organizational memory information systems: a transactive memory approach
TRANSCRIPT
www.elsevier.com/locate/dsw
Decision Support Systems 39 (2005) 549–562
Organizational memory information systems: a transactive
memory approach
Dorit Nevoa,*, Yair Wandb
aS337H, Schulich School of Business, York University, 4700 Keele St., Toronto, ON, Canada M3J1P3bSauder School of Business, University of British Columbia, Vancouver, BC, Canada
Available online 10 May 2004
Abstract
Effective management of organizational memory (OM) is critical to collaboration and knowledge sharing in organizations.
We present a framework for managing organizational memory based on transactive memory, a mechanism of collective memory
in small groups. While being effective in small groups, there are difficulties hindering the extension of transactive memory to
larger groups. We claim that information technology can be used to help overcome these difficulties. We present a formal
architecture for directories of meta-memories required in extended transactive memory systems and propose the use of meta-
knowledge to substitute for the lack of tacit group knowledge that exists in small groups.
D 2004 Elsevier B.V. All rights reserved.
Keywords: Organizational memory information system; Transactive memory; Meta-knowledge
1. Introduction the way organizations store knowledge from the past to
Organizational memory (OM) and knowledge man-
agement are two intertwined topics that have grown in
importance for businesses and academics over the past
few years. Knowledge management encompasses var-
ious practices of managing organizational knowledge
such as knowledge generation, capture, sharing, and
application [2]. Within these practices, effective shar-
ing and use of organizational knowledge depends—to
a large extent—on the organization’s ability to create
and manage its collective memory. This collective
memory is often referred to as organizational memory
(OM). The organizational memory can be described as
0167-9236/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.dss.2004.03.002
* Corresponding author.
E-mail address: [email protected] (D. Nevo).
support present activities [24].
Organizational memory can increase organizational
effectiveness by supporting the coordination of work,
management of information, the organization’s re-
sponsiveness to changes, and the definition and pur-
suit of organizational goals [25]. Such memory
generally resides in different retainers in the organi-
zation and organization members retrieve its content
based on their work needs [28]. To support effective
management of organizational memory, Stein and
Zwass [25] propose the use of information technology
to accomplish four specific processes related to orga-
nizational memory: acquisition, retention, mainte-
nance, and search and retrieval. In addition, they
outline a design for an organizational memory infor-
mation system (OMIS) that includes a ‘‘mnemonic
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562550
functions layer’’ intended to provide the functionality
necessary to support the above four processes. The
general design requirements of this layer include the
ability to capture and represent knowledge in OM, the
ability to communicate knowledge, and the mainte-
nance of the contents of the OM.
While the management of OM is a good candidate
for the use of information technology, the specific
design of such an information system is not a simple
task. We suggest there are five reasons for this.
First, much of the knowledge in the OM is contex-
tualized. When knowledge is transferred, the receiving
end of the communication system often does not know
the original context of the knowledge and therefore
cannot interpret it correctly [3,24]. In order to cross
boundaries—either departmental or organizational—
the knowledge needs to be stripped of its context for the
receiving end to be able to understand it [1]. An
interesting distinction can be made here about the
magnitude of this problem in high versus low context
communication environments. High context commu-
nication is defined as one in which much of the
information is embedded in the context—either phys-
ical or personal—and very little information is explic-
itly coded. Low context communication is on the other
end of the continuum, representing messages in which
most of the information is explicitly coded in the
message [10]. Thus, people in low context cultures
will rely more on formal communications that can be
verbally expressed, while people in the high context
cultures will rely on context variables such as individ-
ual background and associations [16]. This distinction
implies that the effective application of technology to
support organizational memory might be greater in
low- context cultures and more challenging in high
context cultures.
A second problem concerns the locations of knowl-
edge. OM generally resides in five different types of
retainers [28]: Individuals, who retain knowledge in
their memory stores or in their belief structures,
values, or assumptions; culture that stores knowledge
in language, shared framework, symbols, and stories;
transformations, procedures, and rules, which include
embedded knowledge such as the logic behind them;
structure and roles that represent the organization’s
perception of the environment, and social expect-
ations; and finally, the physical settings of the work-
place represent knowledge about status hierarchy and
behaviour perceptions. These retainers of OM may be
in different locations and their memories might be
difficult to combine [3,32].
A third problem with OM management is that
knowledge is often tacit. Tacit knowledge is a highly
personal knowledge and is hard to formalize. It is
rooted in action, commitment, and involvement in a
specific context [20]. Tacit knowledge is difficult to
track and maintain in large organizational memories
[25].
A fourth problem concerns the volatility of orga-
nizational knowledge. This volatility results in fre-
quent changes to the contents of the OM [32]. In
addition, combined with the problem of context de-
pendence, the volatility of knowledge further compli-
cates the search and retrieval of knowledge included
in the OM.
Finally, since some knowledge is retained outside
the organization [28] or from unfamiliar sources
within the organization, a measure of the retainer’s
legitimacy and reliability is required [3]. In fact, an
inquirer is more motivated to retrieve knowledge if he
or she is aware of the knowledge and sees potential
value in the knowledge [24]. This information should
be attached to OM to facilitate the retrieval and use of
knowledge.
These five problems create difficulties for members
of the organization in retrieving and using knowledge
that resides in OM. As well, they complicate the
design of the mnemonic layer of an OMIS, i.e., the
layer intended to support the processes involved in
using the OM. As a result, organizations might not be
able to attain the potential benefits of increased
effectiveness and performance that can be associated
with effective OMIS [25].
This paper explores how technology can be used to
overcome the above problems and proposes a con-
ceptual design for an information system intended to
support effective management and use of organiza-
tional memory. We base our approach on the obser-
vation that small workgroups—called communities of
practice—are usually efficient in their communication
and sharing of collective knowledge, even when the
knowledge is tacit [7,8]. Therefore, to gain a better
understanding of possible ways to overcome the
barriers for efficient OM management, we examine
the processes involved in the management of the
collective memory of smaller workgroups. In partic-
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562 551
ular, we use the concept of transactive memory
systems that has been developed to explain how
individuals in small workgroup form a collective
memory [18,30,31]. We propose what the barriers to
extending transactive memory mechanisms to the
whole organization might be, and suggest how the
use of technology can help to overcome these barriers.
In Section 2 the next section, we review the trans-
active memory literature. In Section 3, we propose an
extension of transactive memory systems to large
groups with the help of information technology. In
Section 4, we examine the potential benefits of such
systems and provide some empirical support for these
benefits. Finally, Section 5 provides some conclusions
and recommendations for future work.
2. The transactive memory approach
A transactive memory system is ‘‘. . .a set of
individual memory systems in combination with the
communication that takes place between individuals’’
([31], p. 186]. Such a system is built on the distinction
between internal and external memory encoding. In
many cases, individuals encode new knowledge in-
ternally, that is —they learn something new and
catalogue it in memory for future retrieval and use.
However, even more often individuals encode knowl-
edge externally either in diaries, in books, or even in
other people’s memory. In these cases, the individual
internally encodes the label (subject) of the knowl-
edge as well as its location, but not the knowledge
itself. Transactive memory systems are built on the
notion of individuals playing the role of external
memory for other individuals who—in turn—encode
meta-memories (i.e., memories about the memories of
others). Knowledge is encoded, stored, and retrieved
through various transactions between individuals.
Three stages are involved in the creation and
maintenance of transactive memory systems: directory
updating, information allocation, and retrieval coordi-
nation [30]. At the first stage, group members create
directories of meta-memories containing information
about the memories held by others. These meta-
memories usually include information about the sub-
ject and location of the knowledge but also—tacitly—
some perceptions about the individual’s own and
others’ expertise on each subject [22].
When new knowledge enters the group, it is
allocated to the person who is perceived by the group
as the expert on the topic. This expertise differentia-
tion can develop naturally within the group or be
imposed by defining roles and allocating responsibil-
ities. For example, a hotel reservation manger is
formally responsible for knowledge related to the
reservations and guests of the hotel. However, she
may also develop an interest in technological innova-
tions and therefore become the informal expert on
new technologies that may be useful for the hotel. In
either case, relevant knowledge that enters the orga-
nization is allocated to her.
Finally, a group member wishing to retrieve some
knowledge will first assess his or her own ‘‘feeling
of knowing’’ on the topic and then—if necessary—
will evaluate other group members that may possess
this knowledge. This again requires some evaluation
of perceived expertise. For example, if the reserva-
tion manager would like to know the credit of a
potential guest, she will probably turn to the hotel’s
credit department for the relevant knowledge.
Initial research on transactive memory focused on
dating couples [12,31] and later extended to small
groups. In group studies, the existence of transactive
memory was measured by assigning groups the simple
task of building the AM portion of an AM/FM radio.
Some of the groups were trained together, while others
were trained individually and assigned to groups after
the training session. In three different experiments, the
results show that training together led to the develop-
ment of transactive memory in the group and also
resulted in improved group performance [18]. An
additional benefit of transactive memory is in creating
a more efficient problem problem-solving mechanism
[17]. Since group members know more about each
other, they are able to match problems with the people
who are most likely to solve them.
While most research on transactive memory focuses
on small groups or intimate couples, some work relates
transactive memory to OM [3]. However, these studies
do not extend a single transactivememory system to the
whole organization, but rather remain within the frame-
work of small groups, for example, by examining the
existence and relations between various transactive
memory systems within organizational memory [3].
To apply the above benefits of transactive memory
to the whole organization, we first examine whether
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562552
there is basis to believe that a transactive memory, in
fact, develops in organizations. To this end, we exam-
ine three measures for the existence of transactive
memory: memory differentiation, which addresses the
specialization of group members in specific topics; task
coordination, which reflects the ability of group mem-
bers to work together smoothly; and task credibility,
which is the level of trust between group members in
each other’s knowledge [18].
When organizations are created, formal roles are
assigned to people and groups. These roles create
memory differentiation in the organization. For ex-
ample, in a business school, knowledge of adminis-
tering students is divided between the undergraduate
office, the Masters office, and the PhD office. The
departmental structure of organizations also creates
coordinated tasks and—in many cases—task credi-
bility issues. In other words, using the three meas-
ures, we can view organizations as a workgroup of
smaller groups that work together. In this case, a
transactive memory system develops [3]. Moreover,
often organizations invest in various knowledge-
creating social activities, such as employee educa-
tion, the creation of social networks or brainstorming
activities [4]. Such activities can contribute to the
natural development of transactive memory within
the organization.
It seems, therefore, that the mechanisms operating
in transactive memory systems in small groups
should have been able to support effective OM
management across the organization. Yet, as the
analysis of problems in the previous section points
out, this is not the case. We propose two explanations
to these observations:
1. The meta-memory directory required for people to
allocate and retrieve knowledge from the right
group might simply be too large for the
individual’s memory capacity. This is especially
the case in large or geographically dispersed
organizations, and even more so if we consider
that individuals may be required to possess some
knowledge about members within every subgroup
in the organization.
2. Uncertainty might exist as to who should be the
‘‘owner’’ of certain types of knowledge. In
particular, certain knowledge may not ‘‘‘formal-
ly’’’ belong to any department or to any specific
role in the organization. In such cases, the
knowledge may be allocated to individuals based
on their personal interest or on internal expertise
definitions within workgroups. While such infor-
mal allocations work well within small groups,
they can cause problems when applied across the
organization.
We propose that these two problems—which hin-
der the extension of the transactive memory mecha-
nisms to large groups such as an organization—can
be alleviated with the use of technology and a more
formalized approach to meta-memory directories. In
other words, we propose to use an information
system to support an organization-wide transactive
memory system. Specifically, we identify the role of
technology in the creation and maintenance of the
directories of organizational meta-memories. We sug-
gest that by using technology to support organiza-
tion-wide directories, we can enable individuals to
identify knowledge retainers across the organization
(i.e., outside their own immediate work groups) even
in very large or geographically dispersed organiza-
tions, and thus leverage the benefits of transactive
memory. In the next section, we propose a mecha-
nism for creating such organizational directories of
meta-memories.
3. Technology-enhanced transactive memory
Using the notion that small groups succeed in
effectively utilizing a shared memory, we view orga-
nizational memory as consisting of clusters (commu-
nities) of individual memory retainers described in
Fig. 1.
As previous research had shown, transactive mem-
ory develops within each community and members
hold updated directories about other group member’s
memories [3,18,22]. To extend this view to the
organization, we suggest that a more general directory
of meta-memories should be formed, linking the
different communities and assigning responsibilities
of knowledge assimilation to specific groups (for
example, the sales department is in charge of cus-
tomer knowledge). This directory can also support
knowledge transfer between individuals in different
communities.
Fig. 1. A view of organizational memory.
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562 553
We begin our discussion of the proposed applica-
tion of transactive memory to the whole organization
by identifying the following three types of knowledge:
1. Role knowledge—this is knowledge that is re-
quired by the definition of the knowledge retainer.
For example, a project manager is required to retain
specific knowledge about projects.
2. Instance knowledge—this is knowledge that is not
required by the formal definition of the knowledge
retainer’s role. For example, the secretary that has
been with the company for year may have
knowledge about former employees or former
experiences of the company that is not required
by her job definition. Individual instances can
therefore have an informal role as knowledge
retainers.
3. Transactive knowledge—this is the directory
knowledge a retainer has about group members.
The availability of transactive knowledge enables
retainers to effectively extend the knowledge
available to them by being able to access their
group members’ knowledge. This knowledge may
be either role or instance knowledge.
In an effective transactive memory system—such
as those that develop within close workgroups—the
transactive knowledge of each retainer will include
references to both the role and the instance knowledge
of all other retainers. However, when extended to the
whole organization, individuals’ transactive knowl-
edge usually will only include partial reference to role
knowledge beyond the immediate workgroup and
likely very little reference to instance knowledge (role
knowledge can be acquired from organizational struc-
ture and from the division of responsibilities between
the various divisions and roles). As our objective is to
use information technology to enhance the individual
transactive memory knowledge and extend it to orga-
nization-wide sources, we propose to create artificial
directories containing transactive knowledge. These
directories will thus provide information about both
the role and instance knowledge of memory retainers
across the organization.
3.1. Artificial directories of meta-memories
Traditionally, there are two main dimensions to the
directories of meta-memories—the subject of the
knowledge and the location of the knowledge [30].
Thus, when new knowledge enters the group, it is
categorized in group members’ memories based on
these two dimensions. In addition to these two dimen-
sions, a third—tacit—dimension also exists that
includes some perceptions of expertise concerning
self and others’ knowledge [22]. Based on this tacit
dimension, new knowledge that enters the group is
allocated to the perceived group’s expert on the topic
and only then are its subject and location stored in
memory. Similarly, an individual who requires knowl-
edge will assess who in the group is likely to have the
knowledge.
To extend transactive memory to a large group
using artificial directories, we therefore need to for-
Table 1
Categories of meta-knowledge
Category Examples References
Conceptual Ontology: set of concepts needed
to describe a domain. Meta-
models: formalized descriptions
of generalized concepts
[14,21,23]
Descriptive Author information, scope of the
knowledge, intended audience,
cost of attaining the knowledge,
format of the knowledge, date
of knowledge, etc.
[5,15,26]
Cognitive Meta memory; meta-cognitive
knowledge—knowledge about
our own knowledge and abilities
[9,31]
Persuasive Source credibility, expertise,
trustworthiness
[11,13]
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562554
malize the two dimensions of meta-memories as well
as to substitute for the third dimension consisting of
tacit group knowledge. The latter, in particular, is
intended to overcome four specific problems of for-
malizing meta-memories. First, members in different
groups might not share a set of concepts to describe
the contents of knowledge (needed or available).
Fig. 2. A model of extended m
Thus, the notion of the subject of knowledge needs
to be further formalized. Second, perceptions of
expertise when group members do not know each
other closely should be formalized. Third, it is diffi-
cult or even impossible to keep track of the expertise
of all members in a large group, as members of one
group might not even know about members of other
groups (let alone know about their knowledge). Fi-
nally, what we termed ‘‘instance knowledge’’ above,
might be easy to perceive within a group, but hard to
formalize across the organization.
We refer to the above four issues, which together
comprise information about knowledge available to
individuals, as meta-knowledge. The literature has
addressed various elements of meta-knowledge and
we summarize some of the most common elements in
four categories presented in Table 1.
We propose that the computerized directories of
meta-memory be based on this formalized meta-
knowledge in order to compensate for the lack of
the group’s tacit knowledge. In other words, we
propose to explicate the third (tacit) dimension of
meta-memories using the notion of meta-knowledge.
The diagram in Fig. 2 depicts a simple model—based
eta-memory directories.
2
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562 555
on the entity-relationship notation—linking the three
meta-memory dimensions of retainers (location),
knowledge (subject), and meta-knowledge.
In this model, retainers of knowledge are charac-
terized by cognitive and descriptive meta-knowledge
(e.g., years of education or perceived self expertise).
These retainers possess knowledge about concepts
and/or instances in the organization—i.e., about sub-
jects of knowledge. The knowledge possessed by
retainers is in the form of predicates (truth statements)
about the state of affairs at a given time, past changes,
or possible changes in a domain of interest. The
knowledge predicates are characterized by descriptive
meta-knowledge, for example, the currency of the
predicate, as well as persuasive meta-knowledge such
as the expertise of the retainer on the specific subject
of knowledge.
The subjects of knowledge (concepts and instan-
ces) can also be characterized by descriptive meta-
knowledge. In addition, knowledge subjects are often
related to other subjects. For example, knowledge
about the concept ‘client’ might be related to knowl-
edge about the concept ‘product’. Some common
types of relations are instantiation, specialization,1 or
complementarities of knowledge (i.e., if you know
how to operate a car, then you also need to know the
traffic rules). The specific meta-knowledge that char-
acterizes this relationship is shown in the model as the
conceptual meta-knowledge that signifies the type of
the relationship that exists between predicates. More-
over, in the context of conceptual meta-knowledge, an
organizational ontology defines the domain of orga-
nizational knowledge, i.e., the concepts about which
knowledge might exist and their relationships. There-
fore, conceptual meta-knowledge is also represented
in Fig. 2 by the ‘concept/instance’ entity itself.
Based on the above view of organizational mem-
ory, our extended directories of meta-memory include
three types of elements and the relationships between
them: subjects of knowledge (i.e., the organizational
ontology), retainers of knowledge, and cognitive,
descriptive, and persuasive meta-knowledge. We
now turn to formalizing these three components to
examine how they can be handled using information
technology.
1 ‘c2’ is a sub-concept of ‘c1’ if the set of properties of ‘c1’ is a
subset of the set of properties of ‘c2’.
3.2. Formalization
We define O as the set representing the organiza-
tional ontology: O={xjx is a term relevant to the
organization}. Effectively, O is the set of terms
needed to describe the universe of discourse of
organizational memory. A term in the ontology may
be a generic concept (e.g., a person), an instance of
another concept (e.g., a specific individual), or a
relationship that might exist between concepts and
or instances (e.g., ‘John is a client’).
In addition to the ontology, the retainer set defines
all the possible memory retainers (based on the five
types of retainers described in Ref. [28]). Again a
retainer can represent a generic concept, in which case
it is termed a Role (e.g., ‘Production Manager’) or it
can be an instance of a role (e.g., ‘Jane, the Marketing
Manager’).
We now define the relation2 knows about: K(x,y),
where x is a subject and y is a retainer. For example:
K(‘Jane’, Client) represents that Jane has knowledge
about the concept Client. Fig. 3 illustrates various
types of the relation ‘‘knows about’’. Formally,
retainers in specific roles are required to know about
concepts and instances defined by their role defini-
tion. Individuals inherit the knowledge requirements
of the roles they play in the organization (e.g., all
individual project managers are required to possess
the basic knowledge that is defined in the job
description of the project manager role). In addition,
Fig. 3. The knowledge relationship.
Generally, a relation between concepts and/or instances is
represented by an n-place predicate that is true and relevant in
the universe of discourse (e.g., ‘‘John works in sales’’!Works
(John, sales)).
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562556
individuals may hold knowledge that is not required
by their formal role but that they have acquired
somehow. In transactive memory terminology, this
knowledge is assigned to them because of their
personal interests or because of some informal role
that they have acquired within the group. For
example, a group member in the accounting division
may be assigned to be the group’s specialist on
technological innovations related to the division’s
work. The formal roles played by memory retainers
have an impact on the way knowledge can be
retrieved from organizational memory. Specifically,
the existence of formal division of knowledge
facilitates the retrieval of knowledge in transactive
memory systems [12].
Using the definitions above, we now can create
simple transactive memory queries. For example,
we can query who are all the retainers that possess
knowledge about project ‘x’? Or what knowledge is
retained by a specific retainer ‘r’? Note that these
queries can relate to role knowledge or to instance
knowledge, to concepts or to instances of concepts.
Examples for transactive memory queries enabled by
the knowledge relation are:
� Role knowledge about a concept: What does a
project manager know about ‘‘projects’’? Which
roles possess knowledge about ‘‘clients’’?� Role knowledge about instance of a concept: What
does a project manager know about a specific
project?� Role instance knowledge about a concept: What
does Jane know about ‘‘projects’’? What individ-
uals know about clients?� Role instance knowledge about an instance of a
concept: What individuals know about the client x?
By now we have formalized the two ‘traditional’
dimensions proposed by transactive memory theory:
—the subject of the knowledge and the retainer of the
knowledge. We are able to create a meta-memory
directory as well as basic transactive memory queries
to locate knowledge in organizational memory. How-
ever, as discussed earlier, in order to capture tacit
group knowledge that exists in small groups, we need
to add a third—meta-knowledge—dimension to the
directories of meta-memories. We now turn to de-
scribe this meta-knowledge.
The final dimension in the proposed directory of
meta-memory includes the four types of meta-knowl-
edge described in Table 1. We further distinguish
between two groups of meta-knowledge elements.
The first group includes conceptual meta-knowledge
that represents the knowledge subjects and their
relationships and is represented by the ontology.
This meta-knowledge appears in Fig. 2 as an entity
type representing concepts or instances or as specific
relationships between these entities. The second
group includes descriptors of roles and knowledge
relationships, specifically the cognitive, descriptive,
and persuasive components of meta-knowledge. To
demonstrate, consider the following example:
A consultant wishes to find knowledge on how to
adjust a specific function in an information system.
When asked to select a potential knowledge source
to turn to (among 10 possible sources) the
consultant enquires about the expertise of the
knowledge source in this domain, the accuracy and
validity of the knowledge, and the cost of attaining
the knowledge.
In this example, the consultant (an instance of a
role) is searching for knowledge on a specific instance
of the concept ‘information system’. The meta-knowl-
edge requested by the consultant is specific to the
problem at hand and includes persuasive and descrip-
tive meta-knowledge about the knowledge requested.
The conceptual meta-knowledge is represented here
by the concept ‘information system’ and the instanti-
ation relationship. In addition, the consultant may
wish to know if there is other relevant knowledge
that she should be aware of. This again requires the
use of conceptual meta-knowledge to identify related
concepts and/or instances.
To summarize, the traditional transactive memory
model focuses on two dimensions: the subject and the
retainer of the knowledge. All other information is
assumed to be tacit, in group-members minds. Our
model intends to make this tacit dimension explicit via
the different aspects of meta-knowledge. Therefore,
our meta-memory directory is three -dimensional, as
illustrated by the cube in Fig. 4.
Every cell in the conceptual cube in Fig. 4 repre-
sents a piece of transactive knowledge. For example:
‘‘The hotel reservations manager knows about the
Fig. 4. The extended meta-memory directory.
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562 557
process of room reservations for large groups’’. Every
cell is therefore characterized by an ordered triplet that
includes subject of knowledge, retainer of knowledge,
and a vector of meta-knowledge components. The use
of ontology as the domain of knowledge subjects
implies that various cells in the knowledge cube are
conceptually tied to each other thus providing addi-
tional meta-knowledge such as knowledge comple-
mentarities or specializations. We claim that using the
cube structure as the basis of the organizational trans-
active memory directories can facilitate updating and
retrieval of knowledge from the organizational mem-
ory. This is explained in the next section.
3.3. Directory updating, information allocation, and
memory retrieval
We follow Wegner [30] to show how the trans-
active memory information system can support the
three defining activities of transactive memory: direc-
tory updating, information allocation, and memory
retrieval. These activities are illustrated in Fig. 5.
As new knowledge enters the organization, it
should be allocated to the right retainer. Using the
knowledge cube, the recipient of the knowledge can
identify the subject of the knowledge and locate the
retainer that has a declared expertise on this topic. The
knowledge can then be allocated to this retainer. The
cube structure enables a very precise allocation of
knowledge. If more than one retainer exist that has a
declared expertise on the topic of the new knowledge,
the allocation of the knowledge can be targeted to the
most suitable retainer by examining the additional
meta-knowledge. For example, many people may
indicate knowledge concerning the company’s prod-
ucts; however, engineering knowledge will be direct-
ed to the production manager while knowledge
regarding competing products will go to the sales
and marketing manager. The knowledge relationship
(‘‘knows about’’) defines the topics in which a spe-
cific retainer is proficient and the organizational
ontology relates these topics to more general areas
of expertise.
Similarly, when an individual seeks some knowl-
edge, they can query the meta-memory directory to
identify the best knowledge for their need. The cube
structure and the transactive memory queries specified
in the previous section support fast retrieval of spe-
cific knowledge from the organizational memory. For
example, we can query the directory for the people
with knowledge about the company’s products. This
in itself may not be different from any other search
engine that exists today. However, using the meta-
knowledge directories, we can then focus the results
on a specific subset proposed by the organizational
ontology (some specialization of the concept ‘prod-
uct’) and reduce the set of results. Finally, we can
select the most suitable source of knowledge by
Table 2
Examples for the use of meta-knowledge in knowledge retrieval
Search goal Search engine
(using keywords)
Transactive memory approach
(using meta-knowledge)
Research a
new topic.
For example,
search for
‘‘research
grant
application’’
Keyword search
brings numerous
results that the
knowledge
seeker needs
to sort through
The ontology (conceptual
meta-knowledge) provides
a set of relevant terms and
the relations among them,
so the knowledge seeker
can refine their search based
on their specific goals. For
example, the search term
‘‘research grant’’ is tied to
‘‘granting agencies’’,
‘‘ethical review process’’,
and ‘‘funding rules’’
Find the best
result from
the search
Results are often
sorted by some
relevance ranking
Results can be sorted by
various criteria based on the
meta-knowledge dimension.
For example, by relevance,
accuracy, or currency of the
knowledge itself, or by
source characteristics such
as background, credibility,
or expertise. The knowledge
seeker can make a more
informed selection of the
knowledge
Fig. 5. Information allocation and retrieval using the meta-memory directory.
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562558
examining the additional meta-knowledge. For exam-
ple, if we are looking for knowledge concerning the
sales of a specific subset of our products, we can
query the system for all the retainers of knowledge
about this subset and then select the retainer that we
feel is most suitable to answer this query. To illustrate
how meta-knowledge would improve the retrieval of
knowledge from organizational memory, consider the
example of a researcher looking for information on
how to write a successful grant proposal. Table 2
shows the differences in the search results between
traditional search engines and the transactive memory
approach.
The main challenge posed by this approach is the
ability to keep the meta-memory directories updated.
Some of the updates can be done automatically since
much of the knowledge in the organization is re-
quired by some formal job definitions (as illustrated
in Fig. 3). The updates of expertise on knowledge
that is not formally required by a specific role will
still be the responsibility of organizational members
and cannot be easily automated. Studies on trans-
active memory mechanisms in small groups indicate
that groups discuss and encode much of the trans-
active knowledge at the early stages of their work. As
the work progresses, the groups have additional
encoding cycles that are initiated by questions about
a task and are followed by the identification and
encoding of relevant experts and coordination of this
new transactive knowledge with previous knowledge
[22]. Following this, we propose that the automated
transactive memory will be set initially to reflect
Table 3
Alleviating the problems associates with OM management
Problem with OM Proposed solution
Knowledge in the OM
is contextualized
Conceptual meta-knowledge
can help reconstruct the original
context of the knowledge
Retainers of OM may be
in different locations
and their memories might
be difficult to combine
Descriptive meta-knowledge
locates different knowledge
retainers as well as provides
information on the most up-to-
date version of the knowledge
Knowledge may be tacit Tacit knowledge can be
located and evaluated using
meta-memory and the four
types of meta-knowledge
The content of OM
changes often
The responsibility to ‘‘keep
track of knowledge’’ is now at
the hand of the relevant expert.
The system maintains only
meta-memory, which is less
volatile than the knowledge itself
Knowledge seekers require Descriptive and persuasive
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562 559
existing knowledge, and then will be updated accord-
ing to the encoding cycles in the various groups. We
also note that the volatility of meta-knowledge is likely
to be lower than the volatility of knowledge itself.
Hence, periodical (rather than continuous) updates
should not diminish considerably the effectiveness of
the system.
A second challenge for encoding transactive
knowledge relates to the combination of memories
of different groups within the organization and the
resolution of conflicting knowledge between these
groups. The existence of ontology in the design of
the system can resolve some of these problems by
identifying related concepts and preventing the ambi-
guity of concepts. Furthermore, it can be used to
identify different instances of knowledge about the
same concepts. However, the decisions on resolving
conflicting knowledge instances are a matter of man-
agerial practices.
a measure of the retainer’slegitimacy and reliability
when retrieving knowledge
meta-knowledge is included in
the meta-memory directory
4. Discussion
The purpose of the model described above is to
demonstrate the ability of information technology to
extend the notion of transactive memory to large
groups. Using technology we can create computer-
supported knowledge allocation process that is based
on the meta-knowledge provided about each of the
knowledge retainers. Similarly, the system can assist
in the retrieval of knowledge from organizational
memory.
We claim that using information technology as
suggested here can help alleviate some of the prob-
lems of managing organizational memories described
in Section 1. To show this, we present in Table 3 the
set of problems described earlier and explain how a
transactive memory information system can help
alleviate them.
An additional benefit from using the transactive
memory approach is in its potential benefits for group
performance. We claim that the use of meta-memory
will lead to improvement of knowledge adoption due
to its effect on knowledge transfer from organization-
al memory to individual users. To demonstrate this,
we use the notion of ‘‘‘knowledge stickiness’’’, a
measure of the difficulty of transferring knowledge
[27].
Studying knowledge transfer in organizations, Szu-
lanski [27] identified eight predictors for knowledge
stickiness. These predictors concern characteristics of
the source of the knowledge, the recipient, the con-
text, and the message (knowledge) itself. The predic-
tors were measured in the settings of best practices
transfers within organizations and the most significant
predictors of stickiness were the inability to pinpoint
the reasons for a success or failure of replicating
knowledge (causal ambiguity) and the recipient’s lack
of ability to identify, value, and apply knowledge
(absorptive capacity). Other significant factors were
the perceived utility of the transferred knowledge,
perception of source as unreliable, lack of motivation
of source and recipient, lack of retentive capacity of
recipient, unsupportive organizations, and distant re-
lationship between parties.
We suggest that by exposing the receivers of
knowledge to some knowledge about the knowledge
(meta-knowledge) before the knowledge transfer takes
place, it is possible to reduce stickiness. For example,
using meta-knowledge, we hope to provide individu-
als with information that will enable them to select
knowledge that they can interpret most easily based
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562560
on their prior knowledge, thus increasing their absorp-
tive capacity. Similarly, the availability of persuasive
meta-knowledge can increase perceived reliability of
the knowledge source.
4.1. Empirical support
Several studies exist that provide empirical support
to the potential benefits of the transactive memory
approach and specifically—the use of meta-knowl-
edge to support the management of organizational
memory. In this last section, we describe some of
these studies as well as propose research questions for
future empirical work.
The first group of empirical studies examines the
benefits of transactive memory in a group setting.
These studies show that transactive memory naturally
develops in groups that train together, and that the
existence of transactive memory leads to improved
group performance and better problem problem-solv-
ing mechanisms [17,18,22]. These early studies on
transactive memory used observations as the measure-
ment method and determined the existence of trans-
active memory using three specific measures: memory
differentiation, task coordination, and task credibility.
A more recent study demonstrated the development of
transactive memory in virtual teams and the positive
effect of this transactive memory on the team’s
performance [33]. Transactive memory was measured
by a three three-item scale, asking respondents about
their perceptions of the team’s transactive knowledge.
This study provides important support for the ideas
presented in our paper as it identifies the development
of transactive memory in the context of computer-
based communications.
A second group of studies providing support for
technology-supported transactive memory looks at
the potential benefits of meta-knowledge for knowl-
edge adoption in organizations. These studies dem-
onstrate the importance of cues, such as source
credibility, for information adoption [29]. In addition,
they empirically identify the specific knowledge and
source attributes that should be included as meta-
knowledge (cues) in order to increase the likelihood
of knowledge adoption [19]. These knowledge and
source attributes mainly fall under the descriptive
and persuasive meta-knowledge types described ear-
lier in our paper and include such items as the
accuracy and currency of the knowledge and the
trustworthiness, knowledgeability, and willingness-
to-help of the knowledge source. Finally, the studies
show that the probability of information seeking will
increase when the seeker is able to identify what
other people know, when the seeker is able to
evaluate the other person’s knowledge, and when
the seeker can gain timely access to the other person’s
knowledge [6].
Following these studies and the conceptual ap-
proach presented in our paper, we propose that future
empirical studies examine the following questions:
(1) Can technology enable the creation of transactive
memory in cases where it does not develop
naturally?
(2) Will this artificially created transactive memory
lead to the same benefits—namely improved
performance and better problem problem-solving
in the organization—as it does in small groups?
The studies described in this section set the ground
for answering these two questions by defining meas-
ures of transactive memory and performance as well
as by identifying the meta-knowledge that would
effectively influence the information adoption deci-
sion of individuals. This meta-knowledge was pro-
posed in our paper as the third dimension of the meta-
memory directory. The model proposed in the current
paper can provide guidance to future empirical stud-
ies, in particular in suggesting experimental settings.
First, we develop the specific design of the organiza-
tional memory system and the three-dimensional
directories of meta-memory to be tested. Second, we
distinguish between three types of knowledge—e.g.,
role knowledge, instance knowledge, and transactive
knowledge—that are related to the development of
transactive memory. Specifically, we propose that the
existence of instance knowledge might inhibit the
development of transactive memory in settings where
this knowledge is not easily available to all group
members. We expect that in settings where instance
knowledge cannot be fully obtained an information
system based on the transactive memory approach
proposed here would improve the sharing of knowl-
edge. This prediction can be tested empirically by
creating situations in which transactive memory
would not develop naturally by controlling the avail-
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562 561
ability of instance knowledge to group members and
testing the effect of a system based on the principles
proposed in this paper.
5. Conclusion
In this paper, we have addressed the use of infor-
mation technology to obviate the problems involved
in the use of organizational memory. We suggest an
approach to the design of the memory-supporting
(mnemonic) layer of an organizational memory infor-
mation system. Our approach is rooted in transactive
memory theory, that is, in a theory of collective
memory of small groups. Specifically, we suggest that
the role of technology is to enable the extension of the
transactive memory mechanisms to the whole organi-
zation. The architecture we propose is based on three
dimensions of meta-memory. First, the two ‘‘tradi-
tional’’ dimensions of transactive memory—subject
and retainer. Second, as a replacement to the tacit
aspect of meta-memory, we propose the use of a meta-
knowledge dimension that includes three components:
descriptive, cognitive, and persuasive. These compo-
nents formalize, respectively, what the knowledge is
about, what the individual believes they are capable of
knowing, and what the individual knows about others’
knowledge and expertise. The three components,
when included explicitly in an OMIS can serve to
help allocate and retrieve knowledge more effectively
by informating individuals in the organization. In
addition, the availability of these components of
meta-knowledge can serve to reduce stickiness and
to improve knowledge transfer.
We believe the justification for this approach is
that it provides technology support for existing
organizational mechanisms. The use of a rich set
of meta-knowledge components also provides a
possible solution to the difficulties of knowledge
search and retrieval. The use of ontology has been
shown in various projects to be useful for the
preservation of the context of knowledge and thus
can facilitate an efficient management of organiza-
tional memory. Finally, we have pointed out that
existing empirical studies provide support for the
potential benefits of our approach and proposed
empirical work that will explicitly test the proposed
design.
Acknowledgements
The research was supported in part by a grant from
the Social Sciences and Humanities Research Council
of Canada.
References
[1] M.S. Ackerman, C.A. Halverson, Reexamining organizational
memory, Communications of the ACM (2000 Jan.) 58–64.
[2] M. Alavi, D.E. Leidner, Review: knowledge management and
knowledge management systems: conceptual foundations and
research issues, MIS Quarterly 25 (1) (2001) 107–136.
[3] V. Anand, C.C. Manz, W.H. Glick, An organizational memory
approach to information management, Academy of Manage-
ment Review (1998 Oct.) 90–111.
[4] L. Argote, R. Ophir, Intraorganizational learning, in: J.A.C.
Baum (Ed.), The Blackwell Companion to Organizations,
Blackwell, Oxford, England, 2002, pp. 181–207.
[5] R. Basch, Measuring the quality of the data: report on the
fourth annual SCOUG retreat, Database Searcher 6 (8) (1990)
18–24.
[6] S.P. Borgatti, R. Cross, A relational view of information seek-
ing and learning in social networks, Management Science 49
(4) (2003) 432–445.
[7] J.S. Brown, P. Duguid, Organizational learning and communi-
ties of practice: toward a unified view of working, learning, and
innovation, Organization Science 2 (1) (1991) 40–57.
[8] J.S. Brown, P. Duguid, Organizing knowledge, California Man-
agement Review 1998 (Spring) 796–809.
[9] J.H. Flavell, Metacognition and cognitive monitoring, a new
area of cognitive—developmental inquiry, American Psychol-
ogist 34 (10) (1979) 906–911.
[10] E.T. Hall, Beyond Culture, Anchor Press/Doubleday, New
York, 1976.
[11] M. Higgins, Meta-information, and time: factors in human
decision making, Journal of the American Society for Infor-
mation Science 50 (2) (1999) 132–139.
[12] A.B. Hollingshead, Perceptions of expertise and transactive
memory in work relationships, Group Processes & Intergroup
Relations 3 (2000) 257–267.
[13] C. Hovland, I.L. Janis, H.H. Kelly, Communication and Per-
suasion: Psychological Studies of Opinion Change, Yale Univ.
Press, New Haven, 1953.
[14] Y. Kalfoglou, T. Menzies, K.D. Althoff, E. Motta, Meta-
knowledge in systems design: panacea or undelivered
promise? Knowledge Engineering Review 15 (4) (2000)
381–404.
[15] W.A. Katz, Introduction to Reference Work, McGraw-Hill,
New York, 1992.
[16] A. Mintu-Wimsatt, J.B. Gassenheimer, The moderating
effects of cultural context in buyer– seller negotiation, Jour-
nal of Personal Selling & Sales Management 20 (1) (2000)
1–9.
[17] R.L. Moreland, L. Myaskovsky, Exploring the performance
D. Nevo, Y. Wand / Decision Support Systems 39 (2005) 549–562562
benefits of groups training: transactive memory or improved
communication? Organizational Behavior and Human Deci-
sion Processes 82 (1) (2000) 117–133.
[18] R.L. Moreland, L. Argote, R. Krishnan, Socially shared cog-
nition at work: transactive memory and group performance,
in: J.L. Nye, A.M. Brower (Eds.), What’s Social About Social
Cognition?, SAGE Publications, Thousand Oaks, CA, 1996,
pp. 57–86.
[19] D. Nevo, I. Benbasat, and Y. Wand, The role of knowledge
and source attributes in the knowledge adoption decisions of
individuals. Working Paper, University of British Columbia,
2003.
[20] I. Nonaka, A dynamic theory of organizational knowledge
creation, Organization Science 5 (1) (1994) 14–37.
[21] R. Plant, R. Gamble, Using meta-knowledge within a multi-
level framework for KBS development, International Journal
of Human–Computer Studies 46 (1997) 523–547.
[22] D. Rulke-Liang, D. Rau, Investigating the encoding process of
transactive memory development in group training, Group &
Organization Management 25 (2000) 373–396.
[23] D. Schwartz, When e-mail meets organizational memories:
addressing threats to communication in a learning organiza-
tion, International Journal of Human–Computer Studies 51
(3) (1999) 599–614.
[24] E.W. Stein, Organizational memory: review of concepts and
recommendations for management, International Journal of
Information Management 15 (1) (1995) 17–32.
[25] E.W. Stein, V. Zwass, Actualizing organizational memory
with information systems, Information Systems Research 6
(2) (1995) 85–117.
[26] D. Stoker, A. Cooke, Evaluation of networked information
sources, in: A.H. Helal, J.W. Weiss (Eds.), Information Super-
highway: the Role of Librarians, Information Scientists and
Intermediaries: Proceedings of the 17th International Essen
Symposium, 24–27 October 1994, 1995Universitatsbiblio-
atsbibliothek Essen, Essen, 1995, pp. 287–312.
[27] G. Szulanski, The process of knowledge transfer: a diachronic
analysis of stickiness, Organizational Behavior and Human
Decision Processes 82 (1) (2000) 9–27.
[28] J.P. Walsh, G.R. Ungson, Organizational memory, Academy
of Management Review 16 (1) (1991) 57–91.
[29] S. Watts Sussman, W.S. Siegal, Information influence in
organizations: an integrated approach to knowledge adoption,
Information Systems Research 14 (1) (2003) 47–65.
[30] D.M. Wegner, A computer network model of human trans-
active memory, Social Cognition 13 (3) (1995) 319–339.
[31] D.M. Wegner, Transactive memory: a contemporary analysis
of the group mind, in: B. Mullen, G.R. Goethals (Eds.), The-
ories of Group Behavior, 1987Springer-Verlag, New York,
1987, pp. 185–208.
[32] F. Wijnhoven, Development scenarios for organizational
memory information systems, Journal of MIS 16 (1) (1999)
121–146.
[33] Y. Yoo, P. Kanawattanachai, Development of transactive
memory systems and collective mind in virtual teams, Inter-
national Journal of Organizational Analysis 9 (2) (2001)
187–208.
Dorit Nevo is an assistant professor of
Information Systems at the Schulich
School of Business, York University, Tor-
onto, Canada. She received her PhD in
Management Information Systems from
the University of British Columbia, her
MSc, in Economics from the Technion,—
Israel Institute of Technology, and BA in
Economics and Business Administration
from the University of Haifa, Israel. Her
current research interests include require-
ments analysis, and design and evaluation of Knowledge Manage-
ment Systems.
Yair Wand is CANFOR Professor of MIS
at the Sauder School of Business, The
University of British Columbia. He re-
ceived his DSc in Operations Research
from the Technion,— Israel Institute of
Technology, his MSc in Physics from the
Weizmann Institute, —Israel, and his BSc
in Physics from the Hebrew University,
Jerusalem. His current research interests
include theoretical foundations for infor-
mation systems analysis and design, de-
velopment and evaluation of systems analysis methods, and
conceptual modeling.