Un document du programme européen M.C.X./A.P.C.


Workshop on Knowledge Management


June 5th, 14h20 18h00

CEA Cadarache

Workshop on Knowledge Management


June 5th, 14h20 18h00

CEA Centre, Cadarache


14h20 - 14h30 :

Jean-Louis Ermine, CEA/DIST, Centre d'études de Saclay, France

Introduction to the workshop

Case studies :

14h30 - 14h50

Viktor L. Herna, Oberholzr O. Mauritz (University of Pretoria, South Afrika)

Management and Structuring issues of Knowledge Production and Refinement (A Case Study)

14h50 - 15h10

Aries S. (COFINOGA, France)

A Knowledge Management Case Study in a Loan Bank

15h10 - 15h30

Holtham C., Courtney C.: (City University of London, United Kingdom) :

Applied Knowledge Management : a case study in executive education

15h30 - 15h50 : Break

Aspects of KM :

15h50 - 16h10

Baets W. (Euro-Arab Management School, Granada, Spain)

Knowledge Networks for Service Companies : The Art of Supporting the "Emergent" with Information and Knowledge Technology

16h10 - 16h30

Carlsson S. (Lund University, Sweden)

Shared Knowledge Creation Spaces around Business Processes

16h30 - 16h50

Blanco S., Lesca H. ( University Pierre Mendès France, Grenoble, France )

Business Intelligence : a Collective Learning Process for the Selection of Early Warning Signals

16h50 - 17h10

Cardon A., Lesage F. (University Paris VI, INSA Rouen, France)

Dynamic Knowledge : the Sense of Communicated Knowledge in Distributed Systems

Management and Structuring Issues

of Knowledge Production and Refinement

(A Case Study)

Herna L Viktor Mauritz O Oberholzer

Department of Informatics

University of Pretoria

Pretoria 0002, South Africa


The acquisition of knowledge and the subsequent refinement and evolution thereof form an important part of any knowledge management project. Knowledge discovery from data (KDD) is increasingly being used to minimize the knowledge acquisition bottleneck. Here data mining tools are used to extract knowledge from raw data rather than using the traditional approach of obtaining the knowledge from the experts themselves. However, experience showed that many of these projects fail because the role of the stakeholders is not adequately acknowledged. This paper introduces an approach that implicitly incorporates the expert opinion into the KDD process by employing the soft systems methodology and cooperative learning.

The contribution will address the following issues, with reference to the lessons learned from applying the proposed approach to a case study concerning the knowledge management of a project conducted by a South African government institution:

1. Introduction

The successful completion of large-scale knowledge management projects mainly depends on the quality of the knowledge acquisition, refinement and evolution procedures (Adriaans, 1996). Some of the weaknesses associated with traditional manual knowledge acquisition approaches include the domain experts' problems in the expression of expertise, access problems, the desire to be able to acquire knowledge from examples and the time-consuming and thus expensive nature thereof (McGraw, 1990). Knowledge discovery from data (KDD) addresses the knowledge acquisition bottleneck by extracting knowledge from raw data, rather than using the lengthy, traditional approach of obtaining the detail knowledge from the experts themselves.

The KDD process consists of six main steps as introduced by Adriaans and Zantinge (1996). The first four steps, namely data selection, data cleaning, data enrichment and data coding, concern the preprocessing of data. The fifth step, i.e., the actual discovery stage of the process, is called data mining. The final reporting step is used to display the results to the environment. In the literature concerning knowledge discovery and the knowledge discovery process, the emphasis is almost always put on the fifth stage of the KDD process. That is, the stage that is specifically used to discover the new knowledge. People are mainly interested in how the knowledge is to be discovered, and tend not to spend time considering what they are looking for. The one thing that we as information system professionals should have learned over the past thirty years concerning the failure of IS projects, is that the lack of continuous user involvement in development projects lead to coordination and communication problems. These inevitably lead to the failure of the project or the overuse of resources. In order to obtain high quality, usable results from the KDD process, it is paramount that the expert opinion and objectives should be obtained using a structured approach such as the soft systems methodology (SSM).

Due to the complexity of organizations, there is an increasing realization by both researchers and practitioners in the field that knowledge management may only be effectively addressed by using a hybrid approach (Khosla, 1997). That is, the processes should be combined into a framework that facilitates cooperation and interaction between a variety of methods and stakeholders. One approach for setting up this framework is by considering educational techniques that facilitate the way in which humans cooperate when they learn. The approach discussed in this paper employs cooperative learning techniques that have been successfully applied in the human environment. These techniques are used to model the interaction between different data mining tools and human role-players, which are modeled as individual learners. The learners cooperate by exchanging data and/or discovered knowledge to find the best solution for the stated problem as defined during the application of the SSM.

The findings of this paper are based on the lessons learned from a case study concerning the knowledge management of a project conducted by a South African government institution. The main objective of the project was to determine the national strengths and weaknesses in a designated field as well as the existing skills and equipment base. The project included publicly funded ventures as well as the contribution and activities of the private sector.

The first section of this paper introduces the KDD process that incorporates cooperative learning techniques. In the second part of this paper, we discuss the need for human involvement and the use of systems methodologies with reference to the case study.

2.	A framework for KDD

The KDD process, as shown in figure 1, is subject to three phases. Note that the process is an iterative one, rather than a number of sequential steps. Therefore, the knowledge engineer usually experiments with more than one approach.

2.1. Data preprocessing The first phase concerns the collection, selection, possible enrichment and coding of data. It is important to determine whether the data are amenable to the data mining tools and to ensure that adequate data exists or can be obtained.

The data preprocessing phase is an ongoing process. Therefore, new data should frequently be incorporated into the framework. This should lead to a dynamically changing knowledge discovery environment.
2.2.	Knowledge discovery using cooperative learning

Combining data mining tools into a hybrid environment usually lead to more powerful results, since each data mining tool makes assumptions about a particular problem domain. These assumptions, also referred to as the inductive bias of the approach, lead to complementary results within the same problem domain. That is, one technique may succeed to extract knowledge from data where another has failed. Currently, the KDD process includes rule induction programs (Sestitto, 1994), decision trees (Mitchell, 1997), neural networks and the ANNSER approach (Viktor, 1998) that extracts rules from trained neural networks. Each of these approaches has its strengths and weaknesses. For example, decision trees may lead to bushy trees if the data contains noise, whereas neural networks perform exceptionally well. However, the numeric nature of neural networks makes them difficult to comprehend.

Our approach utilizes cooperative learning techniques as used in human environments to model the combination of heterogeneous data mining tools. Cooperative learning is based on the philosophy that the power and diversity of four or five minds are greater than that of one. That is, when people learn together in an effective group, they all benefit. Formally, cooperative learning may be defined as an organizational structure in which a group of students pursue a goal through collaborative effort. It can be considered as a formalized extension of group work. The students should not just complete a task in a group, but they are required to engage actively in the group learning process (Sharan, 1994). Readers are referred to Slavin (1991) for a detailed discussion of this topic. 

When employing a cooperative learning method, the data mining tools collaborate during the learning process. The cooperative learning process, as depicted in figure 2, consists of three steps (Viktor, 1998):

By cooperating during the learning process, rather than afterwards, the weaknesses of the individual tools are alleviated. Similarly, high quality results are shared and adopted as necessary. 

2.3.	Reporting

The final results (i.e. the knowledge discovered) should be comprehensible and accurate. The representation should adhere to the following usability principles, as defined by Dix et al (1993):

3. The need for systems methodologies and human involvement

The government institution as well as the KDD process form part of Systems Theory, which includes hard and soft systems approaches. It is important to note that the KDD process is hard systems oriented. The process is means-end oriented, which implies that the objectives are already clearly defined before the process is initiated. However, experience shows that the user objectives are not always clearly defined at the start of the KDD process, especially within a social environment such as an organization. In the case study, this lead to a number of problems, including the following:

3.1. The philosophy of soft systems methodologies

Soft systems methodologies present a broad approach to the analysis of problem situations in a manner that will lead to action decisions on both the levels of how and what - the latter is a system of investigation. If this system of investigation is followed, it ensures that nothing is presumed as obvious beforehand. According to soft systems thinking, problem situations originate because of contrasting perspectives on the same situation held by different persons. The SSM assists the participants in the process to determine their goals and reach consensus before these goals can be pursued. In the case study it was presumed that the goals and objectives were clearly defined and all major stakeholders had a similar view of the project. However, this was not the case. For example, the consultants responsible for the data capturing independently developed diverse questionnaires. Also, the database developers did not consult these questionnaires and proceeded to develop a data model independent of inputs from other stakeholders. This lack of communication, as well as the problems that originated from it, could have been avoided if the difference in perspectives were acknowledged.

The SSM consists of a seven-stage process of enquiry, as depicted in figure 3. Interested readers are referred to Checkland and Scholes (1990) or Flood et. al. (1991) for a detailed discussion of the SSM.

3.2. The principles of the SSM applied to the KDD process

At present, the importance of the human aspect in the context of information systems and related topics is widely acknowledged. However, the continuous and active participation of humans is usually not stressed. Continuous human participation involves much more than a mere input to the KDD process. Rather, it is a goal as such. The SSM provides an approach to ensure continuous human involvement by viewing it as an integral part of the methodology itself. During the case study, the inputs from the owner and other end users were not incorporated into the KDD process. The original databases were created by using the inputs as obtained from five different surveys (questionnaires). JAD sessions with stakeholders were held after the questionnaires have been drawn up. However, the inputs from the JAD sessions were not reflected in the questionnaires. When applying the principles of the SSM, the views from all major stakeholders are considered throughout the KDD process.

Problem situations originate because of contrasting perspectives on the same situation held by different persons. The main purpose of soft systems thinking is to obtain a common solution to the problem that is acceptable to all participants. This view helps soft system thinkers to focus on what must be done and it also facilitates a discussion of the problem situation. Coordination and cooperation are highly significant aspects that should be present when attempting to successfully complete KDD. These two aspects are implicitly part of the soft systems thinking. Therefore, the use of a soft systems methodology when conducting KDD can assist the persons responsible for the completion of the KDD process.

The SSM plays an important role in the regulating of communication and authority. By doing so, it ensures that the following issues are addressed: Communication between different participants in the KDD process is guaranteed, training takes place after the completion of the KDD process, there is continuous feedback to stakeholders, the development process is transparent and fair arbitration of stakeholder differences is ensured. These goals are accomplished by incorporating the inputs from the end users into the KDD framework. The inputs are used to determine the important tasks and focus the search. In the case study, there were no clear communication and power structures in place. Because of the lack of communication between the consultants responsible for the data capturing and the development team, consensus were not reached about the data capturing and data preprocessing phases. Therefore, each of the databases had a unique format and contained missing values, which lead to a considerably increase in the complexity of the data preprocessing phase.

Another characteristic of SSM is a cultural based enquiry stream that explicitly investigates the culture within which the work is being done. This cultural investigation is a continuous process and does not only take place at the outset of the SSM. Rather, soft systems methodologies offer a framework for the conscious consideration of issues that would not usually have been taken into account when conducting the KDD process. The models constructed give the stakeholders of the project the chance to understand the prevailing culture of the organizations participating in the project. It also leads to an understanding of the way in which other stakeholders that participate in the project function, especially on cultural level. The case study was completed without regarding the context within which the main stakeholder (owner) functions. Therefore the KDD process did not take the context of the organization into account: the two entities remained separate.

3.3. Incorporating the SSM into the cooperative learning phase

When incorporating the SSM into the KDD process, the human role-players are modelled as forming part of the system. Their inputs are used during data preprocessing to ensure that the relevant data are selected and that the data integration and coding steps take place without loss of information. The presentation and content of the final results should meet the approval of the role-players. In addition, the human role-players are explicitly modelled as part of the cooperative learning process. The cooperative learning phase is extended to include computational learners (data mining tools) as well as human learners. That is, the role-players actively participate in the cooperative learning process, thus aiding the learning process and possibly adapting their own perception of the system. This process is depicted in figure 4.

4. Conclusion

The knowledge management framework uses cooperative learning techniques to formalize the cooperation of tools and humans during the KDD process. The framework explicitly models the human experts as part of the KDD system. That is, the human experts form, together with the data mining tools, part of the cooperative learning process. By participating, the experts can evaluate the knowledge obtained by the knowledge discovery processes against their collective personal knowledge and experiences. In this way, the explicit knowledge that was obtained by means of the knowledge discovery processes is augmented with the tacit, implicit knowledge of the experts. The expertise, creativity and innovation of the experts are diffused throughout the organization.

It is our experience that, especially in a large project, a formal systems methodology such as the SSM should be used to structure and focus the problem areas before and during the KDD process. As seen from the case study, adequate communication between stakeholders is paramount for successful project completion. If this is not approached in a structured way, the coordination and cooperation between stakeholders do not take place. The culture of the organization should be analyzed to understand the way in which people within the organization perceive the knowledge to be discovered. The SSM provides a mechanism to accomplish the latter.

We are currently using the lessons learned during the first phase of the project to develop a systems approach that incorporates stakeholder participation. The second phase will also involve the development of a data warehouse from the original databases as provided by the data capturing teams.


Adriaans, P. and Zantinge, D. (1996). Data mining, Addison-Wesley, Harlow: England.
Clark, P. (1989). The CN2 induction algorithm, Machine Learning, 3, 261-283.
Checkland, P. and Scholes, J. (1990). Soft systems methodology in action. Wiley, New Jersey: USA.
De Villiers, C. (1996).  The integration of IT in a cooperative learning environment, D Comm dissertation, Department of Informatics, University of Pretoria, Pretoria: South Africa.
Dix, A. et al. (1993). Human-computer interaction, Prentice-Hall, New York: USA.

Flood, R.L. and Jackson, M.C. (1991). Creative problem solving: Total systems intervention. John Wiley & Sons, Surrey: Great Britain.

Jackson, M. C. (1991). Systems methodology for the management sciences. Plenum Press, New York: USA.

Holsheimer, M., Kersten, M. and Toivonen, H. (1995). A perspective on databases and data mining. Proceedings: First international conference on knowledge discovery and data mining, Montreal: Canada, 150-155.
Khosla, R. and Dillon, T. (1997). Engineering intelligent hybrid multi-agent systems, Kluwer Academic Publishers, Boston: USA.
McGraw, K.L. and Harbison-Briggs, K. (1989). Knowledge acquisition principles and guidelines, Prentice Hall, New Jersey: USA.
Mitchell. T.M. (1996). Machine learning, McGraw-Hill, Boston: USA.
Moustaskis, V., Lehto, M. and Salvendy, G. (1996). Survey of expert opinion: which machine learning method may be used for which task? International journal of human-computer interaction, 8 (3), 221-236.
Quinlan, J.R. (1994). C4.5: Programs for machine learning, Morgan Kaufman.
Sharan, S. (1994). Handbook of cooperative learning methods, Greenwood Press.
Slavin, R.E. (1995). Student team learning: an overview and practical guide (second edition), National educational association, Washington, D.C.: USA.
Sestitto, S. and Dillon, T.S. (1994). Automated knowledge acquisition, Prentice-Hall, Sydney: Australia.
Viktor, H.L and Cloete, I. (1998). Inductive learning with a computational network, Journal of Intelligent and Robotic Systems, Kluwer Academic Publishers, 21(2), 131-141.
Viktor, H.L., Cloete, I. and Beyers, N. (1997). Extraction of rules for tuberculosis using an artificial neural networks, Methods for Informatics in Medicine, 36 (2), 160-171.

Wnek, J. and Michalski, R.S. (1994). Comparing symbolic and subsymbolic learning: three studies, Machine Learning (editors Michalski, R.S and Tecuci, G.), 4, Morgan Kaufman, San Francisco: USA, 489-519.

Woodridge, M. and Jennings, NR. (1998). Formalizing the Cooperative Problem Solving Process, Reading in Agents, Morgan Kaufman, 430-442.



Professor Clive Holtham and Nigel Courtney;

City University Business School,

Frobisher Crescent, Barbican Centre,

London EC2Y 2HB, UK.

Tel 00 44 171 477 8622; Fax 00 44 171 477 8629;


This paper reports on collaboration since 1995 between a research university and a consortium of UK public and private sector organisations to raise the capabilities of their managers to extract business benefit from IT investment. The focus is on strategic awareness rather than practical IT skills. The process entails the creation and sharing of knowledge.

This has meant developing and refining theories concerned with knowledge management, strategic information systems and executive learning and then implementing practical tools and methods based on these theories. The results to date are reviewed and proposals for the next stage of the project are explored.


There continues to be a wide range of theories about factors that promote or inhibit the successful application in business of information technology (IT) and communications technologies. We have grouped these theories into three categories; knowledge management, strategic information systems and management learning.

1.1. Knowledge management

The global shift towards knowledge-based work has provoked a flood of books describing a scalar spectrum comprised of data, information, knowledge and wisdom (eg: Wiig et al 1997). But information to one person may be data to another. In this sense, knowledge can be defined as 'information with meaning and context'.

The possession of knowledge and insight provides power and influence to individuals - and to organisations via the knowledge-based theory of the firm (Nonaka & Takeuchi 1995). Grant observes that "knowledge-based theory also permits us to look beyond conventional transaction cost analysis to better understand the optimal boundaries of the firm" (Grant 1997). But knowledge is tacit and exists in people's heads; so knowledge management must be a personal matter. Collectively, it is only feasible to 'manage for knowledge'.

Although expert systems had failed to deliver an 'institutional memory' (Neale & Morris 1988) the 1990s recession brought the issue back into focus. The legions of discarded managers had taken the corporate knowledge with them. Their successors urgently sought practical ways to apply knowledge management. Three main approaches have emerged: knowledge value management, intellectual capital management and knowledge productivity.

The first category is exemplified by the way Dow Chemical has saved $40 million by rationalising its patents (Petrash 1996). The second by Skandia's belief that the difference between market capitalisation and net worth represents an organisation's intellectual capital which must be nurtured like any other asset (Edvinsson 1997). The third was pioneered by the French Atomic Energy Commission (CEA) to maintain its capability following the test ban treaty. The solution enables groups of experts to create a 'knowledge book' that codifies the actions they would take in each circumstance (Ermine & Coquand 1997).

In none of these examples is IT essential - but it certainly helps.

1.2. Theories for strategic information systems

Until the late 1980s almost all of the huge corporate investment in information systems was used to automate tasks. Very few organisations were exploiting the capabilities of IT in strategic ways (Rockart & De Long 1988) or integrating their IS with telecommunications (Keen 1988).

In the US the emphasis turned to creating organisation structures and behaviours that suited the information systems. In close succession corporate America embraced IT-enabled transformation (Scott-Morton 1990), business process re-engineering (Hammer & Champy 1993), distributed IS architectures (Donovan 1994) and organisational renewal (Gouillart & Kelly 1995). In the UK the quest turned to new managerial roles such as the 'hybrid manager' (Earl 1989) - defined as "a person with technical skills able to work in user areas doing a line or functional job, but adept at developing and supplementing IT application ideas". To offset the implied technical bias, Earl added two further roles: 'leaders' - business executives who can drive the exploitation of IT; and 'impresarios' - IS managers who can propel the organisation into strategic consideration of IT (Skyrme 1996).

In either region it was hard to find business leaders who excelled by using computers personally to communicate, coach, convince and compete (Boone 1991).

1.3. Theories of management learning

There is a long-standing tradition in management education of using case studies abstracted from real situations (McNair 1954). The 1960s saw a shift to 'less classroom, less hierarchical' forms (Revans, 1978) requiring a physical environment, not dissimilar to the atelier of an artist, for accelerating practical learning of both novices and more experienced practitioners (Argyris & Schon 1978). This led to the observation that "experiential learning is a process that links education, work and personal development" (Kolb 1984).

Psychologists have found that people naturally turn to metaphors to help them unlearn existing theories as they broaden their worldview (Winnicott 1958; Papert 1980). Pask saw that we are all born with a preferred cognitive style, be it holistic, versatile or serialist (Pask 1986) while personal circumstances will also determine when off-the-job or on-the-job learning options are appropriate (Easteal & Thomas 1984). The idea of the manager as educator, coach and counsellor (Tannehill 1970) has led to the concept of personal responsibility for on-going learning (Senge 1990). From this has emerged the belief that personal development must embody a willingness to share information and knowledge so that the entire organisation can become a 'learning company' (Pedler et al 1997).

1.4. Four hypotheses

From our review of the literature we derived four hypotheses:

  1. there is no single right way to achieve successful management learning. Organisations profit by combining a variety of practical approaches
  2. learning is not linear; it must be a dynamic, iterative process
  3. metaphors enable the experienced executive more rapidly to translate information into knowledge for evaluation
  4. managers will develop strategic IT skills more quickly when they can draw on experiences from other organisations which they perceive to be relevant to their own specific circumstances.


It needs to be emphasised here that the executive IT skills referred to in this paper do not normally include hands-on skills of any type, whether application package, e-mail, office software or co-ordination technologies. The concern is much more with those skills that enable executives to identify business benefits that are potentially available from the deployment of effective information systems as described in the 'IT skills in the 90s' report (West London TEC 1993).

2.1. A consortium of IT users

By January 1995 the stark messages in that report had resonated with executives in some twenty leading UK organisations who decided to create a collaborative consortium to find practical ways to improve the IT-awareness of executives. The opportunity to test our hypotheses was provided when this consortium commissioned research and organised a broad range of collaborative events designed for groups of managers in different industries but with similar levels of responsibility.

2.2. The Executive Learning Ladder

Reviewing our hypotheses on management learning, we developed an integrative framework supported by materials published both on paper and in electronic format drawing on portfolios of some 450 original and public domain case studies plus diagnostic booklets and workbooks illustrated by best practice case examples. We called this framework the 'executive learning ladder' and proposed that for each of the five levels there are common catalysts which can promote the enhancement of knowledge.

  1. At the lowest level the novice learner is largely passive - reading, listening and using computer-based training modules to internalise and start to make sense of generalised data and some information.
  2. The formal educational environment usually emphasises theories, models and frameworks. But many managers place then need practical examples and seek relevant real-world case studies of best practice relating to different sectors, size of company, specific technologies and business drivers.
  3. As management proficiency increases so does the need for 'relevance' of learning materials. This is conveyed by case examples with a larger amount of the meaning and context which enables information to be converted into knowledge. A more highly developed mental model of business requires fewer stimuli to trigger pattern recognition.
  4. As the higher levels of the learning ladder are reached the executive is attuned to metaphors which encapsulate large amounts of meaning and context and allow the knowledgeable executive to understand new situations or concepts very rapidly.
  5. At the highest level the executive is already an expert and develops further degrees of knowledge and insight through collaborative exchanges with other experts.

The circle of arrows in each layer symbolises the dynamics of gathering data, acquiring information, developing knowledge and gaining insight - the information spectrum transformed into a learning system. These flows are between individuals as well as within the brain of each individual. But what is knowledge to one person may merely be information to another. As the writer Eric Hoffer reminds us: "In times of drastic change it is the learners who inherit the future. The learned usually find themselves equipped to live in a world that no longer exists" (Brown 1997).

2.3. Implementation

Having developed this framework, content architecture and knowledge acceleration dynamics were analysed in parallel. A groupware system (Lotus Notes) was used both to store and retrieve cases and provide a search capability. Over the life of the project this architecture has been augmented with web-based access, asynchronous conferencing and audio- and video-tapes of exemplar cases. At the same time consortium delegates consistently demanded networking occasions - often supported by access to the case examples and augmented by electronic meeting room systems. This indicates that many UK executives still do not find online access to be a natural or convenient way to learn.

Over 12-months (1/2/97-31/1/1998) 50 face to face events were delivered. These variously addressed the five levels of the learning ladder and were attended by a total of 561 executives from a wide range of industry sectors. Attendance at any event is entirely optional. The distribution of bookings between senior, middle and novice managers was 51%, 28% and 21% respectively.

Figure 2 groups the 50 events into categories and shows that, left to their own devices, senior managers will seek out networking occasions, middle managers prefer interactive research workshops and novice managers are most likely to attend practical workshops.


Organisational impact has been evaluated by means of assessment forms and reviews with executive sponsors. Although some managers at each level have rejected this learning process, the overall satisfaction rating has remained above 80% despite relatively high degrees of organisational change, both internally and externally. We deduce that, in terms of Earl's definitions, companies need to cultivate the 'learning leaders' and 'learning impresarios' who can play a key role at the middle levels of the learning ladder. The impresarios would take ownership of the alignment of business and IT strategy; the leaders would drive through the enterprise-wide development of skills to use IT. In terms of our four hypotheses:

A variety of approaches is required. Since the consortium embraces many industries an eclectic range of stimuli is required. The evidence from this project is that the more experienced the executive, the fewer the number of stimuli required to trigger recognition and understanding. At the higher levels of the learning ladder the contribution of expertise is as important as the acquisition of fresh information.

The learning process must be dynamic. The 50 events led to the publication of 20 workbooks for consortium members. Most resulted from facilitated electronic meeting room workshops. The online resources have served to stimulate and validate outcomes but moderated asynchronous discussion groups have had little impact. Dynamic learning by executives remains dependent on face-to-face exchanges.

Metaphors speed up managerial learning. Metaphors have proved valuable with senior executives in two ways; to bring an ad hoc group very quickly to a shared understanding of the issues and to make outcomes memorable. More junior managers require concrete examples to stimulate learning.

Development of a knowledge-sharing culture. Our approach has been to 'show what excellent looks like' - firstly via online access to first hand case studies, then by videotaped interviews and finally by visits to exemplar organisations, hosted by the CEO. Perceived business benefit is evidenced by an annual consortium membership subscription renewal rate above 80%.


  1. Despite all the advances in IT-mediated communications, senior executives uphold face-to-face contact as the critical factor for successful development of managerial competencies in the application of IT for business benefit
  2. Although e-mail is prevalent in UK organisations, executives demonstrate by their actions that asynchronous electronic communication across organisational boundaries has very low priority
  3. With experienced executives the immediate impact of metaphors can and does speed up their unlearning of earlier theories and absorption of new
  4. The Executive Learning Ladder has proved to be a powerful concept but requires continuous content development accompanied by the effective deployment within organisations of the key learning roles of leader and impresario.


Argyris, C. & Schon, D.A. (1978) Organizational Learning: a Theory of Action Perspective (Reading MA; Addison-Wesley)

Boone, M.E. (1991) Leadership and the Computer (Rocklin, CA.; Prima Publishing)

Brown, P. (1997) Strategic Planning Society Annual Report: chairman's statement

Donovan, J.J. (1994) Business Re-engineering with Information Technology: sustaining your business advantage (New Jersey; PTR Prentice Hall)

Earl, M.J. (1989) Management Strategies for Information Technology (Prentice Hall)

Easteal, M. & Thomas, M. (1984) The Development of the General Manager: a review of best practice (Nuffield Provincial Hospitals Trust)

Edvinsson, L. (1997) Developing Intellectual Capital at Skandia; Long Range Planning, Vol 30, No 3, pp 366-373

Ermine & Coquand (1997) Corporate Memory and Knowledge Productivity; in the proceedings of the 10th World Productivity Congress; October 12-15; Santiago, Chile

Gouillart, F.J. & Kelly, J.N. (1995) Transforming the Organisation (McGraw-Hill Inc)

Grant, R.M. (1997) The Knowledge-based View of the Firm: implications for management practice; Long Range Planning, Vol 30, Iss 3, June, pp450-454

Hammer, M. & Champy, J. (1993) Re-engineering the Corporation: a manifesto for business revolution (London; Nicholas Brealey Publishing Ltd)

Keen, P.G.W. (1988) Competing in Time: using telecommunications for competitive advantage (Ballinger/Harper & Row)

Kolb, D.A. (1984) Experiential Learning: experience as the source of learning and development (Englewood Cliffs, NJ; Prentice-Hall); p4

McNair, M.P. (1954) The Case Method at the Harvard Business School (New York; McGraw-Hill)

Neale, I.M. & Morris, A. (1988) Knowledge Acquisition for Expert Systems: a brief review; in Expert Systems for Information Management; pp 178-182

Nonaka, I. & Takeuchi, H. (1995) The Knowledge Creating Company (Oxford University Press)

Papert, S. (1980) Mindstorms: children, computers and powerful ideas (Brighton; Harvester Press)

Pask, G. (1976) Styles and Strategies of learning. British Journal of Educational Psychology; 46, pp128-148

Pedler, M., Burgoyne, J. & Boydell, T. (1997) The Learning Company (London; McGraw-Hill)

Petrash, G. (1996) Dow's Journey to a Knowledge Value Management Culture; European Management Journal, Vol 14, No 4, pp 365-373

Revans, R.W. (1978) The ABC of Action Learning: a review of 25 years experience (Bromley; Chartwell-Bratt)

Scott-Morton, M. (ed.) (1991) The Corporation of the 1990s: information technology and organisational transformation (Oxford University Press)

Senge, P.M. (1990) The Fifth Discipline: the art and practice of the learning organisation (Doubleday Currency)

Skyrme, D.J. (1996) The Hybrid Manager; in Earl, M.J. (ed.) (1996) Information Management: the organisational dimension (Oxford University Press)

Tannehill, R.E. (1970) Motivation and Management Development (London; Butterworths)

West London TEC (1993) IT Skills in the 90s: overcoming obstacles to growth (London; West London Training and Enterprise Council)

Wiig, K. et al (1997) Leveraging Knowledge for Business Performance (edited and published by Wits Business School, South Africa)

Knowledge networks for service companies:

The art of supporting the “emergent” with Information and Knowledge Technology

Dr Walter Baets

Euro-Arab Management School

C/ Carcel Baja 3

18001 Granada


Phone: +34 58 80 50 50

Fax: + 34 58 80 01 52

E-mail: baetseams.fundea.es

In the cognitive sciences and even more so in the epistemology a lot of work has been undertaken in order to identify and define knowledge. Unfortunately, in managerial sciences, we do not know a lot about what managerial knowledge really is and though we all have a vague feeling for what this kind of knowledge is, there are not a lot of definitions within a managerial context.

Kim (1993) suggests that knowledge is a combination of "know-how" and "know-why". Other authors identify (two) types of knowledge (Firebaugh, 1989; Nonaka et al., 1994). On the one hand we have explicit knowledge. This kind of knowledge is transmittable in formal, systematic language. This kind of knowledge is dealt with in knowledge based systems. On the other hand we have implicit knowledge or tacit knowledge. This latter kind of knowledge deserves most of our attention. Originally, implicit knowledge was defined as that knowledge which is logically entailed in the system like e.g. the process of deduction applied (Firebaugh, 1989). Tacit knowledge has more of a personalised quality. It is deeply rooted in action, commitment and involvement in a specific context (Nonaka et al., 1994). Tacit knowledge involves cognitive and technical elements, but it also involves context.

Tacit knowledge gets used in managerial tasks and tacit knowledge is the knowledge which makes the difference. Key to acquiring tacit knowledge is experience. A well identified example of tacit knowledge in management is the decision making behaviour of dealers on financial markets. Based on what they learned from their past experience, the things they read and hear, probably the "climate on the market", they make decisions on buying and selling a in few seconds. We like to call this "instinct" or "fingerspitzengefuhl" but the behaviour of different dealers is different. Each individual dealer seems to have his own way of dealing based on his experience and his reference framework. It proves extremely difficult to extract this kind of "knowledge" from dealers and not because they do not want to share that. It seems extremely difficult for dealers to express the knowledge which they use, or in technical term to make tacit knowledge explicit. However, since some dealers are persistently better than others, it would be of interest for a company to understand why they score better, in order e.g. to reproduce the "winning" behaviour. Furthermore, if a dealer acquires his experience/knowledge during his stay in a particular bank, how can this bank keep this acquired knowledge (this asset), in case the particular dealer leaves the company ?

Many different types of cognitive elements are involved. Those of interest for managerial problems seem to centre on "mental models" (Johnson-Laird; 1983) in which human beings form working models of the world by creating and manipulating analogies in their mind. Senge (1990) describes mental models as deeply held internal images of how the world works which have a powerful influence on what we do because they also affect what we see. Mental models represent a person's view of the world, including explicit and implicit understanding. Mental models provide the context in which to view and interpret new material and they determine how stored information is relevant to a given situation (Kim, 1993). Based on these definitions and analogies to individual learning, organisational learning is defined as increasing an organisation's capacity to take effective action (Kim, 1993). What seems to matter is not reality but rather perceptions of reality. It is clear from this description how crucially important context is for learning and knowledge.

This capacity of an organisation to take effective action is based on tacit corporate knowledge (Baets, 1996). The more this corporate knowledge is accessible (which does not necessarily mean explicit) and shared, the easier it becomes to take advantage of it. Perceptions of reality become more important than reality for management. Hence the role of corporate mental models becomes extremely important since the ultimate aim of corporate mental models is to visualise the shared mental model on any chosen subject. In that case fundamental to corporate learning, and hence to proactive management, is a shared mental model (Kim, 1993) more than anything else. Further in this chapter we will detail more how mental models get created and how they appear. If we want to take this reasoning one step further, one could even consider that it is the manager's role to identify the shared elements within the diversity (complexity) (Van der Linden, 1993): this idea introduces management of corporate (tacit) knowledge as a strategic mission.

If we want to understand more about mental models, we need to have some idea how the human brain works. The PDP Research Group (Rummelhart and McClelland, Vol 1 and 2, 1986) demonstrates, based on extensive research, that the human brain is characterised by a high degree of parallelism. This means that a large number of elements (in this case neurones) are used at the same time alongside each other. A second important characteristic of the human brain is the Micro Structure of Cognition (distributed knowledge) of which it is built. The human brain has no clear equation for what happens in a given situation, but is able to reconstruct solutions, actions, etc. fast and easily, based on this micro structure of knowledge. Consequently, we can assume that knowledge is not sequential (but parallel) and deals with variety (and not with averages).

Despite the many problems with definitions, the organisational capability for knowledge creation is gaining momentum in managerial sciences and some consider it as a potential source of competitive advantage for companies (Toffler, 1990; Badaracco, 1991; Quinn, 1992; Baets, 1997). Whereas companies have long been dominated by a paradigm that conceptualises the organisation as a system that "processes" information and/or "solves" problems, we now consider an organisation as a knowledge creating system (Nonaka et al., 1994; Borucki and Byosiere, 1994).

This body of knowledge, this repository of core competencies, this capacity to manage knowledge and to learn from it has recently attracted some attention in management research. A number of concepts point to the body of differentiating ideas, skills, capacities, etc. of a company. Eden (1988) introduced the idea of a "cognitive map" as a sort of pool of knowledge in the company. A cognitive map is a map in which a person expresses via blocks and connecting arrows how he reasons about a particular subject or how he sees that things fit together. Haeckel and Nolan (1993) introduce the "corporate IQ" for a comparable "repository" of corporate knowledge. Keen (1993) argues for a more quantitative representation of this "repository" and calls it a "fusion map". They all aim to identify and make explicit the differentiating core competencies of a company. All of them seem to agree that there exists a "repository of knowledge", which is the resource for the differentiating capacity of a company.

From a knowledge based perspective, the process of doing something is probably more important than the ultimate goal. The process of moving somewhere and of learning about the situation is a process which initiates knowledge. If the ultimate goal becomes too important and central to a company, what does the company do when it reaches the goal (if possible) ? Should it stop operating ? Can one conceive that at a point in time a company chooses to change course ? The process of corporate survival, or call it sustainable development, is therefore more important than the ultimate goal which could be share-holders value or short-term profit.

Instead of over-emphasising procedures and rules, we should allow people to have fun in their job. Knowledge is created by people who share experiences. When people have fun in their job, they identify with the job and they put their soul in the job. They are willing to learn and to adapt since they do not feel threatened by rules and regulations. They can come up with an idea or a proposal. They will pay attention to detail since they will observe things they have not seen before and they will enjoy it. Sense for detail is a basic attitude for quality improvement. Furthermore, quality is also the ability to "listen" to the problem. Managers should take time in order to take distance. If one "listens" to the problem, on many occasions the problem itself suggests solutions. Quality ends up to be a healthy combination of men and machine.

So knowledge management ends up being a healthy combination of a learning culture and an Information and Knowledge Technology (IKT) network, in which the sufficient condition for organising knowledge networks is the adequate use of Information and Knowledge Technology (IKT). IKT is a combination of techniques and technologies which all relate to what is commonly known as IT, telecommunications and Artificial Intelligence, often brought together in a physical network.

Intelligent support tools such as Case Based Reasoning Software (CBRS), Knowledge Based Systems (KBS), Cognitive Mapping Systems (COMS), Group Decision Support Systems (GDSS), Artificial Neural Networks (ANN) together with a knowledge-base , case library and data base can be loaded on to a fileserver which is connected to other computers/ terminals through a local area network where all members have equal access to the different tools and are able to communicate freely. New technologies such as electronic brain storming, group consensus and negotiation software and general meeting support systems can be integrated in the GDSS for supporting group learning. Computer mediated communication systems and wide area networking would enable the companies to store, process, retrieve external information and provides an electronic learning environment where all members can communicate freely. Public domain databases can be incorporated. Technologies such as voice mail, E-mail and video conferencing can be made available for an efficient and effective communication. All these together can serve as environmental scanners. Companies' computers can be networked through Wide-Area-Networking with a relevant external system. Thus, remote access to business environmental knowledge can be made available within the company at any time. Venugopal and Baets (1995) discuss a conceptual framework for integrating these intelligent tools which could facilitate efficient and effective learning processes and support knowledge networks. However, the implementation of such a prototype integrated system will require a fundamental rethinking of the organisational design and transactions.

Examples are discussed in the service sector, e.g. banks, road construction and maintenance, Xerox Corp.


Badaracco J, 1991, The Knowledge Link: Competitive Advantage through Strategic Alliances, Boston, Harvard Business School Press

Baets W, 1996, “Some Empirical Evidence on IS Strategy Alignment in Banking”, Information & Management, Vol 30, Nr 4

Baets W, 1997, Knowledge creation and learning as a corporate strategy: Managing chaos for sustainable development, forthcoming

Borucki C and Byosiere P, 1994, "Toward a Theory of Enlightenment of Middle Management in Globally Competitive Firms", Annual Meetings of the Academy of Management, Las Vegas, Nevada, August

Eden C, 1988, "Cognitive mapping", European Journal of Operational Research, 36, pp 1-13

Firebaugh M, 1989, Artificial Intelligence: A knowledge-Based Approach, PWS-Kent, Boston

Haeckel S and Nolan R, 1993, "Managing by wire", Harvard Business Review, Sept-Oct, Vol 71, Nr 5

Johnson-Laird, 1983, Mental Models, Cambridge University Press, Cambridge

Kim D, 1993, “The link between Individual and Organizational Learning”, Sloan Management Review, Fall

Nonaka I, Byosiere P, Borucki C and Konno N, 1994, "Organizational Knowledge Creation Theory: A First Comprehensive Test", Annual Academy of Management Meetings, Dallas, August

Rummelhart D and McClelland J, 1986, Parallel Distributed Processing: Exploration in the Microstructure of cognition, Vol 1: Foundations, Cambridge, MA: MIT Press

Rummelhart D and McClelland J, 1986, Parallel Distributed Processing: Exploration in the Microstructure of cognition, Vol 2: Psychological and Biological Models, Cambridge, MA: MIT Press

Senge P M, 1990, The Fifth Discipline: The Art & Practice of the Learning Organization, Doubleday, NY

Quinn, James Brian., 1992, " Intelligent Enterprise", The Free Press, New York

Toffler A, 1990, Powershift: Knowledge, Wealth and Violence at the Edge of the 21st Century, Bantan Books, New York

Van der Linden G, 1993, Een Essay over en Empirisch Onderzoek naar Betekenissen van Loon en Beloning. Casus van een postmoderne visie op bedrijfsbeheer, Rijksuniversiteit Groningen

Venugopal V and Baets W, 1995, “An integrated intelligent support system for Organisational Learning - A conceptual framework”, The Learning Organization, Fall

Sven Carlsson


Lund University

Ole Romers vag 6

SE 223 63 Lund



Shared Knowledge Creation Spaces around Business Processes

Sven Carlsson


Lund University

Ole Romers vag 6

SE 223 63 Lund



The motivation for this study was to better understand shared knowledge creation (SKC) around the new product development process in the context of international corporations. The research design is a three country progressive field investigation (Finland, USA, and Sweden) with staged theory building and iterative hypothesis testing -- which are preceded by a field pilot study. The focus is on SKC around the new product development process in high technology companies. The results of the study help us to better understand the nature of SKC spaces around business processes and how to better design them.

The study addresses these four main research questions:

RQ1. How does SKC take place around the new product development process in international corporations? What are its various explicit and tacit modes? What characterizes best practices in SKC?

RQ2. Can knowledge catalysts accelerate or facilitate SKC between and within NPD teams? How? How does the effect of knowledge catalysts change depending on the strength of links between collaborating teams?

RQ3. Should information systems for improved SKC be designed primarily around knowledge catalysts? If yes, how ? If no, how should they be designed?

RQ4. What are the implications of RQ1, RQ2, and RQ3 for the effective use of information technologies for improving SKC?

Currently, we are in the middle of the project. We have collected and analyzed data from Finland. By June we should have been able to collect and analyze the data from Sweden and the US.


Sylvie BLANCO and Humbert LESCA

Ecole Supérieure des Affaires, C.E.R.A.G.

University Pierre Mendès France, Grenoble 2, France

sylvie.blanco@cea.fr ; lesca@esa.upmf-grenoble.fr


This article is an attempt to better understand the relationships between Business Intelligence and Knowledge Management. First, we identify and explain a major problem for organisations facing rapidly changing environments : the selection of Early Warning Signals (EWS). Then, as empirical data about this task are quite unaccessible, we have conceived a support that is an "artefact" enabling us to observe people in an actual selection situation. It is briefly presented alongwith one of its implementation. So, we present the relationships between Business Intelligence and Knowledge Maangement that have been uncovered through this field study. Finally, we formulate theoretical implications and future research avenues.


By introducing the expression "Strategic Management", Ansoff (1975) highlighted the necessity to focus on organisations' ability to anticipate threats and opportunities, in order to cope with the turbulence of the environment.

Actually, various field studies confirm that standpoint. Successful organisations are those who detect major events through "alert sensors" (Hedberg et al., 1976), who gather more information, with more diversity and more frequently (Daft et al., 1988), who acquire strategic information and interpret it in order to use it for action (Thomas et al., 1993).

But practically, this task seems to be quite problematic (Ansoff et al., 1979 ; Porter, 1980 ; Aaker, 1983 ; Vandenbosch et Huff, 1997). Even organisations who have implemented BI systems often do not manage to anticipate strategic suprises (Gilad, 1988 ; Lesca, 1994). In both cases (with and without BI systems), most of them seem to suffer from both information overload and lack of "strategic information" which leads us to question their information gathering strategies.


BI as an uncertainty reduction process

According to Galbraith (1973), "If the organization is faced with greater uncertainty, due to technological change ...increased competition... the amount of information processing is increased" which amounts to increased task uncertainty. So, a major consequence is the necessity to reduce that uncertainty by ensuring the "fit" between information-processing needs and information-processing capabilities.

One of the two approaches identified by Galbraith is to increase the information-processing capabilities until they fit the amount of information-processing needs. It consists in creating processes and mechanisms to acquire and exploit the information required by the considered task. This approach of uncertainty reduction is presented in figure 1.

Business Intelligence is typically an uncertainty reduction process which consists in increasing information-processing capabilities. For instance, March and Feldman (1981) noted that the intelligence an organization possesses about its environment depends on its ability to acquire, analyze and retrieve relevant information at the right place and at the right time. Lesca (1994) defines BI as : "The information process through which companies prospectively monitor their environment by gathering weak signals in order to create opportunities and to reduce their uncertainty". This approach to BI is presented in figure 1.

Figure 1. BI as an uncertainty reduction process : a conceptual framework

Implementing such systems supposes that the "fit" to be obtained has been clearly identified. Nevertheless, as most existing BI systems don't seem to be effective, it appears necessary to reconsider both information processing needs and capabilities that are inherent to the anticipation of unpredictable changes in the environment.

Information processing needs

Authors from strategic management who have dealt with environmental monitoring have highlighted major difficulties. We focus here on two major points often recognized to be crucial : the nature of the information we deal with and the nature of the information processes being used.

Information that is required to anticipate unpredictable changes can be assimilated to weak signals (Ansoff, 1975). They are also called early warning signals (Reinhardt, 1984). EWS can be defined as any actions by a competitor that provide direct or indirect indications about its intentions, its motivation, its objectives or its internal situation (Porter, 1980). However, EWS can also concern the technological, political, economical or social environments (Bright, 1970). Ansoff (1975) and March and Feldman (1981) show that dealing with this kind of information is difficult because of its nature : it is anticipatory, qualitative, ambiguous, fragmentary, of various formats and it may come from very diverse sources. Further explanations are given in table 1.

Table 1. Nature of weak signals

To be more concrete, let consider the following piece of information : "the research director of our main competitor X has jus left his organisation". Various interpretations are plausible : a passed event not anticipative ; an EWS of a new technological orientation with which the director doesn't agree ; an EWS of the creation of a firm where this director will implement a promising technology, etc.. It is an ambiguous piece of information which bears various interpretations. Some of them may transform it into an EWS.

These interpretations take part of the processing needs, related to EWS gathering. First, as it is part of a strategic decision process, it is ill-structured, novel and complex (Mintzberg et al., 1976). Hence, reasoning is heuristic rather than algorithmic and subject to individual biases. Second, it refers to sensemaking rather than problem-solving (Weick, 1995).That means that the environment is not a given reality but rather a collective construction created through a process of attention and interpretation (Weick, 1995). It entails that information gathering cannot rely on exhaustive information requirements. It is not directed and conditioned by a given problem but more explorative, guided by mental models (Jonhson-Laird, 1983) and experience (Cyert and March, 1963). Finally, the information process can be divided into three main stages : perception of a stimulus ; interpretation in order to create sense ; learning or incorporation of new information into existing representations (Billings et al., 1980 ; Kiesler and Sproull, 1982 ; Daft and Weick, 1984 ; Cowan, 1986). A critical point in this process is the selection of "relevant" information in situations of time pressure and information overload (O'Reilly, 1980). The following quotations support that standpoint : "Detection of weak signals requires sensitivity, as well as expertise, on the part of the observers", (Ansoff, 1975) ; "In other words, managers must be able to scan environments selectively so that timely decisions can be made" (Hambrick, 1982) ; "Managers literally must wade into the ocean of events that surrounds the organization and actively try to make sense of them". (Daft et Weick, 1984) ; "Somehow, the tidal wave of environmental data must be funneled down to a small pipeline of information" (Smircich et Stubbart, 1985). So we can state that one of the major information processing need in the context of environmental monitoring lies in the selection of weak signals. It must be considered as a collective information process which consists in transforming raw data into EWS. Hence, environmental scanners' role encompasses an interpretive dimension. They are not only seen as raw information transmitters.

Organisations' information processing capabilities

To face these information-processing needs, organisations possess information-processing capabilites that can be assimilited to their information systems, to the extent that "information systems" is understood in its broader sense. More precisely, concerning the gathering of BI information, information-processing capabilities may include three dimensions which are individuals, organisational structure and information technologies.

Individuals who are assigned to an information gathering task are often called environmental scanners or gatekeepers. They are persons who access potentially interesting information sources (Thiétart and Vivas,1981 ; Aaker, 1983) or who hold specific knowledge about environmental actors (Choudhury and Sampler, 1997). The organisations' structure acts both as means and obstacles to support information systems. In the context of scanning the environment, the organisation can act as an information filter (Wang et al., 1991), both positively and negatively. For instance, the following factors have been presented as critical : the structure of the scanning network (centralized versus decentralized) and the assignment of responsibilities ; the proactive versus reactive culture ; the management style ; the reward systems ; the existing information flows and their quality ; etc. Finally, information technologies constitute the last dimension of information processing capabilities an organisation dispose. It includes electronic, oral and written supports and in the context of information gathering, the relevant information technologies are those related to information accessibility, transmission, receipt, exchange and "transformation".

BI and the selection of weak signals : some limits

Before the implementation of specific information systems for business intelligence, some authors among whom Hedberg (1976, 1978) noted that organisations can not rely on "traditional" information systems to cope with environmental uncertainty. They had better consider them as potential obstacles since they tend to encourage inertia and leave anticipation issues to managers' discretion, which has been revealed as not satisfactory. Actually, authors who have paid attention to individuals' information gathering strategies under time pressure and information overload are generally pessimistic about their abilities to anticipate strategic surprises (Kiesler and Sproull, 1982 ; Schwenk, 1984 ; Barr and al., 1992 ; Weick, 1995). For instance, Weick (1995) explains that uncertainty and complexity affect what people notice and ignore : they may simplify the cues that are extracted, they may implement queuing, filtering, abstracting, escape and chunking and finally alter the construction of the environment.

These individual weaknesses have motivated the creation of BI systems. The remaining questions are to assess whether they have overcome the preceding problems or created new ones and whether mechanisms to select EWS are proposed and satisfactory. A few authors formulate some general recommendations to improve the fit between information processing needs and information processing capabilities in the context of strategic intelligence systems. They can be grouped into two categories :

1. A general heuristic to improve individuals' information-processing capabilities (Mason and Mitroff, 1981 ; Cats-Baril and Huber, 1987) by providing them with knoweldge about BI.

2. A collective learning process to implement the heuristic (Hedberg et Jönsson, 1978 ; March and Feldman, 1981 ; Boland, 1994 ; Brannback, 1994 ) which implies to develop mechanisms of dialog and coordination such as direct supervision, standardization and mutual adjustment through informal communication within the environmental scanners' network (Gibbons and Prescott, 1996).

There is quite an important gap between these recommendations and existing BI systems regarding information gathering (Brannback, 1994 ; Lesca, 1994 ; Bartoli et Le Moigne, 1996 ; Vandenbosch and Huff, 1997). First, current research about BI ignore the problem of selecting EWS and focus on defining exhaustively and precisely information requirements. This is contrary to the nature of weak signals and often leads organisations to obtain "state-of-the-art" results (which tends to eliminate dissonant information) instead of proactive construction(s) of their environment. Actually, current research focus on information technologies, such as the Web, to support the gathering of BI information. It consists in increasing information-processing capabilities by increasing external raw data accessibility. This amounts to increase the information overload and do not enable organisations to reduce their uncertainty. As a response, some authors are currently developing methods to filter information but they keep on relying on the same postivist approach of information. They postulate that EWS can be found objectively into the environment thus ignoring the necessary interpretation process that permits to transform raw data into EWS.

To conclude, we can sum up BI limits regarding the information gathering mechanisms that are proposed as follows : the interpretation of raw data to obtain EWS, which is an heuristic process relying largely on individuals' knowledge is not taken into account ; the collective dimension of the process on which lies the confrontation of individuals' points of view and which requires dialog and communication mechanisms has not been much evoked ; the collective learning process that is necessary to develop appropriate and effective organisational attention is ignored.

Research Question

Hence, it appears necessary to develop specific mechanisms, in the context of BI, to improve organisations' ability to select collectively EWS. We can formulate the following research question : What mechanisms to conceive in order to support the collective selection of weak signals in a business intelligence process ?

This question can be divided into two subquestions which are : what heuristic to propose ? what collective learning process to support the implementation of the heuristic ?


A general heuristic to select EWS

We present this method briefly as it is not really the purpose of this article.This heuristic is articulated around two constructs : the cognitive processes (which are heuristic) by which individuals select information ; the criteria and the categories to be used to prepare and to guide individuals' perception of EWS.

The heuristic must be : easy to implement as environmental scanners cannot afford and are not willing to spend much time on EWS selection ; quite general to fit various contexts and be refined in accordance with contextual factors. Nevertheless, it must also be discriminative enough to help people be selective.

The phases that we have retained are : perception of a potential EWS ; argumentation to transform the perceived piece of information into an EWS ; information evaluation in order to make it subsequently usable by other people and to incorporate it into existing representations of the environment.

After having identified dozens of information quality criteria, we have retained six of them which seem to be the most appropriate. Then, these remaining criteria have been defined and apportioned to the preceding phases. Hence, the perception of a potential EWS is guided by two criteria : information relevance defined as its objective link to the BI scope (which supposes the existence of an explicit scope where actors are namely identified alongwith issues to be scrutinized and keywords) ; anticipative nature of information which is defined as its future-oriented nature and its ability to inform about future changes. The argumentation phase also relies on two indicative criteria which are : the significance of information which supposes that the individual formulates the future event he foresees ; the importance of this future event in terms of its potential impact on the organisation. Finally, the evaluation phase relies on two criteria which are information reliability depending on the information source and information timeliness which is the delay between information creation and information collection.

A collective learning process

The objective of implementing the heuristic through a collective learning process is to make individuals aware of their cognitive biases and of the collective dimension of the gathering process. However, before entering the collective learning, a training session to the constructs and concepts we propose is necessary (Vandenbosch and Huff, 1997). This 30 minutes preliminary session is articulated on three ideas which are the concept of BI, the environmental scanners' role and the heuristic and criteria that are being proposed.

Then, the collective learning phase begins. It includes two moments. First, people apply the heuristic on a case study which involves a BI scope, raw data and the users' guide accompanying the heuristic. Then, a collective work is initiated by a table lap where each persons says whether she has selected the information and argues his results. No commentaries from other persons are admitted. Then, a collective discussion takes place to argue individual differences. Finally, a collective "decision" about the selection of information is made and the experience about the heuristic is capitalized to be reusable. The next section presents our research methodology and a field study where we have implemented this heuristic. It illustrates the kind of knowledge we have produced.

Observing the selection of EWS : an action research

The proposed support (involving the heuristic and the collective learning process) has been implemented within various organisations willing to implement effective BI information gathering. These implementations have been done using an action research methodology. It has been made possible thanks to our support. These experiences allow us a preliminary validation of our support and the deepening of scientific knowledge about BI and knowledge management. In the next section, we illustrate one of our experiences. Voluntarily, we emphasize observations related to the place of knowledge within BI.

Context of the experience : the organisation is a local unit of a big company from the telecommunications industry. The partnership with us has been initiated by the top executives who were willing to improve their BI process. They felt they suffer from both information overload and lack of strategic information. The group of participants is constituted of fifteen persons from various services and different hierarchical levels.

The training session lasted approximatively half an hour. Three points bring out from this phase of knowledge acquisition about BI. First, people feel all the more motivated by their environmental scanners' role as they understand both the structure of the information process and the role they have been assigned. Second, the concept of EWS is largely accepted and seems to match a kind of information they were not able to formalize before. The opportunity people had to give examples of EWS they had in mind seemed to relieve them of a burden. Nevertheless, the limits of the concept have rapidly been reached. The intrinsic ambiguity of EWS motivated both dilemma and suggestions. For instance, participants did not agree to say whether the following piece of information "our main competitor has just opened a distribution center downtown" is an EWS. Another participant suggested to create an open list of EWS to support imagination. Finally, these difficulties introduced the need for a method. The presentation of our support has been well accepted. Even people who dread that this training would be complicated felt positively suprised by both its usability and its utility.

The next phase consists in observing the participants selecting information by applying the support. Individuals are provided with a BI scope and raw data and achieve their own selections. This lasts about twenty minutes. Only two persons did not carry out the exercise. One felt she had not enough expertise with regard to the other participants and concerning the domain under study. The other told us that this work was too unprecise and ambiguous.

Then we started the table lap. Each argumentation is written on the board in a way that highlights differences. Results are quite suprising as people's points of view are often very different.

At the end of these individual presentation, the collective discussion is very rapidly engaged by persons willing to compare their argumentation to others. Once engaged, we organized this phase in accordance with the selection process. We present here the main commentaries as they have been stated :

Commentaries about information relevance :

"According to me, this piece of information is not related to the BI scope. But I think the reason is that I did not understand this issue as you did."

"There is no need to be an expert to identify a link between a piece of information and the BI scope. It is just key-word matching."

"Had I not known that this actor is a subsidiary of one that pertains our scope, and that he is very active abroad, I would not have selected that information."

Differences in the interpretation of the BI scope led us to define it again. Then, reaching a consensus about information relevance was quite easy.

Commentaries about anticipation and significance : One of the dialog between two participants is quite illustrative of the way this criterion has been treated.

D :"I don't agree with you. This is not anticipative but rather a passed event !"

V :"Of course it is anticipative. Don't you think that this action indicates a future offensive towards our customers. The main point is that they are gathering information about our own customers and their satisfaction about our products."

D : "I had not seen that action under this commercial point of view. But now, I do agree with you and I would select this piece of informatio nto the extent that your argumentation is added to the raw data."

Commentaries about the importance criterion : This criterion has raised animated discussions. Three representative commentaries are given :

"I would have enjoyed to share the point of view of persons from the marketing department about these future products."

"This criteria is very important for us. It allows us to reveal and to share our individual richness. "

"I don't know much about our customers information systems. I wonder whether we are able to do as this competitor. I thought we were able and you do not seem to agree. So ?

Commentaries about information reliability and timeliness : A few remarks have been formulated. People were not really preoccupied by these criteria and the differences they have uncover. Among the most "sophisticated" commentaries, we have noted the following ones :

"This piece of information, whatever the source it comes from, is not very reliable. The reason is that the picture do not match what is described in the text".

"Timeliness is important but it is rather difficult to assess."

"I knew that informatioin already. So, I don't think its timeliness is high."

"How to know how long a piece of information has been kept by a journalist without being published".

After reviewing the six criteria, the discussion is recapitulated on the board in order to highlight the way the heuristic has been actually used and the final decision for each piece of information. We have carefully noted the remaining differences between individuals. At the end of this collective learning, participants have given their point of view about the whole session.


As action research has both practical and theoretical objectives, we articulate the presentation of our results around these two dimensions. A special emphasis is given knowledge.

Practical results : knowledge management issues and BI

The proposed support has been largely accepted and perceived as both usable and useful. The training session has allowed participants to understand their information processing difficulties and to improve their knowledge about BI. They have recognized that it has increase their motivation to track down information, and that they felt the task as potentially less difficult.

The collective learning process has been perceived as useful for various reasons. First, it is a way to be more effective as procedural knowledge (how to select information) has been acquired. Moreover, individuals have become aware of their own biases, due to their own knowledge, which tends to favor a collective behavior. People need to share information and knowledge to enhance mutual enrichment instead of keeping a partial representation of their environmnent. This awareness is a positive factor to motivate and to enable individuals to explicit their implicit knowledge. Finally, the procedural knowledge they have acquired is intended to evolve as suggested by the participants themselves.

A major point in the utility of the heuristic lies in the notions of anticipation and significance. These criteria bring added value to the process as they imply that knowledge is integrated into the selection process. Enrichment is quite easy as even when people initially diverge about one piece of information, a consensus is rapidly established. Nevertheless, the lack of familiarity in using such criteria may act as an obstacle, some persons thinking that it is too selective. So, more practical knowledge is required here.

Theoretical resutls : relationships between knowledge and EWS selection

First, as Vandenbosch and Huff (1997) suggested, it appears necessary to train people to the concepts and constructs they will have to implement. The development of preliminary knowledge is a necessary condition for the environmental scanners' network to be motivated and effective, that is to say to fit the information processing needs.

Second, prior knowledge in the form of expertise, privileged relationships with the environment, but also organisational position, affects the use of the proposed criteria. Individuals do feel guided and more effective in their information selection thanks to the heuristic. Nevertheless, as Kiesler and Sproull (1982) suggested, cognitive biases also apply to the use of criteria. For instance, information relevance depends on the knowledge individuals have about actors and issues of the BI scope. We have also observed that some people judge information as not relevant just because it does not enter their specific domain of expertise as mentioned by Choudhury and Sampler (1997) or because of their organisational position as mentioned by Cowan (1986) and Walsh (1988). Hence, prior knowledge may act as an important obstacles to EWS selection.

Concerning the anticipation and significance of information, the willingness to use these criteria depends on individuals' tolerance for ambiguity (Dermer, 1973). The way people use them is intrinsically linked to their knowledge and leads to different points of view about a same piece of information. This shows that EWS is necessarily constructed through interpretation and that this process needs to be guided and supported by appropriate information support.

The collective learning process we have proposed has also allowed people to become aware of their own biases due to their own knowledge. This has been possible by rendering individual tacit knowledge both collective and explicit. A major consequence is that individual become aware that their knowledge about the environment but also about information processing may evolve thanks to mutual enrichment.


To conclude, we can state that EWS can not be approached objectively but rather as a construct that implies individuals' knowledge. Hence, the selection process should necessarily be contemplated as a collective process where interpretation plays a major role. This leads us to formulate both practical and theoretical implications.

First, improving scanners' knowledge about BI, their role and the "procedures" they should implement increases their consciousness about BI and their motivation. So, developing scanners knowledge about BI seems to be an effective way to overcome the barriers identified by Boettcher and Welge (1994). In accordance with this standpoint, we have conceived a didactic software that is currently being tested.

Another interesting phenomenon lies in the necessity to manage prior knowledge in order to improve the selection process. This knowledge management has different levels. First, a kind of knowledge cartography inside the organisation should be established to ensure that the BI scope is covered by the selected environmental scanners. Some redundance may be enhanced to avoid prior knowledge biases in the selection process. Second, interpreting raw data to obtain EWS is perceived as adding value thus motivating. Moreover, participants feel that it is an opportunity to make the organisation benefit from their individual knowledge. So, specific supports and mechanisms should be implemented in order to make this implicit knowledge become explicit and collective. Finally, as people become aware of their own biases in part due to their specific knowledge, they feel that feedback loops should be possible to favor mutual enrichment. Moreover, they wish to organize periodic collective learning sessions in order to refine the heuristic and to enhance mutual adjustments. Such mechanisms should also be supported by the BI process.

Practically, the organisation we have worked with has listed each scanners' addresses, phone and fax numbers, designed a special "journal" to animate the network and organized periodic meetings. This amounts to develop a relational system between environmental scanners to support coordination and mutual enrichment. Furthermore, redundancy between scanners in terms of knowledge as well as network evolutivity should be allowed thanks to specific mechanisms.

Of course we are aware of the limits of this research in terms of generalization of our results. Nevertheless, we have replicated this experience within various organisations and we are about to explore new research avenues. Some of them concern validity conditions of the implementation of our support. For instance, we plan to replicate the experience presented here in a research community and a multinational group.


Aaker, D.A. (1983). Organizing a strategic information system. California Management Review 25(2) 73-83.

Ansoff I. (1975). Managing strategic surprise by response to weak signals. California Management Review, 18(2) 21-33.

Ansoff, H.I, Eppink, J., Gomer, H. (1979). Management of strategic surprise and discontinuity : problem of managerial decisiveness. Economies et Sociétés, G.1 459-488.

Barr P.S., Stimpert J.L., Huff A.S.(1992). Cognitive change, strategic action, and organizational renewal. Strategic Management Journal 13 15-36.

Bartoli J.A et Le Moigne J.L. (1996). Organisation intelligente et système d'information stratégique. Paris, Ed. Economica.

Belkin N.J., Croft W.B. (1992). Information filtering and information retrieval : two sides of the same coin ? Communication of the ACM 35(12) 29-38.

Cyert, R.M. and March, J.G. (1963). A behavioral theroy of the firm. Englewood Cliffs : NJ.

Billings R.S., Milburn T.W., Schaalman M.L. (1980). A model of crisis perception : a theoretical and empirical analysis. Administrative Science Quarterly 25 300-316.

Boettcher R., Welge M.K. (1994). Strategic information diagnosis in the global organization. Marketing International Journal 34(1) 7-24.

Boland R.J., Tenkasi R.V., Te'eni D. (1994). Designing information technology to support distributed cognition. Organization Science 5(3) 456-475.

Brannback M. (1994). Decision support systems for strategic management. Journal of Decision Systems 3(2) 95-114.

Bright, J.R. (1970). Evaluating signals of technological change. Harvard Business Review January 62-70.

Cats-Baril W., Huber G. (1987). Decision support systems for ill-structured problems : an empirical study. Decision Sciences 18 350-373.

Choudhury V., Sampler J.L. (1997). Information specificity and environmental scanning : an economic perspective. Management Information Systems Quarterly 21(1) 25-53.

Cowan D.A. (1986). Developing a process model of problem recognition. Academy of Management Review 11(4) 63-776.

Daft R.L., Sormunen J., Parks D. (1988). Chief executives scanning environmental charracteristics and company performance : an empirical study. Strategic Management Journal 9 123-139.

Daft R.L., Weick K.E. (1984). Toward a model of organizations as interpretation systems. Academy of Management Review 9(2) 284-295.

Dermer J.D. (1973). Cognitive characteristics and the perceived importance of information. The Accounting Review 48 511-519.

Galbraith J. (1973). Designing complex organizations. Addison-Wesley.

Gibbons P.T., Prescott J.E. (1996). Parallel competitive intelligence processes in organisations. International Journal of Technology Management, Special Issue on Informal Information Flow 11(1/2) 162-178.

Gilad B and Gilad T. (1988). The business intelligence system : a new tool for competitive advantage. Amacom 1988.

Hambrick D.C. (1982). Environmental scanning and organizational strategy. Strategic Management Journal 3 159-174.

Hedberg B.L.T., Nystrom P.C., Starbuck W.H. (1976). Camping on seesaws : prescriptions for a self-designing organizations. Administrative Science Quarterly 21 march 41-65.

Hedberg B., Jönsson S. (1978). Designing semi-confusing IS for organizations in changing environments. Accounting , Organizations and Society.

Johnson-Laird P.N. (1983). Mental models. Harvard University Press : Cambridge.

Keen P.G.W. and Scott-Morton M.S. (1978). Decision support systems : an organizational perspective. Addison Wesley Ed.

Kiesler S., Sproull L. (1982). Managerial response to changing environments perception on problem-sensing from social cognition. Administrative Science Quarterly 27 548-570.

Lesca H. (1994). Veille stratégique pour le management stratégique de l'entreprise. Economies et Sociétés, Série Sciences de Gestion, SG20(5) 31-50.

March G.J., Feldman M. (1981). Information in organizations as signal and symbol. Administrative Science Quarterly 26 171-186.

Mason R., Mitroff I.I. (1981). Challenging strategic planning assumption. Wiley :NY.

Mintzberg H., Raisinghani S et Théorêt A. (1976). The structure of unstructured decision processes. Administrative Science Quarterly 21(2) 246-275.

O'Reilly C.A. (1980) .Individuals and information overload in organizations : is more necessarily better ?. Academy of Management Journal 23(4) 684-696.

Porter M.E. (1980). Competitive Strategy. New York, Free Press.

Reinhardt, W.A. (1984). An early warning system for strategic planning. Long Range Planning, 17(5) 25-34.

Schwenk C.R. (1984). Cognitive simplification processes in strategic decision making. Strategic Management Journal 5 111-128.

Simon H. (1977). The new science of management decision. Prentice Hall, Revised edition.

Smircich L., Stubbart C. (1985). Strategic management in an enacted world. Academy of Management Review 10(4) 724-736.

Thiétart R.A., Vivas R. (1981). Strategic intelligence activity : the management of the sales force as a source of strategic information. Strategic Management Journal. 2.

Thomas, J.B., Clark, S.M., Gioia, D.A. (1993). Strategic sensemaking and organizational performance : linkages among scanning, interpretation, action and outcomes. Academy of Management Journal 36(2) 239-270.

Vandenbosch B., Huff S.L. (1997). Searching and scanning : how executives obtain information from executives information systems. Management Information Systems Quarterly 21(1) 81-107.

Weick K.E. (1995). Sensemaking in organizations. Sage Publications.

Walsh J.P. (1988). Selectivity and selective perception : an investigation of managers' belief structures and information processing. Academy of Management Journal 31 873-896.

Wang P. et al. (1991). Filtering strategic environmental information processing using EIS. Proceedings of the 27th IEEE Conference on System Science 3 126-134.

Figure 1. A conceptual framework to study uncertainty reduction problems

Table 1. Characteristics of weak signals
related to future potential events that may affect the organization's survival. So, a piece of information has no intrinsic relevance as it is not related to immediate decisions
it does not consist in numbers, nor in extrapolations. As it concerns events that have not occured yet, factual or quantitative data are not available.
they can be interpreted in various ways, which make them difficult to catch and to exploit
present in the form of fragments, each of which are unsignificant and suspect.
in various formats and sources
they can be picked up in any shape or form, such as snatches of conversation, electronic data, messages from conferences, customers, suppliers, etc.

Distributed Knowledge: the sense of communicational knowledge in distributed systems

Alain Cardon, Franck Lesage

Alain Cardon


Case 69, 4 place Jussieu

75252 Paris Cedex 05


Franck Lesage


INSA de Rouen BP 08

76131 Mont-Saint-Aignan



The Distributed Information Systems raise the problems of meaning in simultaneously exchanged knowledge. We try to solve this problem by extending each message with the aim to express the knowledge and pragmatics according to the sender's intentionality. We make a communicational layer able to interpret all these extended messages, using an artificial interpretation process generating the meaning of all the exchanged messages. Then, the system can distribute the result of these interpretation processes, allowing recipients to recognize the meaning and intentionality of senders.


Distributed Information Systems, Intelligent Information Systems, Multiagent systems.


Distributed Information Systems (DIS) are computerized tools used for management of complex situations: they allow communication between many distant decision makers in order to achieve a cooperative management of emergency situations such as industrial crisis, financial flow and strategic management. They should allow the structuration and the representation of the decisional actors' knowledge on the common problem to be solved, by the exchange of information knowledge through high bandwidth networks (WF86).

We are interested in the exchange of knowledge between users that are solving a common problem. The DIS architecture we propose should allow the expression of two knowledge categories. A classical architecture will allow the access to and the exchange of factual - objective - data on the situation through the handling of various databases. The other, containing the previous, must allow the expression of the perception of the phenomenon by the decision makers (later referred as actors), that is, their subjective appreciation of the situation. The system must allow multiple simultaneous accesses facilitating the decision making at various responsibility levels, that is, including the actors' intentions and that of groups of actors. It must then know the different types of exchanged knowledge and behave like a decisional entity able to unravel, from exchanged knowledge between actors, the global representation that can be reasonably conceived on the emergency situatio!!n.

Thus, we are in the context of the so-called complex DIS (CAR97), i.e. management of situations whose evolution is not well planned beforehand and whose component representating the current state is to be constructed, augmented and dynamically modified during the use of the system. In order to represent the link between the phenomenon to be studied, structured, managed by actors and the system which informs and provides help to the decision makers on the phenomenon and the knowledge on what the different actors have, we insert in the DIS an interpretation system on the knowledge and judgments on the situation as perceived by the actors. This is the system we present in this article.

Distributed Information Systems problematics

A DIS's purpose is to allow the effective management of various emergency situations. It is a complex system in the way that it is composed of numerous interdependant subsystems with different characteristics (CAR97). Some belong to domains whose rational modelization operates very well, like in the planification of intervention ressources or the exchange of secure informations, the others belong to the Artificial Intelligence domain, for instance, the acknowledgment and insertion of human values in intervention decisions.

In the classical approach, DIS's only purpose is: putting into adequation the currently appreciated situation with a distinguished one from a vast set of given situations. Thus, the intelligence part is particularly developped. Information sources mainly rely on objective informations coming from image captors or other reliable informators. These pieces of information are stored in databases and are then transfered into knowledge bases. This knowledge is synthetized and represented in objective entities that are to be used in strategic and tactic decision making, thus assuring the cohesion of actions of the different units on the ground. The interconnection and interoperability levels between the various operational entities are well taken into account. The problem of the effective cooperation between the operators now constitutes a key research topic but raises the problem of the acknowledgment of the actors' effective motivation and intention.

Here, we explore a new research area that has been pointed out by professionals: the mutual correct understanding of the exchanged information.

Indeed, when only one operator uses a system, this system is merely a factual knowledge database. The symbolic forms used in the system only exist for the user, this is typically the case of the de Saussure or Frege's diadic semiotics where knowledge is introduced in the system by experts. This knowledge is symbolically treated and eventually returned to a human who will use it. The symbols refer to what they signify.

Difficulties arise when there are two distant users of a same system and when they must communicate decision making knowledge. Indeed, information passed from one to the other may not have the same meaning for both. With time, the users may achieve a common understanding ground but it is not satisfying since there is no time for that in emergency situations.

The general functions of a DIS seen as a computerized system for the management of evolving knowledge, are based on the two following criteria:

1  There is an explicit knowledge for the operational treatment of typical emergency situations, made of numerous pre-established plans. For instance, in France, there are "Plans Particuliers d'Intervention" (Particular Intervention Plans) set in the different regions and allowing to lead coordinated actions of civil protection services in a well defined canvas. With this knowledge, an emergency situation management system has a certain degree of efficiency and unfortunately of inefficiency. In other words, it is only operative in domains already studied. The construction of a DIS of emergency situation management based on the exchange of annotated and augmented informations is a significative evolution of the planification and involves a mutation of the organisations.

2  The knowledge that is used and developped in a emergency situation management system has a structure which must be much like the one of the organisational complexity of the ressources engaged on the ground and augmented with the complexity of the representation of the situation of the institutinal decision makers. The system has an important metalevel of manipulation of this knowledge.

In the field of civil DIS, the traditional characters, based on data with a tendency to exhaustive planification, are considered, by professionals, as insufficient. The system must not only consider the current situation like in problem solving, but must also operate a decision generation process of analysis on the situation with a certain intentionality, and not limitating itself to strictly rational decisional pattern in four steps, considered as linear: project conception, deliberation, decision, satisfaction (SFE92). It must take into account the behaviors, intentions, engagements and multiple points of view of the actors and groups of actors in a situation being structured. The augmentation of the quantity and of the qualification of exchanged information will lead the system towards a necessary adaptivity to this very dynamical flux of manipulated knowledge.

The problem of the interpration of exchanged knowledge

In the emergency situation management by various institutional services, the initial situation is cleary not structured. The incoming data in the DIS, considered as fuzzy and inacurate knowledge coming from potentialy unreliable sources, will lead to the structuration of the problem domain. But since the actors that provide data belong to different institutions, are geographically distant, operate in emergency and have only a partial view of the situation, the exchanged knowledge is strongly dependant on the emission context, that is, it is encapsulated in some pragmactics it is necessary to consider (ECO93).

In the case of an emergency situation management system, there are numerous institutional actors who communicate simultaneously, at a potentially high rate and who state some facts considered as complex since badly perceived, badly recognized and full of question. All statement of a message is thus encapsulated with the operator's intentionality.

Two problems arise when dealing with the communicated information:

1  how the final recipient can, from the received symbolic form, trace back the real object that has been perceived and designated by the author ?

2  how can a computerized system take into account the achievement of this lecture, when it takes place in a heavily charged information flux ?

As a beginning, it seems necessary to introduce pragmatic knowledge in the system in charge of the communications. An actor generates a message that is a certain statement relatively to the situation, but loaded with intentions, judgments and subjective opinions. It is then necessary, in a first step, to take into account the pragmatics' characteristics of the message generation, in other words, what is actually the speech act of the emiting actor (SEA69). This is feasible by categorizing statement contexts and by asking every actor to qualify, with approriate and explicit hints, his communication: he makes a Communicational Data, which qualitatively augments the denotative meaning of his message (CD97).

Moreover, the system must be able to interpret the set of Communicational Data in order to exhibit the functional and organisational structures, to make interpretations and to send the synthesis to the different actors.

By necessity, the system's communicational layer must then achieve an interpretation process, in the sense of the Peirce's triadic semiotics (ECO93). The system then generates meaning for itelf (expressed by the structures of signs it manipulates) during an auto-adaptation process relatively to the outside world from which it takes into consideration the communicationnal data: it links the object (the communicational data) to the sign (the internal representational structures) according to a certain self-intentionality (Figure ). The architecture and implementation of such a communicational adaptive layer will be based on dynamic multiagent systems, reifying the different semantic categories used in the communicational data and producing an artificial interpretation process.

Ontology of the communicated knowledge

The interpretation system of the communicated knowledge must operate on the exchanged messages between professionals, by representating the meaning of the situation to be cooperatively managed. In other words by representating the structure and the organisation of the perception of this situation, during its evolution. For this, it is first necessary to specify the typology of the messages and to define their standard form. This task is to be done for each of the complex domains we want the system to operate on. The following applies to emergency situations (natural or industrial disasters for example). We identified five types of statements constituting the messages:

1  the type "stated fact",

2  the type "asked question",

3  the type "affirmed judgment",

4  the type "given order",

5  the type "annotated action".

Each message is made of one or more statements refering to these five types. The cognitive content of the classified messages according to the typed statements is grouped in cognitive categories relatively to the emergency situation domain. We make the assumption that the exchanged informations have their place in those four cognitive categories:

- spacial knowledge,

- temporal knowledge,

- organisational knowledge,

- knowledge on actors' subjective perception of the situation.

Indeed, in all messages we can identify, distinguish typical statements. For instance, one may give an order, ask a question, emit a judgment or annotate one's actions. As a consequence, we are led to a categorization of the actor's message.

These categories are declined according to various typical characters or semantic traits (JAC83). The last category seems to be fundamental since it deals with opinions that belong to each actor, which might be misunderstood or badly resented by others.

This categorization of the speech allows the structuration of the message into statements that belong to the five types, each statement being made of various semantic traits relatively to the four previous categories (JAC83). To each semantic trait, it will be possible to link an estimation of its importance, on a subjective scale set by the actor who generates the message. The communication language is thus constrained, which is in adequation with the technical aspects of the problem we are dealing with.

The standardisation of exchanged messages: Communicational Data

Each communication sent by an actor is made of a statement with an objective characteristic and different qualifications expressing the pragmatics and the interpretation of the event that lead to the emission of this message. The language constituants belong to the previously defined categories. This communicational form, extending the message to elements of its pragmatics, constitutes what we named a Communicational Data (CD97).

A communicational data is a list of triplets, structurable into a graph, each triplet being composed of the following entities:

 - the current object of attention,

 - the qualifications of the object,

 - the intensity of these qualifications.

Here is a typical example of a communicational data. It is taken from a real situation in France at the beginning of spring holidays with heavily loaded roads.

The Prefet ( roughly the equivalent of a Governor ) orders to "immediately take into account the risks of important icing rains because the neighbouring department has serious problems". The corresponding Communicational Data is divided into two parts: the objective part which is filled with the sender's message and the interpreted part which is built thereafter by the interpretation system. agents. The interpreted part is the result of the attempts made by the interpretation system to qualify the message: it tries to attach, to enhance the objective part with facts that are implied by the sender's role and the content (stated or not) of his message (NUT91). According to the categories we defined before, the message is interpreted as a communicational data, here is the objective part:

1  Identification

 text: the text with its emission date

 sender: Prefet

 recipient: all civil protection services of the department

2  Given order

 type: stated order on a potentially serious situation

 subject: risks of important icing rains

 qualification of the subject: judgment on situation = important, judgment's intensity = very high

Now, the interpretation process tries to attach some knowledge to the objective part:

1  Spatial knowledge

 scale: department

 accuracy: fuzzy

2  Temporal knowledge

 acknowledgment date: in the coming hours

3  Organisational knowledge

 prerogatives: valid order because it is the Prefet

 order justification: explicit cause

 justification accuracy: fuzzy

The two parts thus can be seen as slots of a frame: they constitute the actual Communicational data.

The representation of the interpretation process of Communicational Data

In the model of DIS (Figure ), each decisional operator is symbolized by a node of the communication network. He has access to the network via an interface that allow him to access various local databases or Geographical DIS and, particularly, to communicate with other operators.

The communicational data are explicit knowledge concerning a structuring situation. These are fuzzy knowledge chunks, very contingent on the moment, the place and the operator who states them. They must lead in an incremential manner to the correct global representation of the situation. It is necessary, in order to preserve the dynamics and the complexity of this kind of knowledge, to have an interpretation system of the Communicational Data. The communication medium will thus assume an interpretating function of the communicational data by dynamically self-adapating to their different cognitive aspects, achieving the augmentation of the messages.

To achieve this goal, the construction of the communicational data is essentially done by agents that reify the various semantic traits, in order to enhance them by introducing some elements of their pragmatics and to preserve the action characteristic of the communicational data.

The interpretation process of Communicational Data is represented by a set of four multiagent systems somehow encapsulating all generation of communicational data in order to interpret it (ASH60). These agent systems are to exhibit the meaning of the set of the communicational data and to produce contextual lectures: synthesis, evaluations, positionning, pertinences, interests, values ... What the system has to do is to understand the meaning of the exchanged information relative to a situation described parts by parts, taking into account the pre-established plans, historical data and knowledge on operators themselves ( typical actor's behavior, typical actors' points of view).

The interpretation process: an agent based system architecture

The system architecture we propose for helping users to achieve cooperative decising making is an adaptive MAS (HC93). We propose a definition for such adaptive systems:

An adaptive system has an organisation (a structure which evolves with time) it modifies according to its own perception (a motivated interpretation and conception) of its environment and by taking into account the current state of its organisation (it is actually self-referring).

The interpretation process is modelized as a complex agent system (TZA95). The result of this interpretation process - the global meaning of the situation - is an organisational emergence in the whole agent system. We define the emergence in adaptive systems as follows:

It is the acknowledgment at a certain time, of a global stabilization aspect in the organisation's modifying process.

The whole system can be decomposed into four dynamic MAS. The first one is the "Aspectual MAS". It is composed of various aspectual agents whose functions is to reify the semantic traits in the system. These agents actually represent the previous knowledge categories on the domain. The second one is the "Morphology MAS". It is made of agents whose task is to classify through some kind of main component analysis; they have to exhibit significant organisational structures in the aspectual MAS, it is also made of chreod agents that actually are some kind of aggregations of morphology agents and whose task are to generate the "care about" in the system (JOR89).

The third one is the "Analysis MAS", it is composed of analysis agents. These agents represent the rational reasoning of the system, they assume many forms: expert systems, neural networks... In other word, they consider the system in its temporality. Their role is to compare the previously conceived situation with the current situation and thus to point out the differences and/or the adequations in order to express the global meaning of the situation (WEY95).

Another MAS is spread throughout the system. It is the decisional MAS. Its role is to influence the emergence of the meaning about the situation. It does so with polymorph agents that can assume the form of the agents that constitute the four other MAS.


A system implementing the previously exposed notions has been implemented in SmallTalk. It is dedicated to emergency situation management and will be validated by the professionals. Communication nodes being installed in the various institutions will be linked by a high bandwidth network.


Concerning the DIS, the approach based on the communication focused on the representation of exchanged knowledge perception is the way to go beyond the weaknesses of symbolic accesses inherent to classical Information Systems. It allows the mutation of lonesome access into dialogic loops extented to many simultaneous users. Thus preserving the different points of view and leading to a common negociated decision making.


[ASH 60] Ashby W.R., Design for a brain: the origin of adaptive behavior, Chapman & Hall, London, 1960.

[CAR 97] A. Cardon, A Multi-agent Model for Co-operative Communications in Crisis Management System: the Act of Communication, 7th European-Japanese Conference on Information Modelling and Knowledge Bases, p. 111-123, Toulouse, 27-30 Mai 1997.

[CD 97] Cardon A., Durand S., A Model of Crisis Management System Including Mental Representations, AAAI Spring Symposium, Stanford University, Californie, USA, 23-26 mars 1997.

[ECO 93] H. Eco, Sémiotique et philosophie du langage, PUF, 1993.

[HC 93] Hayes-Roth B. , Collinot A., A Satisficing Cycle for Real-Time Reasoning in Intelligent Agents, Expert Systems with Applications, 7, p. 31-42, 1993.

[JAC 83] Jackendoff R., Semantics and Cognition, Cambridge, M.I.T. Press, 1983.

[JOR 89] P. Jorion, Principes des systèmes intelligents, Masson, 1989.

[NUT 91] Nutin J. Théorie de la motivation humaine, PUF 1991.

[SEA 69] Searle J. R., Speechs Acts, Cambridge University Press, 1969

[SFE 92] Sfez L., Critique de la décision, Presses de la fondation nationale des sciences politiques, 1992.

[TZA 95] Tzafestas E., Vers une systémique des agents autonomes: des cellules, des motivations et des perturbations, Thèse de doctorat, Université de Paris VI, 1995.

[WEY95] Weyrauch R., Building conscious artifacts. In Consciousness, Distinction and Reflexion. Bibliopolis, 1995.

[WF 86] T. Winograd, F. Flores, Understanding Computers and Cognition: A new Foundation for Design, New-Jersey, Abley Press, 1986.