Sharing Knowledge in the Net through a Collaborative System

Ruth Cobos, Xavier Alamán and Jose A. Esquivel

Departamento de Ingeniería Informática, Universidad Autónoma de Madrid. Spain.

{Ruth.Cobos|Xavier.Alaman|Jose.A.Esquivel}@ii.uam.es

 

ABSTRACT

In this paper we present a groupware system for sharing and structuring knowledge in a collaborative, distributed and non-supervised way, The system, KnowCat, stands for "Knowledge Catalyser" and its purpose is enabling the crystallisation of collective knowledge as the result of user interactions.

One application of the system is the generation of quality educational materials as the automatic result of student interactions with the materials. Another application of the system is the generation and the maintenance of collective knowledge of a research group.

KEYWORDS: Knowledge Management, Groupware, Virtual Communities, Knowledge Crystallisation, Collaborative Systems.

 

1. INTRODUCTION

The necessity of systems that implement models for knowledge management is primarily related to business strategic solutions [5][7]. A comparison between different tools shows that most of these tools are oriented to the management of the corporative knowledge [3] . Although, these tools can be used in other areas, such as educational and research activities, we think that it is necessary to construct tools specially designed for supporting academic communities.

In this paper we presented a Web-based system called KnowCat, that stands for "Knowledge Catalyser". This system has two applications: one of them is the generation of quality educational materials as the automatic result of student interactions with the materials; another application of the system is the generation and the maintenance of collective knowledge of a research group.

 

2. DESCRIPTION OF THE SYSTEM

KnowCat, it is a distributed non-supervised system for structuring knowledge and its purpose is to enable the crystallisation of collective knowledge as the result of user interactions, so it may be classified as a groupware tool.

One of the main goals of the system is to allow the building of places where we can find the relevant knowledge about an area or topic. With the word "relevant" we are referring to quality knowledge about the area. These places are called "KnowCat sites" or KnowCat nodes.

KnowCat is a web-based client-server system, which allows asynchronous work. It is characterised by its portability, i.e., it can run in any platform; adaptability, i.e., it can be adapted to the user needs; and scalability, i.e., KnowCat nodes may be combined to form higher order nodes.

The system is based on the concept of "Knowledge crystallisation" [1], supported by virtual communities of experts. The main goal of the tool is to capture "established" knowledge about a given topic without the need of an editor managing the task.

2.1 KNOWLEDGE IN THE SYSTEM

KnowCat does not deal with any type of knowledge, but is specialised in that kind of knowledge that is "explicit" [2] - it can be transmitted from one person to other through documents, images and other elements -, and is reasonably stable in time. Examples of this type of knowledge are currently implemented in the form of encyclopaedias or reference books. KnowCat intends to encourage the co-operative creation and organisation of this type of knowledge in the Web.

How is the knowledge stored in the system? KnowCat organises the knowledge in the form of a tree structure. The root of this tree is the main topic of the area in which we are interested. Each node of the knowledge tree represents a topic and contains two items:

KnowCat provides a "knowledge workspace" (see figure 1) where users can add contributions and propose changes in the structure of the knowledge tree. Usually this workspace shows the knowledge tree with the best topic descriptions (contents) displayed ordered by their crystallisation degree. In each moment the descriptions of a topic are competing with each other for being considered as the "established" or the best description of the topic.

 

Figure 1. An example screen of the system.

KnowCat also provides with a voting system, where users can collaborate with their opinions about the knowledge. Any person may contribute with new knowledge to a node, but only the members of the virtual community can give their opinion about knowledge elements in the node.

And there are other auxiliary services in the system: users can see and change their personal information; reports may be obtained about system activity; users may subscribe to events and the system will notify them by e-mail messages; users may communicate with their virtual communities. All the characteristics of KnowCat are easily configurable by means of parameters.

2.2 KNOWLEDGE CRYSTALLISATION

The tree structure and the contents (descriptions of each topic) are knowledge elements. The use of these elements of knowledge, the opinions of other users about them and the time they have endured in the system, decide whether these elements of knowledge are useful, in which case they will stay longer in the system, or if they are useless, in which case they will eventually disappear from the system. This process is called Knowledge Crystallisation.

However, as important as the number of users or their opinions is the "quality" of these users. We would like to give more credibility to the opinions of experts than to the opinion of occasional users. KnowCat tries to establish categories of users by the same means than the scientific community establishes its member's credibility: taking into account past contributions. So the system deals with "virtual communities".

A virtual community [6] is a group of users that are considered experts in one or more related topics. We are interested in opinions from experts because they should have more impact than opinions from novices or occasional users. Furthermore, as Collins, Mulholland and Watt say, "learning about a topic becomes synonymous with learning to be a member of a community of people who are experts on that topic" [4] .

Virtual communities of experts are constructed in terms of the knowledge tree. For each node (topic), the community of experts in this topic is composed of the authors of the crystallised documents on the topic, on the parent of the topic, on any of the children of the topic or on any of the sisters of the topic. There is a virtual community for each node of the tree, and any successful author usually belongs to several related communities.

The mechanism of knowledge crystallisation is based on these virtual communities. When one of your contributions crystallises, you receive a certain amount of "votes" that you may apply for the crystallisation of other documents (of other authors) in the virtual community where your crystallised document is located. As in the previous phase the descriptions may be added both by experts and collaborators; in fact all users start using the system as collaborators and when a document of an user crystallises s/he becomes an expert in the topic where the document is located.

The other aspect of knowledge crystallisation is the evolution of the structure of the knowledge tree. If a member of a virtual community proposes to add a new subject to a topic, remove a subject from a topic or move a subject from one topic to another topic, then a minimum quorum of positive votes from other members of the community will be necessary for the change to be made.

2.3 KNOWLEDGE EVOLUTION

The system deals with knowledge in evolution, because not only the tree structure but also the contents (descriptions of each topic) are elements that have been contributed by the users of the system, and their lifetime in the system depends on the patterns of their usage.

When we start an area of knowledge we only have a root node with the main topic. Probably, there will not be enough people and interactions to make the crystallisation process credible. In relation with this bootstrapping problem, virtual communities have also proven to be handful. Virtual communities behave in a different way when they are just beginning, and also (possibly) in their late days. KnowCat proposes a maturation process that involves several phases. Next figure shows this evolution.

Figure 2. Knowledge evolution in phases.

At the beginning stages the system works in a "supervised" mode. During this supervised phase there will be a steering committee in charge of proposing knowledge structures (initial refinements of the root node) and voting for them. The members of the steering committee are defined in the moment of creation of the area; new members can be added by consensus of the current members.

In this phase, descriptions (documents about some topic) may be added to the system both by the members of the steering committee and by other users that are considered as collaborators. However, only the members of the steering committee have the complete capability of voting on the documents, and thus in deciding which documents crystallise. Collaborators may have limited capability of voting, if the steering committee decides so.

Eventually, the steering committee may decide to advance the area of knowledge to the "active" mode, possibly when a critical mass of participants and interactions is achieved. In this moment there should be a single tree structure for the area, decided by consensus. Then the steering committee is dissolved and the subsequent crystallisation of the knowledge is based on virtual communities.

When one of your contributions crystallises, you receive a certain amount of "votes" that you may apply for the crystallisation of other documents (of other authors) in the virtual community where your crystallised document is located. As in the previous phase the descriptions may be added both by experts and collaborators; in fact all users start using the system as collaborators and when a document of an user crystallises s/he becomes an expert in the topic where the document is located.

The other aspect of knowledge crystallisation is the evolution of the structure of the knowledge tree. If a member of a virtual community proposes to add a new subject to a topic, remove a subject from a topic or move a subject from one topic to another topic, then a minimum quorum of positive votes from other members of the community will be necessary for the change to be made.

Finally, an active community may reach the "Stable" phase. Many of the community members are not active any longer, so different rules should be applied to ensure some continuity of the crystallisation. Changes are rare, and most of the activity is consultation. Few new contributions arrive, and they will have much more difficulties to crystallise comparing to the previous phase. However, if activity raises to a minimum again, the node may switch to "Active" status, and engage in a new crystallisation phase.

 

3. SOME APPLICATIONS OF THE SYSTEM

The main goal of the system is to capture "established" knowledge about an area or topic without the need of an editor managing the task, constructing an active knowledge repository of quality that will be improved over the years.

If the system is used by students (a group of students and the teacher) that are enrolled in a formal course, the system may be used as a repository of the set of topics and contents of this course. This repository will be constructed and improved by the different groups of students over the years.

We have tested KnowCat during the last three years with several communities at Universidad Autonoma de Madrid, they can be consulted in the web address http://www.ii.uam.es/~rcobos/investigacion/knowcat/eng/fKC.htm:

3.1 COMMUNITY ABOUT OPERATING SYSTEMS

In the first experience, we started by creating a KnowCat site about "Operating Systems", containing twelve topics (it was created by the course instructor). We wanted to check the hypothesis that when you get enough documents and enough votes from "knowledgeable" peers, the result is a reasonable description of the topic.

The students were grouped in communities, each related to one of the topics. Moreover, each student had to produce a small paper (fragment or description) on an assigned topic and vote for the three best papers in that same topic. The instructor graded papers independently, and this grading was used to check the adequacy of the voting system to capture the quality of the paper.

At the end of the first year, in 11 of the 12 topics the votes of the students converged to a small set of documents. There was a remarkable consensus. For most topics the two most popular documents collected 50% of the total votes. Furthermore, in 10 out of the 12 topics at least two of the three documents selected as "the three best documents" by the course instructor were also selected by the students.

During the second and third years, new groups of students used this knowledge tree, where KnowCat periodically removed the documents with a low crystallisation degree. Students added new documents about the twelve topics, and they scored both the new ones and the "veteran" ones through the tool voting mechanism.

The result of the collaborative work during these years is that 50 % of the topics of the initial tree structure - it was created by the course instructor - have in their first ranking position documents that have been added during different years. This shows some evidence that the knowledge in the system is in evolution and is possible for a document that arrives later to crystallise and achieve the first positions of the rank.

However, in the other 50 % of topics the description selected as the best during the first year obtained so high a crystallisation degree that descriptions added in the following years were not able to reach it. It is justifiable that about 10 % of these topics may contain the best description since the first year. However the other 40 % needs further explanation. If participation is scarce in a given topic, then it may be expected that veteran descriptions preserve their high crystallisation degree. This has happened in 20 % of these topics. But, the behaviour of the other 20% of topics shows the need of some rearrangement in the crystallisation algorithm.

3.2 COMMUNITY ABOUT UNCERTAIN REASONING

For the second experience we created a KnowCat root node labelled "Uncertain Reasoning" with no initial topics. The students in this experience were considered as a group of researchers and they worked in the supervised phase because we wanted to test the crystallisation of the knowledge tree structure. The aim was to check the feasibility of a group of researchers making a good structure for the topic by using our proposed voting mechanism.

At the end of the year there were 14 topics in the KnowCat node, which were distributed over a tree four levels deep. The course instructor's opinion was that the structure of the knowledge tree was quite appropriate for the topic.

The second year, these students changed to the active mode and some virtual communities appeared about certain topics. They contributed with new documents, new structures for the topics and opinions about these issues. Student participation was uniformly distributed in time. However, there were several periods in which the participation was markedly greater, and these coincided with times in which students engaged in discussions by means of the messaging service. The experiment continued for one more year with a new group of students.

The structure has been improved along these years successfully. The number of topics in the current structure is twice the quantity of the initial number of topics of the structure that was created the first year. Now the knowledge tree is five levels deep. In the opinion of the instructor, the resulting tree contains a credible overview of the topics of the system, and the crystallised papers show a high quality.

3.3 THE OTHER COMMUNITIES

The third and fourth experiences have been carried out during the last year. In both cases, the course instructor proposed an initial tree structure without documents, and they worked on the active mode. The aim was to check the document crystallisation and the creation of virtual communities of experts.

In the third experiment each student chose a topic and wrote a document about it. They had an unlimited number of votes for voting other documents. At the end of the course, some discrimination was achieved through the crystallisation mechanism: although most of the documents of each topic had a similar crystallisation degree, two small groups of documents could be identified, one in the top and the other in the bottom of the ranking. This classification coincided with the course instructor opinion: "Although I am not sure that the ranking obtained over the whole set of documents reflects the relative quality of each one, it is certainly true that the top documents are the best ones in my opinion, as well as the bottom ones are the worse."

In the fourth experience, the crystallisation mechanism was used by the instructor in a particular way. The root node and their subjects, that were called group1, group2 and so on, were all about the same topic. Students were distributed into these groups because the goal of the instructor was to obtain a good and definite document on the area as the result of the work of students in small groups first, and then among all of them.

Students of this experience made a document about the area and put it in her/his assigned topic. Then they voted for the best description of the topic where her/his document was located and finally they voted to the best description of the area, that could be in another topic, i.e. they voted to the best document between the best one in each subject. The classification of documents is good in each topic, and the two more voted documents among all the documents in the system were the best ones in the opinion of the course instructor.

 

4. CONCLUSIONS AND FUTURE WORK

This paper describes a Web-based system called KnowCat that allows us sharing, evaluating and structuring our community knowledge. This is possible through the knowledge crystallisation process, supported by virtual communities of experts.

KnowCat enables the building of Web sites where relevant knowledge about an area or topic can be found, without the need of an editor for managing the task. The system has being tested with several communities of undergraduate and graduate students at Universidad Autonoma de Madrid. The experiments have shown evidence that the system is useful for motivating communities in sharing their knowledge and in incrementally constructing a knowledge repository of quality.

The obtained results through the experiences carried out provide some support for three of the hypothesis that underlies the design of KnowCat. Firstly, if a set of "knowledgeable" people engage in a reasonable interaction with our system, the result converges to some consensus; secondly, this consensus is closely related to some objective measure of "quality" of the contributions; and finally, the knowledge classification through a tree structure has been exposed as a suitable approach for managing and organising the knowledge.

Currently, the system is available for the RedIRIS's user distribution list (RedIRIS is the main group of the National Programme of Applications and Telematic Services in Spain http://www.rediris.es/). Thus, the system is being used by research groups to create active repositories where they expose their common research. In this way, we will be able to obtain conclusions from investigation uses.

We have identified several issues opened in the context of our approach:

This work has been partially funded by the Spanish National Plan of R+D, project numbers TIC98-0247-C02-02 and TIC2001-0685-C02-01.

 

5. REFERENCES

[1] X. Alaman, R. Cobos, KnowCat: a Web Application for Knowledge Organization, in: Proc. of the World-Wide Web and Conceptual Modeling (WWWCM'99), Lecture Notes in Computer Science 1727, P.P Chen et al (Eds). Paris, November, 1999. pp. 348-359.

[2] V. Allee, The Knowledge Evolution. Butterworth Heinemann, Boston. 1997.

[3] R. Cobos, J.A. Esquivel, X. Alamán. IT Tools for Knowledge Management: A Study of the Current Situation. Journal of Novática and Informatik/Informatique, special issue on Knowledge Management, Vol. III, no 1, February 2002.

[4] T. Collins, T. Mulholland, S. Watt. Using genre to support active participation in learning communities, in: Proc of European Perspectives on Computer-Supported Collaborative Learning (CSCL' 2001). Maastricht, the Netherlands, March 22-24, 2001. pp. 156-164.

[5] J.M. Firestone, Knowledge Management Process Methodology: An Overview, Knowledge ad Innovation: Journal of the KMCI, vol. 1, No. 2, January 15, 2001

[6] W. Hill, L. Stead, M. Rosenstein, G. Furnas, Recommending and Evaluating Choices in a Virtual Community of Use, in: Proc CHI95, ACM Press, New York, pp. 194-201.

[7] Y. Malthotra, From Information Management to Knowledge Management: Beyond the 'Hi-Tech Hidebound' Systems. In K. Srikantaiah & M.E.D Koenig (Eds.), Knowledge Management for the Information Professional. Medford, N.J.:Information today Inc. pp.37-61

 

6. ABOUT AUTHORS

Ruth Cobos. In 1997 graduated from Universidad Autónoma de Madrid (Spain) on Computer Science Engineering. Currently, she is a PhD student and an Associate Professor of Computer Sscience and Intelligence at the School of Computer Science and Engineering of Universidad Autónoma de Madrid. She is a researcher in the KnowCat project and her research interests include knowledge management systems, Web-based groupware systems, teaching and learning.

Xavier Alamán holds a position as tenured Associate Professor in Computer Science and Artificial Intelligence at the School of Computer Science and Engineering of Universidad Autónoma de Madrid. He obtained a MSc degree in Physics (Univ. Complutense de Madrid 1985), a MSc degree in Computer Science (Univ. Politecnica de Madrid 1987), a MSc degree in Artificial Intelligence (Univ. Of California Los Angeles 1989) and a PhD in Computer Science (Univ. Complutense de Madrid 1993). His research interests include knowledge management tools and groupware systems.

Jose A. Esquivel. In 1996 graduated from Intituto Tecnológico de Zacatecas (Mexico) on Computer Science Engineering. Currently, he is a PhD student at School of Computer Science and Engineering of Universidad Autónoma de Madrid. He is an auxiliary investigator in the KnowCat project and his research interests include particularly the knowledge structure in the knowledge management systems and Web-based groupware systems.