Content uploaded by Claudia Roda
Author content
All content in this area was uploaded by Claudia Roda on Sep 10, 2013
Content may be subject to copyright.
Matching competencies to enhance organisational knowledge sharing:
An Intelligent Agent approach
i
Claudia Roda Albert Angehrn Thierry Nabeth
Senior Research Fellow Director of CALT Research Fellow
Claudia.Roda@insead.fr Albert.Angehrn@insead.fr ThierryNnabeth@insead.fr
Tel. +33 (0)1 60724451 Tel. +33 (0)1 60724361 Tel. +33 (0)1 60724312
Fax +33 (0)1 60745544 Fax +33 (0)1 60745544 Fax +33 (0)1 60745550
INSEAD - Centre for Advanced Learning Technologies
Bd. De Constance, F-77300 Fontainebleau France
http://www.insead.fr/calt/
Contact Author: Thierry Nabeth
Abstract
This paper describes an agent-based system designed to guide, monitor, and stimulate managers
towards the understanding of knowledge management concepts and the adoption of knowledge
management practices in organizational contexts. In particular, we focus on the mechanisms needed
to support the adoption of knowledge sharing processes, which are vital to knowledge management
practices and often encounter the strongest resistance. A competency matching function, integrated
in the presented agent system, allows dynamic matching of suppliers and consumers of information.
Profiling (of people as well as of other resources) is used to supply knowledge-sharing partners
with contextual information necessary for efficient knowledge exchange. Finally, we discuss
another distinctive feature of the system, its capability to propose knowledge sharing actions within
the context of user’s activities and in relation with their level of adoption of knowledge sharing
practices.
Keywords: knowledge management, knowledge sharing, software agents, change management,
competency matching.
1. Introduction
Industrialised countries, over the last decade, have increasingly experienced a shift from a
production-based economy to a service-based economy. As a consequence companies have come to
value intellectual capital as much as their physical assets. Knowledge distributed within
organisations (whether in formal storage systems, or in the head of people) is a part of this
intellectual capital and it can retain its value only if it is well managed. Companies have therefore
recognised the need to engage in some knowledge management processes [Davenport 98,
Koulopoulos 97, Leonard 95, Nonaka 95] aimed at encouraging efficient acquisition, organisation,
sharing, and use of knowledge. Many digital tools have been developed in the effort to support
some parts of the knowledge management processes. However, it is widely recognised that one of
the major barriers to the adoption of knowledge management practices is (active or passive)
resistance to change [Dore 01, Lawel 01]. Knowledge workers are required to systematically
implement practices (e.g. sharing knowledge, acquire and use knowledge produced by others) that
are new to them or that they currently implement only sporadically or that plainly contrast with
what they are used, or they perceive is in their personal interest, to do.
In section 2, we first describe an agent based system (K-InCA – Knowledge Intelligent
Conversational Agents) which addresses the problem of resistance to change by providing to each
individual (human) user a non-human entity (software agent) capable of guiding the transition from
their current practices, to the knowledge management practices specified by their organisation. In
particular, we focus on the mechanisms needed to support the adoption of knowledge sharing
processes, which are vital to knowledge management practices and often encounter the strongest
resistance. In sections 3 and 4 we discuss a competency matching function we are integrating in the
K-InCA system to support and enhance the adoption of knowledge sharing practices in
organizations.
2. Overview of the K-InCA system
The K-InCA system behaves as a personal Knowledge Management coach guiding, stimulating and
monitoring the user towards the understanding of knowledge management concepts and the
adoption of knowledge management practices. K-InCA agents can be seen as experts on
organisational behaviour and change management assisting users in the transition from their current
working habits to new habits that integrate some new behaviour (e.g. good knowledge management
practices, entrepreneurial attitude, etc.). In our first implementation, the user initiates a K-InCA
coaching sessions after a practice period. During the practice period the user performs his normal
activity. During the coaching session the user reports his activities and interacts with the K-InCA
agent who supplies advice and stimulus towards the adoption of a set of given knowledge
management behaviours. As the system develops, K-InCA agents will be able to directly observe
some of the user’s actions (e.g. sending emails, browsing, etc.) and they will integrate direct
observation with user’s statements of activity supplied in the coaching session.
Agent’s environment
Personal
agent
Resource
Model
User
Interface
Organizational context
User Model
Learning
Domain
Model
Intervention
Model
Coordination
Mechanism
Figure 2 – An overview of the K-InCA system
The overall architecture of the K-InCA system is shown in Figure 1. The agent coordination
mechanism (allowing agents to communicate and coordinate) and the user interface
ii
are not
discussed in this paper. The domain of expertise of the system is represented in the learning domain
model. The current domain of expertise of K-InCA is knowledge management, however other
learning domains can be “plugged-in” the K-InCA system. The user model and resource model are
discussed in section 3. In this section we discuss the current intervention model which is based on a
specific model of the change process.
The system models the change process as defined by Rogers [Rogers 95]. Rogers introduces the
innovation-decision process as “the process through which an individual […] passes (1) from first
knowledge of an innovation, (2) to forming an attitude towards the innovation, (3) to a decision to
adopt or reject, (4) to implementation of the new idea, and (5) to confirmation of this decision”
[ibid, p.161]. We use the innovation-decision process as a model for organisational learning
[Manzoni 98, Angehrn 97]. This model represents the change process as composed of five learner
states:
• unaware - the user does not know about the innovation
• awareness - the user is introduced to the innovation (i.e. some knowledge management practice)
• interest - the user actively seeks information about the innovation or demonstrates interests in
information about the innovation "pushed" by the system
• trial / appraisal - the user actively experiments in applying the innovation
• adoption - the user incorporates the innovation in his normal practices
K-InCA agents monitor and guide the user as he goes through these phases for each one of the
desired behaviours/practices (those included in the knowledge management process). Figure 2
illustrates the strategies applied by the agent to guide the user towards adoption.
Figure 2 - Strategies guiding the adoption process
Awareness is raised by supplying information about the desired behaviours (introduction strategy).
Interest is generated by situating the new knowledge and demonstrating to the learner that it is
applicable to his situation (situate strategy). The transition to trial is pursued by proposing practical
examples/exercises/problems that the learner can resolve, by changing his behaviour, using the new
knowledge (practice strategy). Finally, adoption is solicited by evaluating the trial process and
demonstrating the advantages of the new skill, process, behaviour or knowledge (evaluation
strategy).
The final goal of the agent is to bring the user to adopt all the desired behaviours. In order to
achieve this goal, K-InCA agents react to the current user activity on the basis of the information
contained in the domain model and the user model, as well as interacting with other agents (see
figure 1 and section 3). Target behaviours, those that the user should adopt, are described in the
learning domain model. This model, which can be updated or completely replaced by selected users
/ tutors, includes a description of all the behaviours (what the behaviour is, why and for whom it is
relevant, how to practice it, which are the benefits for the user who adopts the behaviours, etc.),
Unaware
Introduce
Awareness
Situate
Interest
Practice
Trial
Evaluate
Adoption
behaviour in the organizational context. Behaviour descriptions are used by the agent to supply the
user with the information necessary to achieve awareness and interest. Behaviour relationships
(such as “respond to colleagues request for information” is a sub-behaviour of “share knowledge”)
allow the agent’s evaluation of the adoption state for one behaviour, given the adoption state for
another. The actions associated to behaviours form the basis for the agent’s recognition of relevant
behaviours with reference to a [sequence of] user’s action (i.e. behaviour that have been assumed,
or should have been assumed, by the user).
3. Enhancing knowledge sharing conversations
As mentioned above, one of the vital knowledge management practices which encounters the
strongest resistance in organisations is knowledge sharing. Obstacles to the adoption of knowledge
sharing behaviours generally include: organisational barriers, lack of transparency, culture, habits,
lack of incentives, etc. For instance, knowledge may be available (existence of a supplier) but
access to it may be limited by factors such as the consumer/demander unawareness about it, or by
the unwillingness of the supplier to make the information available. A recent survey on the
implementation of knowledge management practices in banks and insurance companies [Dore
2001] reports that the main barriers to knowledge sharing are the “lack of understanding of the
benefits derived from knowledge sharing” and the “technology inadequacies” due to the fact that
“knowledge is held in too many formats and repositories”. In order to achieve the widespread
adoption of efficient, timely and relevant knowledge sharing practices, several conditions need to be
met. First of all suppliers and consumers must be dynamically matched so that consumers are aware
of possible suppliers and suppliers identify possible consumers. Secondly, the initiators of the
knowledge sharing process (whether consumers or suppliers) must be given enough information to
enable them to ask for knowledge or to package knowledge in a form that is understandable to the
knowledge sharing partner. Thirdly, the involved parties must be motivated enough to engage in the
knowledge sharing process. In order to meet this last requirement, K-InCA agents propose
knowledge sharing actions which are relevant to the current user activity and, when necessary,
supply the motivation information stored in the learning domain model. In order to meet the first
two requirements, the system stores and maintains, in the user and resource models, information
about knowledge sharing parties.
3.1. User and resource models supporting knowledge sharing conversations
The user model embedded in the K-InCA system includes a description of: (1) The adoption state of
the user (e.g. the learner has entered the interest phase for the behaviour share knowledge, or he is
in the trial stage for behaviour acquire knowledge from people outside the company, etc.) (2) The
user's type: personality and attitudes towards innovation (users are classified along five categories
which, in increasing order of resilience to change, are: innovators, early adopters, early majority,
late majority, and laggards [Rogers 95, p.262]). (3) The user's social network (personal network,
e.g. friends and acquaintances; and work environment, e.g. boss, colleagues, acquaintances, etc.).
(4) The user's preferences: expertise, learning goals, interests, communication modes, and learning
modes. (5) A record of previous user's actions. (6) A user profile: a presentation of the user. (7) The
knowledge management agenda. This last element is central in the K-InCA system in that it
mediates the interaction between the user and the agent and provides a concrete context to it. The
knowledge management agenda contains the list of knowledge management actions proposed by
the agent to the user during the coaching session.
The information contained in the user model is dynamically updated by the agent. Particularly
relevant for the competency matching mechanism is the fact that the agent updates the user’s
preferences and social network on the basis of the user’s actions (and his declarations). For
the agent adds project Alpha to the user’s interests or it raises the level of interest of the user for
that project. Similarly, Mr. Smith is added to the user’s acquaintances or, the level of acquaintance
of the user to Mr. Smith is raised. As a result, user’s preferences and social network are assigned
weights so that the agent can represent facts such as the user being “very interested” in a given
subject and only “marginally interested” in another one; or the user “just knowing” Mr. Smith
whilst being “well acquainted” with Mr. Black. All agents use the same weight assignment
procedure (which cannot be detailed here for lack of space) so that, for instance, it is possible to
compare the expertise level of several users on the same subject.
A set of Resource Manager Agents (RMA) are associated to resources such as information
repositories about external people holding relevant competencies, formal/non-human organisational
systems, knowledge bases or training systems, etc. Resource Manager Agents maintain a resource
model and encapsulate their resources making them visible from within the K-InCA system. The
resource model includes, amongst others, the following information: (1) a keywords description of
the information that the resource may supply, (2) a keywords description of the information that the
resource may be interested in acquiring, and (3) instructions on how the resource can be accessed
(4) the resource type (e.g. external people, formal system, knowledge base, etc.)
4. Competency matching
Competency matching is achieved in the K-InCA system through a distributed consultation
mechanism amongst the software agents associated with each user (Change Agent) and the
Resource Manager Agents (RMA). The competency matching consultation is initiated by one of the
Change Agents when it recognises that its user should act as a consumer or supplier of some
knowledge. When this happens, the Change Agent issues a distributed query asking respectively for
suppliers or consumers of that knowledge. Each one of the agents taking part in the consultation is
capable of deciding if the associated resource (whether human or not) could act as supplier or
consumer in that knowledge exchange and responds to the request. Queries for competency
matching may take two forms: (1) User X is looking for consumer of knowledge Y under
constraints Z; or (2) User X is looking for producers of knowledge Y under constraints Z. The
knowledge identifier Y may be, in the simplest instance a (set of) keyword(s). The constraints Z
allow the querying agent to specify the type of producer / consumer it is looking for. Consider, for
example the exchange of figure 2. If the agent recognises the need to situate its suggestion “Have
you though about acquiring information outside the company?” it may issue the query “User userid
is looking for producers of Knowledge about Project Alpha under constraint external people”.
Responses to competency matching queries may either indicate availability for production or
consumption (depending on the original request) and take different formats when generated by
Change Agents or Resource Manager Agents. In particular, Resource Manager Agents respond
indicating the resource availability and the access instructions. For instance, by matching the query
content against the resource model, a Resource Manager Agent may respond with: “External user
Mr.Black is a supplier for knowledge about Project Alpha , ‘Mr. Black is the director of Project
Alpha for the University of Agag, he also collaborates in the EU standardization committee. He can
be reached at 2233454 or email Black@ac.Agag.gr’ “.
Change Agents are capable of supplying a more informative response to competence matching
queries. For instance, responses to requests for suppliers include: the level of expertise of the
supplier in the subject; the level of acquaintance of the supplier with the consumer, his preferred
communication modes, his interests, his user profile, his social network, etc. After receiving the
responses to its competency matching query, the change agent will select the consumers or
suppliers best suited to engage in the knowledge sharing process and it will collate the responses
interested in the information you have found about product TK3. Paul works in the project Alpha
and you seem to know him quite well. Leo shows a very high interest level for this product, he
should be contacted by email at Wilter@company.com. You can access the profile of the above
users by clicking on their name. You should also store information about product TK3 in the special
products database which can be accessed […], rewards are given for certain types of contributions
to this database see […]”.
The dynamic elicitation of knowledge (e.g. user interests and expertise) built in the K-InCA system
requires a maintenance mechanism which is partially automated. When users report their actions
ii
,
e.g. “I have talked to Mr. Smith about project Alpha” they may introduce elements unknown to the
system. For example, the Change Agent may not know who Mr. Smith is, nor what the project
Alpha is. A repository of classified known people and information objects is maintained by
periodically collecting keywords and resource identifiers from Resource Managers and system
administrators. Unknown objects are collected for classification. The classification system allows to
specify facts such as object1 is the same as object2, or object3 is more specific than object4. This
mechanism controls the system as it grows and learns about the environment.
5. Conclusions
This paper has introduced the functionalities and information used by K-InCA agents in order to
support knowledge sharing. We believe that such agents, which are currently under development,
can be particularly effective thanks to their ability to propose knowledge sharing practices within a
context where such practices have been recognised to be relevant for the user as well for the
organization. Furthermore K-InCA agents are able to supply users with the practical information
necessary to efficiently implement knowledge sharing processes within an organization.
References
Angehrn, A., Nabeth, T., Razmerita, L., Roda C. (2001) K-InCA: Using Artificial Agents for
Helping People to Learn New Behaviours, Proceedings of IEEE International Conference on
Advanced Learning Technologies - ICALT, August 2001, Madison, USA, to appear.
Angehrn A., Nabeth T. (1997) Leveraging Emerging Technologies in Management Education:
Research and Experiences, European Management Journal, Vol.15, No. 3, pp. 275-205, June
1997
Dore L., (2001) Winning Through Knowledge: How to Succeed in the Knowledge Economy, Special
Report by the Financial World. The Chartered Institute of Bankers in Association with Xerox.
London: March 2001.
Davenport, T. H. and Laurence P. (1998) Working Knowledge: How Organizations Manage What
They Know, Harvard Business School Press, 1998
Koulopoulos, T. M., Spinello, R. A. and Wayne T. (1997) Corporate Instinct: Building a Knowing
Enterprise for the 21st Century, Van Nostrand Reinhold, 1997
Leonard-Barton, D. (1995) Wellsprings of Knowledge: Building and Sustaining the Sources of
Innovation, Harvard Business School Press, 1995
Lawton G., (2001) Industry Trends: Knowledge Management: Ready for Prime Time ?, Computer,
Vol. 34, No. 2, February 2001.
Manzoni, J.F., Angehrn, A., (1998) Understanding organizational dynamics of IT-enabled change:
a multimedia simulation approach Journal of management information systems, vol. 14, no.
3, (winter 97-98) pp. 109-140
Nonaka, I. and Hirotaka T. (1995) The Knowledge-Creating Company, Oxford University Press,
Rogers, Everett M. (1962) Diffusion of Innovation, 4th edition, Free Press, NY, 1995. First edition
by Everett 1962, same title.
i
The work described in this paper was sponsored by Xerox Corporation.
ii
For a more detailed description of the user interface of K-InCA see [Angehrn 2001]
