing the
knowledge and sending content trough a system
network (repositories and active allocation).
Finally, TMS at the organizational level raises
two main questions (Nevo and Wand, 2005; Jackson
and Klobas, 2008; Ren and Argote, 2011). How an
organizational TMS can be developed with the
assistance of social network and information
technologies? How manage the ability to keep the
meta-knowledge directories updated? These two
questions that remains unanswered at the
organizational level become more crucial at the
cluster level. Indeed, clusters are larger than
organization and composed by multiple entities
which both compete and cooperate.
This section describes the KMP experience which
was conducted in the well-known technology park of
Sophia Antipolis (SA) in France (Castells and Hall,
1994). In this project, we applied an integrative
design science methodology (Pascal et al, 2013) to
create an interactive map of competencies to
enhance knowledge creation through partnerships
within the Telecom Valley cluster.
3.1 The Knowledge Management
Platform Project
Since the mid 1990s, the SA cluster has
progressively developed from a computer industry to
a telecom and IT industry cluster (Krafft, 2004). As
such, Telecom Valley, a non-profit organization,
was founded in 1991 by eight leading firms and
other organizations in order to facilitate
In 2000, the main characteristics of the Telecom
Valley (TV) cluster could be summarized as follows
(Lazaric et al, 2008). First, firms in this cluster were
evolving in a diverse technological context, covering
a wide range of industries (e.g. computing,
multimedia, space, information processing, on-line
services and networking, and microelectronics).
Given that most parent companies were located
elsewhere, the participants in the cluster had been
developing strong external links. The internal
dynamics of the cluster arose from the interactions
in several communities, associations, clubs, and so
forth, but also revealed a huge potential synergy
between agents in the cluster that was still largely
The lack of internal dynamics was the starting
point of the KMP project, launched in 2001 by TV.
Because they only have a partial view of the
different flows of knowledge developed by the
actors of the cluster, members of TV asked a map of
competencies to create strong local links with local
high-tech SMEs and research institutes. The
objective of the KMP project was thus to build an
interactive map of competencies which suggests a
lack of shared representation of who knows what
within the cluster.
3.2 An Integrative Design Science
Design science research develops knowledge in the
service of action and problem solving in
organizational settings. To address the research
objectives, we thus rely on an integrative design
science methodology that connects two perspectives
on design: science-based design drawing on design
propositions grounded in research and human-
centred design emphasizing an active and systematic
participation by users and other stakeholders (for
more details on the methodology see Pascal et al,
2013). This methodology is relevant in the case of
designing an innovative solution, where there
generally is no or limited scientific and practical
knowledge that is closely tied to the design goals at
hand (Pascal et al, 2013). It is also pertinent because
it assumes that technology per se and therefore TMS
based on an IT cannot determine work practices and
thus incorporates an enlarging network of users at
different stages of the design project (Newell et al,
2009; Nevo and Wand, 2005).
This methodology involves six steps. These steps
typically need to be taken in many iterations,
acknowledging that each step overlaps and is highly
intertwined with other steps.
1. Problem awareness. Before one can identify
any knowledge relevant to address a particular
design challenge or assignment, a clear
understanding of the nature of this assignment is
2. Developing design propositions. The scientific
knowledge relevant to the key problem addressed is
identified and synthesized into design propositions
thanks to the CIMO logic. CIMO involves four
components: (1) a problematic Context, in terms of
the surrounding (external and internal environment)
factors and the nature of the human actors
influencing behavioural change, (2) which suggests
a certain Intervention type that managers have at
their disposal to influence behaviour, (3) to produce,
by way of particular generative Mechanisms, the
processes that in a certain context generate (4) the
intended Outcomes (Denyer et al, 2008).
3. Creating scenarios of use. Scenarios of use
serve to explore the organizational context where
work practices are meaningfully accomplished
(Pascal and Rouby, 2006) and serve to convert and
articulate tacit knowledge of practitioners, and as
such, provide input for enriching the design
propositions (Plsek et al, 2007). Scenarios of use
therefore prevent the IT to interfered with the
implicit encoding system of the actors (Ren and
Argote, 2011).
4. Designing and developing artefacts. Drawing
on input from the (initial set of) scenarios of use and
design propositions, design work on artefacts is
conducted. Artefacts are the tangible result from the
design process and arise from contextualizing and
applying design propositions to particular practices.
5. Experimenting with prototypes. For any
information technology (IT) artefact, the design
evaluation process can not be limited to IT
performance but has to involve an in-depth study of
the (intended) artefact in its business environment
(Hevner et al, 2004; Pandza and Thorpe, 2010). As
such, the experimentation process exploits the
potential role of prototypes, extending the similar
role of other artefacts (e.g. drawings) developed and
used in earlier stages of the design process.
6. Organizational transformation. Finally, the
collaborative learning process may progressively
change the organizational context (or fail to do so).
As a result, the initial managerial problem typically
evolves, leading to redesign efforts or an entirely
new design cycle. At the same time, these
transformational processes modify, and possibly
enlarge, the network of users that support and apply
the tool (Akrich et al, 2002; Tatnall and Gilding,
3.3 Main Actors, Data Collection and
Researchers from different academic fields
composed the project team: economics and
management, computer science and ergonomics,
telecommunication sciences. The number of users
engaged in the project has gradually grown from two
TV working groups and several pilot users to
representatives of all TV’ actors. At the end of the
project, all other TV members, several clubs and
associations in the SA territory, and IT firms located
outside SA participated in the project but without a
direct involvement as pilot users.
We gathered data from three types of sources:
(1) interviews (26 open interviews with key
stakeholders, 52 semi-structured interviews with
pilot users, and 21 interviews with users as well as
other stakeholders to evaluate the prototypes); (2)
steering committee and working group meetings;
and (3) occasional meetings. Overall, we employed a
purposeful sampling strategy (Kumar et al., 1993)
towards all key stakeholders of the KMP project. In
particular, we drew on an iterative process of
simultaneously collecting data, analyzing data,
building conjectures (the design propositions) and
testing them through action (via artefacts). At the
same time, we were seeking new users to embed and
integrate the KMP portal in the users’ network in
and around TV.
Data analysis follows the different design cycles.
The first round of data analysis was guided by the
central research question on the dynamics of
knowledge creation inside a cluster. This initial
stage was based on the method of constant
comparison (Conrad, 1982; Glaser and Strauss,
1967). In a large number of iterations, data from
many informants are compared to identify
differences and anomalies and to identify and define
major categories, dimensions, themes, or processes
(Agar, 1986; Miles and Huberman, 1984; Spardley,
1980). At this stage, five researchers analyzed the
data. As such, by examining the congruency of data
patterns among informants, we obtained a clear
picture of the cluster, its dynamics of innovation,
and barriers and difficulties in knowledge sharing
and creation.
During the second and the third design cycles,
operating logics and practices have been described
within scenarios of use. These scenarios of use were
used first to build the tool and subsequently to
evaluate it. Two types of scenarios of use were
analyzed: the process of looking for a partner and
the process of co-evolution between the firms and
the cluster. The method of constant comparison
served to identify differences and anomalies.
Specifically, the analysis strategy was the synthetic
one (Langley, 1999) and was mainly based on the
different categories of the scenarios of use : (1) who
was the informant; (2) how the informant describes
his/her activities; (3) the information needed for
performing these activities; (4) the problem
encountered while performing these activities. Two
researchers coded each semi-structured interview to
develop the scenarios of use. Once developed, the
scenarios of use served to ask critical questions and
introduce alternative interpretations regarding
regularities, contrasts and/or anomalies in the data
(Nemeth et al., 2001). The data analyzed in
interaction with the literature served to create new
design propositions for developing and
implementing new functionalities of the portal.
These design propositions were evaluated by testing
and using the different prototypes. Data obtained
from the (evaluation-oriented) semi-structured
interviews with users as well as other stakeholders
(e.g. associations), steering committee meetings and
meetings with potential users were analyzed and
synthesized by examining the congruency of data
across informants, in order to inform the design team
about the usefulness of the portal and potential
modifications that would enhance this usefulness in
a pragmatic view (Dewey, 1938; Rorty, 1999).
Based on the literature review and our first local
practices analysis, we rapidly defined a meta-design
proposition that ensures the development of the
KMP platform: in a multi-actor cluster with a broad
scope of technologies (C), an interactive map of
competencies (I) will serve to foster knowledge
creation through R&D collaboration (O) by
reinforcing the four potential mediators of
knowledge creation: opportunity, anticipation
ability, motivation, and combinative capability (M).
This proposition does not specify the intervention
modalities, in terms of what kind of solution is
needed to activate each of the generative
mechanisms, and how to develop it. Three new
design propositions have thus been developed
through three successive design cycles between
2002 and 2006 (see Pascal et al, 2013) aiming at
foster TMS in the TV cluster in order to enhance
4.1 Mapping Competencies:
Highlighting “Who Knows What”
Following the first design proposition, we
investigate the development of a 'competencies map'
in order to create the computerized directories of
meta-memories of “who knows what” and thus to
foster TMS. We choose to describe competencies
instead of knowledge because competencies
combine knowledge in action for the output at hand.
As such, describing competencies enriches the
comprehension of “who knows what” by linking
task, expertise and person (Brandon and
Hollingshead, 2004).
However, mapping competencies within a
cluster raises many challenges. First, it is necessary
to identify the appropriate level of competencies
description (individual or collective). The second
challenge is to describe competencies across the
cluster in sufficient detail without disclosing
strategic know-how. Finally, using an IT mapping
raises the issue of data collection and updating.
Combining literature review and local practices, we
established the following design proposition.
DP1: In a multi-actor cluster with a broad
scope of technologies (C), an interactive map of
competencies (I) provides relevant information that
enhances opportunities (M) for finding the good
partner for R&D collaboration (O). To trigger the
opportunity mechanism, a competency is defined as
an action that mobilizes technical, scientific and
managerial resources to produce deliverables that
are likely to create value in a business activity.
Given the size of the cluster, we decided to
describe collective competencies at the team level
which is the appropriate level when looking for a
R&D project partner. We defined an abstract model
of competencies, based on the four abstract
categories: action / resources (including knowledge)
/ deliverables / business activity (Rouby and
Thomas, 2004). These abstract categories are the
first codes shared by the community and the first
bricks for building shared representations. They
constitute the roots of the elaboration of four
specific ontologies (action, resources, deliverable
and business activity). An ontology is “an object
capturing the expressions of intensions and the
theory accounting for the aspects of the reality
selected for their relevance in the envisage
applications scenarios” (Gandon, 2001). The model
of competences and its four constitutive ontologies
permit to locate the competencies and to compare
them depending on the interest and vision of the
actor which can choose in its queries its relevant and
appropriate combination of categories. Users
scenario points different kind of queries: simple
queries on, for example, a particular technology (e.g.
“which firms are working on J2ME?”), a delivery
(e.g. “who has successfully produced video
games?”) or a business activity (e.g. “which firms
are doing work for the 3G mobile sector?”) as well
as more complex queries which combined several
categories such as technology and business activity.
Once competencies are identified and located, an
accurate description is suggested including what is
the problem solved by the competency (for instance
the storage of data on electronic chips), how this
problem is solved (the know-how, skills, equipment
on the building of chips), and the patents,
publications, R&D collaboration, and industrial
partnerships involved. These additional details are
essential to a proper understanding and to reinforce
the credibility of a partner’s competencies. This
description is not based on formal categories
allowing firms to be more or less precise on this
strategic aspect and to use natural language in this
Ontologies used in the KMP project are relevant
A semantic representation of information
allows for more precise research and increases
the degree of answer liability. Ontologies
improve the retrieval of knowledge because
they can focus the results on a specific subset
and then reduce the set of results (Nevo and
Wand, 2005) or conversely to enlarge it if
A semantic representation of information
allows for more precise research and increases
the degree of answer liability. Ontologies
improve the retrieval of knowledge because
they can focus the results on a specific subset
and then reduce the set of results (Nevo and
Wand, 2005) or conversely to enlarge it if
Ontologies allow to acknowledge different
points of view held by spatially distributed
and heterogeneous actors. For example, for
actors belonging to the commercial
professions, 3G (third generation in
telecommunication) and multimedia are
ontologically equivalent. Conversely, in the
technologically oriented professions, 3G and
multimedia are quite distinct because they
belong to separate technological trajectories.
This is the reason why the two terms will be
considered distant in the ontology of the
technological resources, while they will be
very close in the business activity’s ontology.
Finally, to face the size and the large scope of
technologies characterizing the TV cluster, data
collection and updating were highly decentralized
and manage by teams composing the different
organizations. Each team described its own
competencies (between 5 and 10) and added when
necessary new concepts in ontologies. Several expert
groups first agreed on the basic roots of these
ontologies. New concepts were then integrated as
competencies’ description increases and were
regularly validated by expert groups.
4.2 Common Space: Highlighting
Similarities and Complementarities
The second major issue in designing the
competencies map involved developing a shared
identity of the cluster. Members of TV’s board
raised two problems regarding this lack of identity:
(1) “It has always been ambiguous whether Sophia
is more telecom or software”; (2) “We never know
when we have to accept the entry of a consultancy
firm. Generally, the decision depends on the size of
the firm. Thus, we lean more on political aspects
than on industrial or innovation logics. We are not
satisfied by this way of thinking, but we don’t know
how to do it otherwise”.
Regarding the cluster identity issue, the literature
reveals that the representation of a common space
may help individual to develop a shared meaning of
membership and a shared representation of the
cognitive interdependences of labour (Kogut, 2000;
Dyer and Noboeka, 2000). After several design
loops, a design proposition on the cluster’s common
space representation was therefore stabilised.
DP2: In a multi-actor cluster with a broad scope
of technologies (C), building a common space
representation of the cluster (I) reinforces the
motivation of actors and their ability to anticipate
value created from knowledge exchange and
combination (M) to effectively engage in R&D
collaboration (O).
This common space has to exhibit the following
properties: (a) it represents all actors in terms of
their main competencies: scientific and technical
competencies (key stakeholders), managerial
competencies (support) and relational competencies
(facilitators); and (b) it positions the competencies of
stakeholders in technological poles (similarity
concept) as well as value chains (complementarity
To evaluate the degree of similarity and
complementarity, the map of competencies draws on
the following definitions: competences are similar
when they share the same resources, and
complementary when sharing the same business
Figure 1: the cluster common space representation.
As figure 1 shows, the common space
representation identified three kinds of actors:
1. The stakeholders who participated in
knowledge creation in the cluster; that is to
say those who had technical competencies
such as firms and public research
laboratories. The competencies of the
stakeholders are positioned on the
technological poles depending on the main
resources they mobilized. As such, a firm
can be present on different technological
2. The facilitators, including all associations,
clubs or service providers, whose goal was
to help find partners (relational
3. Support organizations in the area of law,
finance and management that would ensure
partnerships by providing managerial
This representation identifies technological poles
with actors who shared similar competencies and
value chains which combined complementary
competencies. Value chains are business-driven and
composed with competencies which shared the same
business activity. These competencies are different
but complementary by producing interrelated
outputs for the same business activity. Value chains
are not given, but dynamically built from the
particular competencies described by the users in the
platform. In addition, the representation allows,
through ontologies, to define general overviews on
business activity sector (eg the telecom value chain)
or precise overviews on market segment (eg the
bluetooth or 3G market). It also provides a diagnosis
of the weaknesses and strengths of the cluster in
terms of the nature and number of competencies in
each technological domain for a specific value
The distinction between similar and
complementary competencies supports the
development of a shared understanding of the
cognitive division of labour. It resolves the
perceived ambiguity on the cluster identity by
showing that the cluster enjoys a lot of software
competencies (which contribute to a technological
pole) which mainly addressed the telecom market.
More generally, the interactive representations of the
common space have effects on motivation and
ability to anticipate. By increasing the actors’ self-
consciousness about the competencies distributed in
the cluster and their interrelations, it reveals actor’s
games of interests. For example, firms in the
software pole realized that they could gain more in
being partners than being fierce competitors and
began to develop partnerships about joints solutions,
aiming to reach more and bigger customers, within
and outside Sophia Antipolis.
Finally, the KMP project showed that the
progressive design of the common space
representation really mobilized all the TV’ actors
and consequently motivated them to engage more
actively in the project (Pascal et al, 2013).
This study provides empirical support to the
potential benefits of the TMS approach at the cluster
level and specifically the ability of an IT to support
the development of an effective TMS in a cluster. A
TMS cluster exhibits specific characteristics
different to those of groups or organizations. It needs
appropriate directory structures and shared models
to an organizational context including a wide range
of actors (and expertises) who both cooperate and
compete. In line with Nevo and Wand (2005), our
results demonstrate the ability of IT to extend the
notion of TMS to large groups including clusters.
Our study suggests that artificial directory of
meta-memories can be formalized in order to link
different organisations and sub-groups (eg teams)
and to facilitate knowledge exchange and
combination. It provides specific guidelines to create
an artificial directory based on an appropriate
formalization of knowledge about who knows what
in a cluster. This formalization integrates the three
main dimensions highlighted by Nevo and Wand
- the competency model based on four abstract
categories (action / resources / deliverables /
business activity) and their ontologies constitute the
conceptual dimension of knowledge. It allows to
accurately identify and locate the expertise within
the cluster.
- this expertise described at the team level is then
completed by additional information (eg know-how,
skills, equipments, partnerships, patents…)
constituting the descriptive dimension.
- these additional details which are more or less
precise regarding the firms’ communication
strategies reinforce the credibility of a team’s
competence ie the persuasive dimension.
While using a simple model (based on only four
categories) enables sharedness, the building of
ontologies for each category and the combination of
these categories in the encoding and retrieval of
competencies allows accuracy. In addition,
ontologies allow knowledge to become more
specialized or differentiated among members even in
context where members in different groups not share
concepts to describe the contents of knowledge. The
formalization of competencies in the KMP project
thus achieved accuracy and sharedness, two
dimensions of TMS effectiveness (Brandon and
Hollingshead, 2004).
Our project also reveals that building an artificial
directory to identify and locate expertise is not
sufficient to support an effective TMS. The IT has to
create structures that highlight the coordination or
combination of distributed expertise. In the KMP
project, this structure is based on the common space
representation which creates a share cluster identity
and a mutual understanding of the division and
coordination of labour and expertises in the cluster.
This identity and shared understanding constitute
motivational factors that affect the development of
TMS. Indeed, according to Ren and Argote (2011:
204), “in groups where members identify with the
group, they are more likely to invest in developing
the specialized division of labour that is defining
characteristics of TMS”. In addition, identity and
mutual understanding affect members’ motivation to
engage in collective processes of communication
and knowledge exchange and combination (Kogut,
2000; Dyer and Nobeoka, 2000). As such, the
common space representation supports the
interpersonal side of the TMS development.
In sum, an IT that supports TMS at a cluster
level must rely on two main characteristics: to
combine both accurate descriptions of knowledge at
a micro level (team) and macro representations of
the cluster knowledge; to favour both interpersonal
and technological approaches of TMS.
Several paths for future research can be derived
from the work described in this paper. One strand of
research might be to study the dynamics of cluster
TMS development with attention to how the
technical and interpersonal approaches of TMS
evolved and enriched each other over time. Other
research should focus on clusters with a well-
developed identity and architecture in order to
analyse the existence of an effective TMS in its four
dimensions: accuracy, sharedness and also
validation and convergence. Finally, a promising
new direction is the inclusion of innovation as an
outcome of TMS. Whereas team performance is
traditionally the focus of analysis, our study
examines the effect of TMS on how knowledge is
combined and recombined. Further investigations
are needed to analyse the relationship between TMS
and innovation at the team, organisational or cluster
Agar, M. H. (1986). Speaking of Ethnography. Beverly
Hills, CA: Sage.
Akrich M, Callon M and Latour B (2002) The key to
success in innovation, International Journal of
Innovation Management 6(2), 207-225.
Antonelli C and Calderini M (2008) The governance of
knowledge compositeness and technological
performance: the case of the automative industry in
Europe, Economics of Innovation and New
Technology, 17 (1), 23-41.
Baskerville R and Dulipovici A (2006) The theoretical
foundations of knowledge management, Knowledge
Management Research & Practice 4, 83–105.
Brandon DP and Hollingshead AB (2004) Transactive
Memory Systems in Organizations: Matching Tasks,
Expertise, and People, Organization Science 15(6),
Castells M and Hall P (1994) Technopoles of the World:
The Making of 21st Century Industrial Complexes.
London: Routledge.
Conrad, C. F. (1982). ‘Grounded theory: An alternative
approach to research in higher education’, Review of
Higher Education, 5, pp. 259-269.
Crevoisier O and Jeannerat H (2009) Les dynamiques
territoriales de connaissance: relations multilocales et
ancrage regional, Revue d’Economie Industrielle 128
(4), 77-99.
Denyer D, Tranfield D and van Aken JE (2008)
Developing design propositions through research
synthesis, Organization Studies 29(3), 393-414.
Dewey, J. (1938). Logic: The Theory of Inquiry. New
York: Henry Holt.
Dyer J.H. and Nobeoka K. (2000) Creating and managing
a High-performance Knowledge-sharing Network: The
Toyota Case, Strategic Management Journal 21, 345-
Gandon F. (2001) Engineering an Ontology for a Multi-
agents Corporate Memory System. INRIA Research
Report, INRIA, France -,
Glaser, B. G. and A. L. Strauss (1967). The Discovery of
Grounded Theory: Strategies for Qualitative Research.
Chicago: Aldine.
Hevner AR, March ST, Park J and Ram S (2004) Design
science in information systems research, MIS
Quarterly 28(1), 75-106.
Jackson P and Klobas J (2008) Transactive memory
systems in organizations: Implications for knowledge
directories. Decision Support Systems 44, 409-424.
Kogut B (2000) The Network as Knowledge: Generative
Rules and the Emergence of Structure, Strategic
Management Journal 21(3), 405-425.
Kogut B and Zander U (1992) Knowledge of the firm,
combinative capabilities, and the replication of
technology, Organization Science 3(3), 383-397.
Krafft J (2004) Entry, exit and knowledge: Evidence from
a cluster in the info-communications, Research Policy
33(10), 1687-1706.
Kumar N, Stern LW and Anderson JC (1993) Conducting
interorganizational research using key informants,
Academy of Management Journal, 36(6), 1633–1651.
Lazaric N, Longhi C and Thomas C (2008) Gatekeepers of
Knowledge versus Platforms of knowledge: from
potential to realized absorptive capacity, Regional
Studies 42(6), 837-852.
Langley, A. (1999). ‘Strategies for theorizing from process
data’, Academy of Management Review, 24/4, pp.
Maskell P and Lorenzen M (2004) The Cluster as Market
Organization, UrbanStudies 41, 991-1009.
Miles, M. B. and A. M. Huberman (1984). Qualitative
data analysis. Newburry Park, CA: Sage.
Mitchell R and Nicholas S (2006) Knowledge creation
through boundary-spanning, Knowledge Management
Research & Practice 4, 310–318.
Moran P. and Ghoshal S (1996) Value Creation by Firms.
In Keys JB, Dosier LN (eds.) Academy of
Management Best Papers Proceedings, 41-45.
Nahapiet J and Ghoshal S (1998) Social Capital,
Intellectual Capital, and the Organizational
Advantage, Academy of Management Review 23(2),
Nelson R. Winter S. G. (1982) An evolutionary theory of
economic change. Cambridge, Mass.: Belknap Press
of Harvard University Press.
Nemeth, C., K. Brown and J. Rogers (2001). ‘Devil’s
advocate versus authentic dissent: Stimulating quality
and quantity’, European Journal of Social Psychology,
31, pp. 707–720.
Nevo D and Wand Y (2005) Organizational memory
information systems: a transactive memory approach,
Decision Support Systems 39, 549-562.
Newell S, Robertson M, Scarbrough H and Swan J (2009)
Managing knowledge work and innovation,
Basingstoke ; New York : Palgrave Macmillan.
Nooteboom B (2005) Innovation, learning and cluster
dynamics, Discussion Paper 44, Tilburg University,
Center for Economic Research
Pandza K and Thorpe R (2010) Management as design,
but what kind of design? An appraisal of the design
science analogy for management, British Journal of
Management 21(1), 171-186.
Pascal A and Rouby E (2006) Construire des scénarios
d’usage: une approche gestionnaire – Le cas KMP –.
In R. Lehn, M. Harzallah, N. Aussenac-Gilles and J.
Charlet (eds), 17èmes Journées Francophones
d'Ingénierie des Connaissances (CD Proceedings).
Nantes, June 2006.
Pascal A, Thomas C and Romme AG (2013) Testing a
human-centered and science-based design
methodology, British Journal Management 24(2),
Plsek P, Bibby J and Whitby E (2007) Practical methods
for extracting explicit design rules grounded in the
experience of organizational managers, Journal of
Applied Behavioral Science 43, 153-170.
Ren Y and Argote L (2011) Transactive Memory Systems
1985-2010: An integrated framework of key
dimensions, antecedents, and consequences, The
Academy of Management Annals 5(1), 189-229.
Rorty, R. (1999). Philosophy and Social Hope. London:
Penguin Books.
Rouby E and Thomas C (2004) La codification des
compétences organisationnelles: l’épreuve des faits,
Revue Française de Gestion 149(2), 51-68.
Shawney M and Prandelli E (2000) Communities of
Creation: Managing Distributed Innovation In
Turbulent Markets. California Management Review
42(4), 24-54.
Spardley, J. P. (1980). Participant observation. New York:
Holt, Rinehart and Winston.
Tatnall A and Gilding A (1999) Actor-network theory and
information systems research, Paper presented at 10th
Australasian Conference on Information Systems
(ACIS), Wellington, Victoria University of
Van de Ven AH (2005) Running in packs to develop
knowledge-intensive technologies, MIS Quarterly
29(2), 365–378.
Wegner DM (1987) Transactive Memory: A contemporary
analysis of the group mind. In B. Mullen & G.R.
Goethals (eds.), Theories of group behavior 9, 185-
208. New York: Springer.